JP2007017282A - Biological data processor, biological data processing method, learning device, learning control method, program and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To learn the judgment of the adoption or non-adoption of a spot based on reliability. <P>SOLUTION: Developed profile data 744 including feature quantity data is provided to SVM 91 and the images of the respective spots of a developed profile image 741 are displayed on a display part. The indication related to whether the respective spots are set to adoption spots 742 or non-adoption spots 743 is received as the teacher data of a learning mode from a user. The non-adoption spots 743 are the spots judged not to be adopted because a large deburi or the like detracting the reliability of a measured result and the adoption spots 742 are the spots wherein deburi or the like is present and effective hybridization is produced. The SVM 91 learns the relation of the teacher data indicated by the user with the feature quantity data at every spot of the developed profile data 744 to store the learning result in a spot removing pattern database 92. This invention can be adapted to a biological data processor. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a biological information processing apparatus, a biological information processing method, a learning apparatus, a learning control method, a program, and a recording medium. The present invention relates to a biological information processing apparatus, a biological information processing method, a learning apparatus, a learning control method, a program, and a recording medium that can be measured in a continuous manner.

  In recent years, a DNA (deoxyribonucleic acid) chip or a DNA microarray (hereinafter simply referred to as a DNA chip when it is not necessary to distinguish between them) has been put into practical use. A DNA chip is a DNA chip in which various and many DNA oligo chains are integrated and immobilized on a substrate surface as nucleic acids for detection. Comprehensive analysis of gene expression in collected cells by detecting hybridization between probes immobilized on spots on the substrate surface and targets in samples collected from cells using a DNA chip Can do.

For example, Patent Documents 1 to 3 are prior art documents related to a method for analyzing and correcting a gene expression level obtained by a DNA chip or the like.
JP 2002-71688 A JP 2002-267668 Gazette JP 2003-28862 A

  Conventionally, when the amount of hybridization to a target probe in a DNA chip is measured, a fluorescent image of a spot associated with hybridization of the DNA chip is displayed on the display. The amount of hybridization is determined based on the average value of the fluorescence intensity in the spot region.

  By the way, when a foreign substance exists on the DNA chip, it becomes difficult to accurately measure the fluorescence intensity necessary to detect the hybridization between the probe and the target. In the conventional method, for example, a person performing the measurement visually confirms whether or not a foreign substance exists on the DNA chip. For this reason, various analysis processes using the DNA chip not only take time, but also the measurement results vary depending on the person performing the measurement.

  The present invention has been made in view of such a situation, and it is possible to prevent a measurement result from being varied by a person who performs measurement and to perform accurate measurement without complicating the configuration. .

  The biological information processing apparatus according to the first aspect of the present invention includes an image input unit that receives input of image information of a reaction region, and a feature amount extraction unit that extracts a feature amount of image information of the reaction region input by the image input unit. Storage means for storing determination criterion information that is a determination criterion for the presence or absence of reliability when the image information of the reaction area is used for measurement, and a feature amount extraction means based on the determination criterion information stored by the storage means Determining means for determining the presence or absence of reliability when the image information of the reaction region having the feature amount extracted by the measurement is used for measurement, and the first biological material and the first biological substance in the reaction region based on the determination result by the determination unit And calculating means for calculating reactivity information representing the state of biological reaction with the biological material of 2 and the criterion information is obtained by using image information for learning and image information of the reaction region for measurement. The presence or absence of reliability And to information, which is obtained by learning the correlation between the feature quantity of the image information of the reaction region, the feature quantity is a value having a correlation to the reliability of the image information of the reaction region.

  The criterion information can be obtained by SVM.

  The feature amount extracted by the feature amount extraction unit may include the fluorescence intensity obtained when the reaction region is irradiated with excitation light.

  The feature amount extracted by the feature amount extraction means may include the fluorescence intensity obtained when the region around the reaction region is irradiated with excitation light.

  The feature amount extracted by the feature amount extraction unit may include a value corresponding to the area of the image corresponding to the substance that is an obstacle to perform the measurement included in the reaction region.

  The feature amount extracted by the feature amount extraction unit may include a value corresponding to the biological reaction amount of the first biological material and the second biological material in the reaction region.

  The feature amount extracted by the feature amount extraction means may include information indicating the ease of biological reaction between the first biological material and the second biological material in the reaction region.

  Information indicating whether or not the reaction area image information is reliable as teacher information by a display means for displaying the learning image information input by the image input means and a user referring to the image information displayed by the display means And a learning means for obtaining criterion information by learning the correlation between the teacher information input by the teacher information input means and the feature quantity extracted by the feature quantity extraction means. Further, the storage unit can store the determination criterion information obtained by the learning unit.

  The first biological material and the second biological material may be genes having base sequences complementary to each other or materials derived therefrom, and the reactivity information includes the first biological material and the second biological material. Information on the hybridization of the biological material.

  The information of hybridization can be information uniquely determined based on a function from the fluorescence intensity obtained by hybridizing the first biological material and the second biological material.

  A biological information processing method or program according to a first aspect of the present invention includes an image input control step for controlling input of image information in a reaction region, and image information in a reaction region whose input is controlled by processing in the image input control step. A feature amount extraction step for extracting a feature amount, an acquisition control step for controlling the acquisition of determination criterion information that is a determination criterion for the presence or absence of reliability when the image information of the reaction region is used for measurement, and processing of the acquisition control step A determination step for determining the presence or absence of reliability when the image information of the reaction region having the feature amount extracted by the processing of the feature amount extraction step is used for the measurement based on the determination reference information whose acquisition is controlled by: Calculation for calculating reactivity information representing the state of biological reaction between the first biological material and the second biological material in the reaction region based on the determination result obtained by the determination step. The criterion information includes learning information, information indicating the presence or absence of reliability when the image information of the reaction region is used for measurement, and the feature amount of the image information of the reaction region. It is obtained by learning the correlation, and the feature amount is a value having a correlation with the reliability of the image information in the reaction region.

  In the first aspect of the present invention, when the image information of the reaction area is input, the feature amount of the image information of the reaction area is extracted, and the presence / absence of reliability when the image information of the reaction area is used for measurement is determined. Based on the determination criterion information as a reference, the presence or absence of reliability is determined when the image information of the reaction region having the extracted feature amount is used for measurement, and the first living body in the reaction region is determined based on the determination result. Reactivity information representing the state of biological reaction between the substance and the second biological substance is calculated. At this time, the determination criterion information uses the learning image information to correlate information indicating the presence or absence of reliability when the image information of the reaction region is used for measurement and the feature amount of the image information of the reaction region. The feature amount is obtained by learning, and the feature amount is a value having a correlation with the reliability of the image information in the reaction region.

  The learning device according to the second aspect of the present invention includes an image input unit that receives input of image information of a reaction region, a display unit that displays image information input by the image input unit, and image information displayed by the display unit. There is a correlation with the reliability of the image information in the reaction area for each reaction area, and the teacher information input means that receives the information indicating the presence or absence of the reliability of the image information in the reaction area by the user who referred to Feature amount extraction means for extracting feature amounts, and learning means for learning the correlation of the feature amounts extracted by the feature amount extraction means with respect to the information indicating the presence or absence of reliability of the reaction region input by the teacher information input means Prepare.

  The feature amount extracted by the feature amount extraction unit may include the fluorescence intensity obtained when the reaction region is irradiated with excitation light.

  The feature amount extracted by the feature amount extraction means may include the fluorescence intensity obtained when the region around the reaction region is irradiated with excitation light.

  The feature amount extracted by the feature amount extraction unit may include a value corresponding to the area of the image corresponding to the substance that is an obstacle to perform the measurement included in the reaction region.

  The feature amount extracted by the feature amount extraction unit may include a value corresponding to the biological reaction amount of the first biological material and the second biological material in the reaction region.

  The feature amount extracted by the feature amount extraction means may include information indicating the ease of biological reaction between the first biological material and the second biological material in the reaction region.

  The learning means can perform learning processing by SVM.

  The image input means can be input with a plurality of pieces of image information on the same substrate obtained when the intensity of the excitation light irradiated to the substrate is different in order to obtain image information. Can extract feature amounts for each reaction region of a plurality of pieces of image information obtained when the intensity of excitation light is different, and is displayed on the teacher information input means by the display means. The user who refers to the image information can receive one class of teacher information for one reaction area, and the learning means includes one class of teacher information input by the teacher information input means. On the other hand, it is possible to learn the correlation between the respective feature amounts obtained when the intensity of the excitation light extracted by the feature amount extraction means is different.

  A learning control method or program according to a second aspect of the present invention controls an image input control step for controlling input of image information in a reaction region, and display of image information whose input is controlled by processing of the image input control step. Teacher information input for controlling the input of information indicating the presence or absence of reliability of the image information of the reaction region as the teacher information by the user who refers to the display control step and the image information whose display is controlled by the processing of the display control step The reliability of the reaction region whose input is controlled by the control step, the feature amount extraction step for extracting the feature amount correlated with the reliability of the image information of the reaction region for each reaction region, and the processing of the teacher information input control step A learning step for learning the correlation of the feature quantity extracted by the process of the feature quantity extraction step with respect to the information indicating the presence or absence.

  In the second aspect of the present invention, the image information of the reaction area is input, the image information is displayed, and the reliability of the image information of the reaction area as the teacher information by the user who refers to the displayed image information. Information indicating presence / absence is input, and for each reaction region, a feature amount having a correlation with the reliability of the image information in the reaction region is extracted, and the correlation of the extracted feature amount with respect to information indicating the presence / absence of the reliability of the reaction region is extracted. Is learned.

  As described above, according to the first aspect of the present invention, the state of the biological reaction between the first biological material and the second biological material that undergoes a biological reaction with respect to the first biological material is measured. In particular, it is possible to determine the presence / absence of reliability of the image information in the reaction region using the determination reference information obtained as a result of learning.

  In addition, according to the second aspect of the present invention, it is possible to perform learning for determining the presence or absence of reliability of image information, and in particular, using a feature quantity having a correlation with the reliability of image information in a reaction region. Thus, it is possible to learn the correlation of the extracted feature quantity with respect to the information indicating the presence or absence of the reliability of the reaction region.

  Embodiments of the present invention will be described below. Correspondences between the configuration requirements of the present invention and the embodiments described in the detailed description of the present invention are exemplified as follows. This description is to confirm that the embodiments supporting the present invention are described in the detailed description of the invention. Therefore, although there are embodiments that are described in the detailed description of the invention but are not described here as embodiments corresponding to the constituent elements of the present invention, It does not mean that the embodiment does not correspond to the configuration requirements. Conversely, even if an embodiment is described here as corresponding to a configuration requirement, that means that the embodiment does not correspond to a configuration requirement other than the configuration requirement. It's not something to do.

  The biological information processing apparatus according to the first aspect of the present invention (for example, the biological information processing apparatus 1 in FIG. 1) is used for a first biological material (for example, a probe) fixed to a reaction region provided on a substrate. A biological information processing apparatus for measuring a biological reaction state between a biological material) and a second biological material that reacts biologically with the first biological material (for example, a biological material used as a target), Image input means (for example, the excitation light intensity estimation unit 81 in FIG. 1) that receives the input of the image information, and feature amount extraction means (for example, the feature quantity extraction means that extracts the feature amount of the image information of the reaction region input by the image input means) A feature amount extraction unit 86 in FIG. 1 and storage means (for example, spot reliability determination information) for storing determination criterion information (for example, spot reliability determination information) that is a determination criterion for the presence or absence of reliability when the image information of the reaction region is used for measurement. For example, FIG. Based on the spot removal pattern database 92) and the criterion information stored in the storage means, the presence or absence of reliability when the image information of the reaction region having the feature quantity extracted by the feature quantity extraction means is used for the measurement. Based on the determination means (for example, SVM91 in FIG. 1) for determination and the determination result by the determination means, the reactivity information representing the state of the biological reaction between the first biological material and the second biological material in the reaction region is calculated. And the presence or absence of reliability when the image information for the reaction region is used for the measurement using the learning image information. Is obtained by learning the correlation between the image information and the feature amount of the image information in the reaction region, and the feature amount is a value having a correlation with the reliability of the image information in the reaction region.

  In the biological information processing apparatus according to the first aspect of the present invention, the feature amount extracted by the feature amount extraction unit includes an image (for example, debris) included in the reaction region and corresponding to a substance that hinders measurement. ) (For example, the number of pixels in the spot area).

  In the biological information processing apparatus according to the first aspect of the present invention, the feature amount extracted by the feature amount extraction unit is a value corresponding to the biological reaction amount of the first biological material and the second biological material in the reaction region ( For example, a hybridization value) is included.

  In the biological information processing apparatus according to the first aspect of the present invention, the feature amount extracted by the feature amount extraction unit includes the ease of biological reaction between the first biological material and the second biological material in the reaction region. Information (eg, a probe-target bond strength matrix) is included.

  The biological information processing apparatus according to the first aspect of the present invention includes a display unit (for example, the display unit 92 in FIG. 1) for displaying learning image information input by the image input unit, and an image displayed by the display unit. Input by a user who has referred to the information as teacher information by teacher information input means (for example, the input unit 29B in FIG. 1) that receives information indicating the presence or absence of reliability of the image information of the reaction area, and by teacher information input means Learning means (for example, SVM 91 in FIG. 1) for obtaining determination criterion information by learning the correlation between the teacher information thus obtained and the feature quantity extracted by the feature quantity extraction means, and the storage means includes learning The criterion information obtained by the means is stored.

  The biological information processing method or program according to the first aspect of the present invention includes an image input control step (for example, step S31 in FIG. 6) for controlling input of image information in a reaction region, and an image input control step process. A feature amount extraction step (for example, the process of step S253 in FIG. 28) for extracting the feature amount of the image information of the reaction region whose input is controlled by the input, and the presence or absence of reliability when the image information of the reaction region is used for the measurement The acquisition control step (for example, the process of step S11 in FIG. 5) for controlling the acquisition of the determination reference information (for example, the spot reliability determination information) serving as the determination reference, and the determination in which the acquisition is controlled by the process of the acquisition control step Based on the reference information, it is determined whether or not there is reliability when the image information of the reaction region having the feature amount extracted by the feature amount extraction step is used for the measurement. The degree of reactivity representing the state of the biological reaction between the first biological material and the second biological material in the reaction region based on the determination step (for example, the processing of step S254 in FIG. 28) and the determination result of the determination step. A calculation step for calculating information (for example, the process of step S39 in FIG. 6), and the criterion information uses the image information for learning, and the reliability of the image information of the reaction region when used for the measurement. It is obtained by learning the correlation between the information indicating presence / absence and the feature amount of the image information in the reaction region, and the feature amount is a value having a correlation with the reliability of the image information in the reaction region.

  The learning apparatus according to the second aspect of the present invention (for example, the biological information processing apparatus 1 in FIG. 1) is provided for a first biological substance fixed to a reaction region provided on a substrate, and the first biological substance. In order to measure the state of the biological reaction with the second biological substance that reacts biologically, the learning device executes a learning process necessary for determining the presence or absence of the reliability of the reaction region. An image input unit that receives an input (for example, the excitation light intensity estimating unit 81 in FIG. 1), a display unit that displays image information input by the image input unit (for example, the display unit 92 in FIG. 1), and a display unit A teacher information input means (for example, the input unit 29B in FIG. 1) for receiving, as teacher information, information indicating the presence or absence of reliability of the image information in the reaction area by a user who refers to the displayed image information; For each of the reaction area image information Features for feature quantity extraction means (for example, feature quantity extraction unit 86 in FIG. 1) for extracting feature quantities having a correlation with reliability, and information indicating presence / absence of reliability of reaction regions input by a teacher information input means Learning means (for example, SVM 91 in FIG. 1) for learning the correlation between the feature values extracted by the quantity extraction means.

  In the learning device according to the second aspect of the present invention, the feature amount extracted by the feature amount extraction unit includes an image (for example, debris) included in the reaction region and corresponding to a substance that becomes an obstacle to perform measurement. A value corresponding to the area (for example, the number of pixels in the spot area) is included.

  In the learning device according to the second aspect of the present invention, the feature amount extracted by the feature amount extraction means is a value corresponding to the biological reaction amount of the first biological material and the second biological material in the reaction region (for example, Hybridization value).

  In the learning device according to the second aspect of the present invention, the feature amount extracted by the feature amount extraction unit includes information indicating the ease of biological reaction between the first biological material and the second biological material in the reaction region. (Eg, a probe-target binding strength matrix).

  The learning control method or program according to the second aspect of the present invention includes an image input control step (for example, the processing in step S351 in FIG. 38) for controlling the input of image information in the reaction region, and the image input control step. As teacher information by a user who refers to the display control step (for example, the process of step S353 in FIG. 38) for controlling the display of the image information whose input is controlled and the image information whose display is controlled by the process of the display control step. A teacher information input control step (for example, the process of step S353 in FIG. 38) for controlling the input of information indicating the presence / absence of reliability of the image information in the reaction region, and the reliability of the image information in the reaction region for each reaction region. A feature amount extraction step (for example, the process of step S352 in FIG. 38) for extracting feature amounts having a correlation with each other, and a teacher information input control step. A learning step (for example, the process of step S355 in FIG. 38) learns the correlation of the feature amount extracted by the feature amount extraction step with respect to the information indicating the presence or absence of the reliability of the reaction region whose input is controlled by the step processing. ).

  The meanings of terms used in the present specification will be described below.

  The probe refers to a biological substance fixed on a bioassay substrate such as a DNA chip, which reacts with a target.

  The target refers to a biological material that bioreacts with a biological material fixed on a bioassay substrate such as a DNA chip.

  Biological substances include genes generated in vivo such as proteins, nucleic acids, and sugars, as well as genes having mutually complementary base sequences or substances derived therefrom.

  Biological reaction means that two or more biological substances react biochemically. A typical example is hybridization.

  Hybridization refers to a complementary strand (double strand) forming reaction between nucleic acids having a complementary base sequence structure.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 illustrates a configuration example of a biological information processing apparatus according to an embodiment of the present invention. The biological information processing apparatus 1 includes a DNA chip 11, a pickup unit 21, a fluorescence intensity acquisition unit 22, an excitation light intensity calculation unit 23, a hybridization amount estimation unit 24, an expression amount calculation unit 25, a standardization unit 26, an output unit 27, It comprises an expression profile data storage unit 28, a user interface (UI) unit 29 having a display unit 29A and an input unit 29B, a fluorescence intensity-hybridization amount conversion type storage unit 30, and a mechanical learning unit 31.

  The biological information processing apparatus 1 has two modes, a learning mode and a determination mode. The learning mode is a mode for causing the mechanical learning unit 31 to perform learning for automatically determining reliability for each spot. In addition, the determination mode is to automatically determine the reliability for each spot based on the learning by the mechanical learning unit 31 as necessary, so that the person can visually determine the reliability for each spot. Compared with the case of performing the above, various analysis processes using the DNA chip 11 can be performed accurately and in a short time by preventing the occurrence of variations in measurement results.

  Note that the learning result obtained by the learning mode of the biological information processing apparatus 1 may be used in the determination mode of another biological information processing apparatus. In addition, the biological information processing apparatus 1 can perform determination mode processing using a learning result obtained by a learning mode of another biological information processing apparatus.

  The DNA chip 11 has a spot 12 and a guide 13. FIG. 2 shows a more detailed configuration example of the DNA chip 11.

  The DNA chip 11 has an expression analysis reaction tank 101 and a cell number counting reaction tank 102 on its substrate 11A. A linear start position guide 13A is provided at the lower end of the substrate 11A in the figure, and an end position guide 13B is provided at the upper end in the figure. Specifically, the guide 13 shown in FIG. 1 includes a start position guide 13A and an end position guide 13B.

  The expression analysis reaction tank 101 and the cell number counting reaction tank 102 are arranged between the start position guide 13A and the end position guide 13B.

  In the expression analysis reaction tank 101, a plurality of spots 12 as reaction regions are formed, and in each spot 12, a hybridization verification probe 111 as a biological material (first biological material), an expression analysis A probe 112 and an expression standardization control probe 113 are fixed. When a sample is dropped into the expression analysis reaction tank 101, a target 111A as a biological material (second biological material) having a base having a complementary structure to the base is hybridized to the hybridization verification probe 111. To do. Similarly, the target 112A as a biological material (second biological material) having a base having a complementary structure to the base is hybridized to the expression analysis probe 112. In addition, the expression standardization control probe 113 is hybridized with a target 113A as a biological material (second biological material) having a base complementary to the base.

  When the DNA chip 11 is produced, a predetermined amount of a synthesized target is dropped by a predetermined amount and irradiated with excitation light having a predetermined intensity to obtain the fluorescence intensity. A probe-target bond strength matrix is obtained that indicates a relationship with the amount of each of the plurality of targets that bind to the target. The probe-target binding strength matrix indicates the ease of hybridization between the probe and the target, and is a value that is uniquely obtained for each DNA chip 11.

  In the cell number counting reaction tank 102, a hybridization verification probe 114 as a biological material (first biological material) and a cell number counting control probe 115 are respectively attached to a spot 12 as a reaction region. When a sample is dropped into the cell number counting reaction tank 102, a target 114A as a biological material (second biological material) having a base complementary to the base is hybridized to the hybridization verification probe 114. The target 115A as a biological material (second biological material) having a base complementary to the base hybridizes with the control probe 115 for counting the number of cells.

  An intercalator 116 is coupled to the probe and target as the hybridized (bioreacted) biological material. The intercalator 116 generates fluorescence when irradiated with excitation light.

  FIG. 2 shows a state in which the target is hybridized to each probe as described above. In FIG. 2, for convenience, only one probe is shown for one spot 12, but actually, a plurality of probes of the same type are fixed to one spot 12. Further, in each reaction tank, an arbitrary number of spots to which the same type of probe is fixed are arranged at predetermined positions.

  The pickup unit 21 shown in FIG. 1 includes a fluorescence intensity acquisition pickup 41, a guide signal acquisition pickup 42, a control unit 43, an objective coordinate calculation unit 44, and a convolution expansion unit 45. The pickup unit 21 performs the same processing in both the learning mode and the determination mode.

  The fluorescence intensity acquisition pickup 41 is a pickup that acquires images of the expression analysis reaction tank 101 and the cell number counting reaction tank 102 of the DNA chip 11 of FIG. On the other hand, the guide signal acquisition pickup 42 is a pickup for reading the start position guide 13A and the end position guide 13B.

  The fluorescence intensity acquisition pickup 41 includes an objective lens 51, a prism 52, a semiconductor laser 53, and a photodiode 54. Laser light (excitation light) emitted from the semiconductor laser 53 is incident on the objective lens 51 via the prism 52, and the objective lens 51 irradiates the incident laser light on the substrate 11A (spot 12). The objective lens 51 also makes light from the spot 12 incident on the photodiode 54 via the prism 52. A plurality of probes are fixed to each spot 12, and when the probe and the target are hybridized, an intercalator 116 is further coupled to both. That is, when the probe and the target are not hybridized, there is no intercalator 116 between them, and only when they are hybridized, there is an intercalator 116 between them. The intercalator 116 generates fluorescence when irradiated with excitation light. The fluorescence condensed by the objective lens 51 is separated from the excitation light by the prism 52 and is incident on the photodiode 54.

  The greater the amount of hybridization, the greater the amount of intercalator 116, and hence the greater the amount of fluorescence generated therefrom. Therefore, it is possible to measure the state of hybridization (obtain information on hybridization) based on the intensity of fluorescence.

  The control unit 43 controls the current of the semiconductor laser 53 and adjusts the intensity of the excitation light. Further, the control unit 43 reads the output (current amount change) of the photodiode 54.

  The convolution developing unit 45 receives a signal based on the change in the amount of current output from the photodiode 54 from the control unit 43, and generates image data in units of pixels.

  The guide signal acquisition pickup 42 includes an objective lens 61, a prism 62, a semiconductor laser 63, and a photodiode 64. The semiconductor laser 63 generates laser light based on control from the control unit 43 (this laser light functions as guide detection light). The prism 62 makes the laser beam from the semiconductor laser 63 incident on the objective lens 61, and the objective lens 61 irradiates the substrate 11A with this laser beam. The objective lens 61 receives the reflected light from the substrate 11A, and the prism 62 separates the reflected light from the irradiated light and emits it to the photodiode 64. The photodiode 64 photoelectrically converts the reflected light incident from the prism 62 and outputs it as a guide signal to the control unit 43. The control unit 43 outputs the guide signal input from the photodiode 64 to the objective coordinate calculation unit 44. The guides 13 (the start position guide 13A and the end position guide 13B) are formed so that the reflectance is higher (or lower) than the other areas of the substrate 11A. The objective coordinate calculation unit 44 determines the positions of the start position guide 13A and the end position guide 13B and the start position guide 13A based on the level of the guide signal supplied from the guide signal acquisition pickup 42 via the control unit 43. The position (coordinates) of the guide signal acquisition pickup 42 moved at a constant speed toward the end position guide 13B is calculated.

  The control unit 43 controls the position of the fluorescence intensity acquisition pickup 41 (objective lens 51) based on the position of the guide signal acquisition pickup 42 calculated by the objective coordinate calculation unit 44. The guide signal acquisition pickup 42 and the fluorescence intensity acquisition pickup 41 are fixed to each other in a predetermined positional relationship, and the fluorescence intensity acquisition pickup 41 is arranged between the start position guide 13A and the end position guide 13B in FIG. The guide signal acquisition pickup 42 is disposed at a predetermined position between the start position guide 13A and the end position guide 13B.

The fluorescence intensity acquisition unit 22 receives the input of the fluorescence intensity (pf _{x, y} ) from each spot 12 (its coordinates (x, y)) output from the photodiode 54 of the fluorescence intensity acquisition pickup 41, and receives this fluorescence intensity. Is output to the excitation light intensity estimation unit 81 of the hybridization amount estimation unit 24. The fluorescence intensity data supplied from the fluorescence intensity acquisition unit 22 to the excitation light intensity estimation unit 81 of the hybridization amount estimation unit 24 includes a fluorescence image (fluorescence image (pf _{x, y} ) from the coordinates (x, y)). Alternatively, it is also referred to as an expression profile image used for obtaining the expression level). Details of the fluorescence intensity data supplied from the fluorescence intensity acquisition unit 22 to the excitation light intensity estimation unit 81 of the hybridization amount estimation unit 24 will be described later with reference to FIG.

  The fluorescence intensity acquisition unit 22 also receives control signals for controlling the objective coordinates (x, y), the objective area radius (r), and the excitation light intensity of the objective lens 51 of the fluorescence intensity acquisition pickup 41 on the substrate 11A. Output to 43. The control unit 43 controls the objective lens 51 based on this control signal. Thereby, the objective lens 51 is arranged at predetermined coordinates (x, y) on the substrate 11A, and the radius (objective area radius) (r) of the irradiation range of the laser light emitted from the objective lens 51 becomes a predetermined value. The laser light intensity (excitation light intensity) is adjusted to a predetermined value.

  The fluorescence intensity acquisition unit 22 outputs the fluorescence intensity supplied from the control unit 43 to the excitation light intensity calculation unit 23. Based on the conversion formula stored in the fluorescence intensity-hybridization amount conversion formula storage unit 30, the excitation light intensity calculation unit 23 is optimized based on the fluorescence intensity input from the fluorescence intensity acquisition unit 22 during the prescan. The excitation light intensity is calculated, and the excitation light intensity obtained by the calculation is output to the fluorescence intensity acquisition unit 22. During the main scan, the fluorescence intensity acquisition unit 22 controls the current of the semiconductor laser 53 based on the excitation light intensity from the excitation light intensity calculation unit 23, and causes the semiconductor laser 53 to emit excitation light having a predetermined intensity.

  The hybridization amount estimation unit 24 includes an excitation light intensity estimation unit 81, a creation unit 82, an image processing unit 83, a verification unit 84, a hybridization amount calculation unit 85, and a feature amount extraction unit 86. The hybridizing amount estimation unit 24 executes partially different processes in the learning mode and the determination mode.

  The excitation light intensity estimation unit 81 is the image information of the reaction region, specifically, the image data based on the fluorescence intensity supplied from the fluorescence intensity acquisition unit 22 or the expression profile data in both the learning mode and the determination mode. Upon receiving input of image information such as expression profile data stored in advance in the storage unit 28, and as necessary (for example, when the excitation light intensity of the expression profile data obtained by another device is unknown), it will be described later Using the fluorescence intensity-hybridization amount conversion formula stored in the fluorescence intensity-hybridization amount conversion storage unit 30, a process for estimating the excitation light intensity is performed.

  In the expression profile data storage unit 28, the hybridization amount is estimated by the hybridization amount estimation unit 24, the expression amount is calculated by the expression amount calculation unit 25 described later, standardized by the standardization unit 26, and output. In addition to the expression profile data output by 27, it is possible to store expression profile data composed of fluorescence images captured by different devices and information relating to the fluorescence images.

  In both the learning mode and the determination mode, the creation unit 82 uniquely determines the amount of hybridization from the fluorescence intensity as necessary (for example, when fluorescence images are acquired at a plurality of excitation light intensities). Execute the process to create the expression hybridize (pf).

  In the learning mode, the image processing unit 83 performs predetermined processing on the image data input from the creation unit 82 and executes various types of image processing necessary for feature amount extraction. Further, in the determination mode, the image processing unit 83 performs predetermined processing necessary for displaying an image, estimating a hybridizing amount, or extracting a feature amount, on the image data input from the creating unit 82.

  As will be described later with reference to FIG. 11, the image processing unit 83 removes debris (a substance that hinders measurement) from the image of the DNA chip 11, and decomposes it into an image for each spot 12. Process. Further, the image processing unit 83 can perform a process of removing the debris component based on an input instructed by the user via the user interface unit 29.

  The user interface unit 29 displays the image input from the image processing unit 83 on the display unit 29A. In addition, the user interface unit 29 is also provided with an input unit 29B that receives a user operation input. In the learning mode and the determination mode, the user interface unit 29 receives an input of a probe-target binding strength matrix corresponding to the DNA chip 11 being used, This is supplied to the feature quantity extraction unit 86. The probe-target bond strength matrix is information indicating the ease of hybridization between the probe provided in each spot of the DNA chip 11 and the dropped target.

  In addition, the input unit 29B receives the spot from the user who refers to the expression profile image displayed on the display unit 29A in the learning mode, for example, when large debris is present or the fluorescence intensity is extremely low or high. SVM91 is input as information for distinguishing a spot that is considered to have a low reliability of the hybridization amount obtained from the fluorescence intensity of the light, that is, a spot having a low reliability and a spot having a high reliability, and is used as teacher data for learning processing. To supply.

  The input unit 29B receives an operation input instructing removal of debris from a user who refers to the expression profile image displayed on the display unit 29A in both the learning mode and the determination mode, and supplies the operation input to the image processing unit 83. It is also possible to do.

  In both the learning mode and the determination mode, the verification unit 84 performs hybridization based on the amount of hybridization in the spot 12 of the hybridization verification probes 111 and 114 in the image data input from the image processing unit 83. Verify that is done correctly.

  In the learning mode, the hybridization amount calculation unit 85 decomposes the in-spot region and calculates the hybridization value and the reliability in units of spots. As the feature amount used in the learning process, for example, the fluorescence of the spot region is calculated. The intensity, the fluorescence intensity of the area around the spot, the number of pixels included in the area within the spot, the hybridization value, the spot reliability, and the like are acquired. Further, in the determination mode, the hybridization amount calculation unit 85 divides a spot area, selects a spot area as necessary, and hybridizes in units of spots as described later with reference to FIG. The soybean value and reliability are calculated, and the hybridization value and reliability in spot units are output. When determining the reliability of the spot, the hybridization amount calculation unit 85 uses the determination result of the reliability of each spot by the mechanical learning unit 31 as necessary.

  In the learning mode, the feature amount extraction unit 86 acquires feature amount data at each spot of the expression profile image displayed on the display unit 29A, and supplies the feature amount data to the SVM 91 for the learning process. In addition, the feature amount extraction unit 86 acquires feature amount data at each spot of the expression profile image to be measured in the determination mode, and supplies the feature amount data to the SVM 91 to determine the reliability of the spot. The feature amount data extracted by the feature amount extraction unit 86 is a feature amount having a correlation with the reliability of the spot, and details thereof will be described later.

  In the determination mode, the expression level calculation unit 25 estimates the expression level corresponding to the fluorescence intensity by obtaining the binding strength of the target to the probe based on the output from the hybridization level calculation unit 85. In the determination mode, the standardization unit 26 performs standardization processing using the expression standardization control probe 113 and the cell number counting control probe 115. The output unit 27 supplies the standardized data to the expression profile data storage unit 28 in the determination mode.

  The expression profile data storage unit 28 stores the data supplied from the output unit 27 as expression profile data. In addition, the expression profile data storage unit 28 can store expression profile data obtained by other devices. It is preferable that the expression profile data obtained by another apparatus includes information on a probe-target binding strength matrix indicating the ease of hybridization between the probe and the target.

  The data stored in the expression profile data storage unit 28 is supplied to the excitation light intensity estimation unit 81 of the hybridization amount estimation unit 24 and can be used for both the learning mode and the determination mode. The data stored in the expression profile data storage unit 28 is supplied to the user interface unit 29 as necessary and displayed on the display unit 29A. Data output from the expression level calculation unit 25 is also displayed on the display unit 29A as necessary.

  As will be described later with reference to FIG. 9, the fluorescence intensity-hybridization amount conversion storage unit 30 is a conversion equation that uniquely determines the relationship between the fluorescence intensity and the corresponding hybridization amount. Or may be data for conversion) in advance.

  The mechanical learning unit 31 includes an SVM (Support Vector Machine) 91 and a spot removal pattern database (DB) 92.

  In the learning mode, an expression profile image is displayed on the display unit 29A of the user interface unit 29. Further, the feature amount extraction unit 86 extracts the feature amount for each spot of the displayed expression profile image, and supplies it to the SVM 92. The user refers to the expression profile image displayed on the display unit 29A, for example, when large debris is present, or when the fluorescence intensity of the spot is very high or low, or the fluorescence intensity within the spot is increased. Distinguish between those that seem to have low reliability of the hybridization amount obtained from the fluorescence intensity of the spot, such as bias, that is, those that have low reliability as spots and those that have high reliability as spots input. In response to a plurality of expression profile images (the number of samples used for learning increases, the reliability of learning increases), the SVM 91 receives the reliability of each spot as teacher data from the user interface unit 29. A pattern for determining the presence or absence of reliability for each spot based on the supply of feature amount data of each spot from the feature amount extraction unit 86. Do learning. The SVM 91 stores spot reliability determination information as a learning result in the spot removal pattern database 92.

  In the determination mode, the SVM 91 is also supplied with the feature amount data supplied from the feature amount extraction unit 86, and based on the spot reliability determination information stored in the spot removal pattern database 92, the reliability of the corresponding spot The determination result is output to the hybridization amount calculation unit 85. That is, in the determination mode, the user does not need to determine the spot reliability with reference to the expression profile image.

  In the biological information processing apparatus 1, a device that executes the learning mode may be distinguished from a device that executes the determination mode. That is, as a result of executing the learning mode, the spot reliability determination information stored in the spot removal pattern database 92 can be supplied to another biological information processing apparatus and used in the apparatus, or other biological information can be used. The spot reliability determination information obtained by the learning mode executed by the processing apparatus is acquired and stored in the spot removal pattern database 92, so that it can be used for the determination mode, or the spot removal pattern database. The spot reliability determination information stored in 92 can be used in its own determination mode.

  The quantitative measurement of the gene expression level and the learning process for performing the measurement are performed by the experimental process processing device 131 shown in FIG. The biological information processing apparatus 1 in FIG. 1 constitutes a part of the experimental process processing apparatus 131 in FIG.

  That is, the experimental process processing device 131 includes an adjustment unit 141, a hybridization unit 142, an acquisition unit 143, an expression level estimation unit 144, a standardization unit 145, an output unit 146, a storage unit 147, a feature amount extraction unit 148, and a learning processing unit 149. And the operation input unit 150. Among these, the acquisition unit 143, the expression amount estimation unit 144, the standardization unit 145, the output unit 146, the storage unit 147, the feature amount extraction unit 148, the learning processing unit 149, and the operation input unit 150 are included in the biological information processing apparatus 1. It is configured.

  Specifically, the acquisition unit 143 includes a pickup unit 21, a fluorescence intensity acquisition unit 22, an excitation light intensity calculation unit 23, and a fluorescence intensity-hybridization amount conversion expression storage unit 30, and the expression level estimation unit 144 includes: The hybridizing amount estimating unit 24 includes a part other than the feature amount extracting unit 86 and the expression amount calculating unit 25, the standardizing unit 145 includes the standardizing unit 26, and the output unit 146 includes the output unit 27. The unit 147 includes the expression profile data storage unit 28. The feature quantity extraction unit 148 includes the feature quantity extraction unit 86 of the hybridization amount estimation unit 24, the learning processing unit 149 includes the mechanical learning unit 31, and the operation input unit 150 includes the user interface unit 29. The input unit 29B.

  The adjustment unit 141 adjusts the target in both the learning mode and the determination mode. The hybridizing unit 142 hybridizes the probe and the target in both the learning mode and the determination mode. The acquisition unit 143 acquires the fluorescence intensity in both the learning mode and the determination mode. In the learning mode, the expression level estimation unit 144 executes various processes necessary for extracting feature quantities used for learning based on the fluorescence intensity obtained by the hybridized learning DNA chip 11. The expression level estimation unit 144 also performs an expression level estimation process using the determination result of the spot reliability determination information obtained as a result of learning by the learning processing unit 149 in the determination mode.

  The feature amount extraction unit 148 extracts the feature amount obtained by the processing of the expression amount estimation unit 144 in both the learning mode and the determination mode. In the learning mode, the learning processing unit 149 performs learning based on the feature amount data extracted by the feature amount extracting unit 149 and the reliability determination result for each spot as the teacher data input by the operation input unit 150. And remember the learning results. The learning processing unit 149 further determines the reliability of the corresponding spot based on the feature amount data extracted by the feature amount extraction unit 149 in the determination mode. The operation input unit 150 receives an input of a determination result for each spot as teacher data in the learning mode of the learning processing unit 149, and supplies it to the learning processing unit 149. In addition, the operation input unit 150 can receive an input of a value (for example, a probe-target bond strength matrix) serving as a feature amount and supply it to the learning processing unit 149 in both the learning mode and the determination mode. The standardization unit 145 standardizes data. The output unit 146 outputs expression profile data. The storage unit 147 stores expression profile data.

  Note that the device that executes the learning mode or the device that executes the determination mode may be configured independently of the experimental process processing device 131 shown in FIG. That is, the spot reliability determination information obtained as a result of learning by the learning processing unit 149 can be supplied to another experimental process processing apparatus and used in the apparatus, or the learning mode executed by the other experimental process processing apparatus. It is also possible to acquire the spot reliability determination information obtained by the above and use it in the determination mode. Also, the spot reliability determination information obtained as a result of learning by the own learning processing unit 149 It can also be used in its own judgment mode.

  Next, processing 1 of the experimental process executed by the adjustment unit 141, the hybridization unit 142, and the acquisition unit 143 in the processing of the experimental process processing device 131 of FIG. 3 will be described with reference to the flowchart of FIG. . Process 1 of the experimental process is executed in both the learning mode and the determination mode.

  First, in step S1, the adjustment unit 141 adjusts the target. Specifically, a sample containing cells is taken out, processed to denature and remove proteins from it, RNA (ribonucleic acid) extraction, fragmentation, DNA (deoxyribonucleic acid) extraction, fragment The target (target 112A for the expression analysis probe 112) is generated.

  In step S2, the hybridizing unit 142 executes a hybridizing process. Specifically, in the solution containing the target generated in step S1, the targets 111A and 114A for the hybridization verification probes 111 and 114, the target 113A for the expression normalization control probe 113, and the cell count counter A target 115A for the control probe 115 is added, and this solution is dropped into the expression analysis reaction tank 101 and the cell number counting reaction tank 102, whereby the target and the probe are hybridized. Then, the intercalator 116 is introduced and coupled to the hybridized target and probe to obtain the DNA chip 11 as shown in FIG. As shown in the figure, in the spot 12 of the expression analysis reaction tank 101, the target 112A is hybridized to the expression analysis probe 112, and the target 113A is hybridized to the expression standardization control probe 113. The target 111A is hybridized with the probe 111 for hybridization verification. And the intercalator 116 has couple | bonded between the probe and target which couple | bonded with those double strands.

  Similarly, also in the spot 12 of the cell number counting reaction tank 102, the target 114A is hybridized to the hybridization verification probe 114, and the target 115A is hybridized to the cell number counting control probe 115. ing. An intercalator 116 is also coupled between these hybridized probe and target.

  In step S3, the acquisition unit 143 acquires the fluorescence intensity. Specifically, the fluorescence intensity acquisition unit 22 drives the fluorescence intensity acquisition pickup 41 via the control unit 43 and causes the semiconductor laser 53 to emit laser light as excitation light. The excitation light is incident on the objective lens 51 via the prism 52, and the objective lens 51 irradiates the expression analysis reaction tank 101 on the substrate 11A.

  The intercalator 116 generates fluorescence when irradiated with excitation light. This fluorescence is collected by the objective lens 51 and is incident on the photodiode 54 via the prism 52. The photodiode 54 outputs a current corresponding to the fluorescence. The control unit 43 causes the convolutional expansion unit 45 to convert a signal corresponding to the current into an image signal, and outputs a signal corresponding to the fluorescence intensity generated by the conversion to the fluorescence intensity acquisition unit 22.

  The control unit 43 moves the position of the objective lens 51 from the start position guide 13A toward the end position guide 13B. At this time, laser light as guide detection light emitted from the semiconductor laser 63 of the guide signal acquisition pickup 42 is incident on the objective lens 61 via the prism 62, and the objective lens 61 irradiates the substrate 11A with this guide detection light. To do. The intensity of the reflected light of the guide detection light increases when it is applied to the start position guide 13A and the end position guide 13B. This reflected light is incident on the prism 62 via the objective lens 61, and is incident on the photodiode 64 from the prism 62. The objective coordinate calculation unit 44 acquires a guide signal from the photodiode 64 via the control unit 43, and on the basis of this signal, the guide signal acquisition pickup 42 (therefore, the fluorescence intensity acquisition pickup 41 integrated therewith). Is located between the start position guide 13A and the end position guide 13B of the substrate 11A. Based on the coordinates, the control unit 43 moves (scans) the guide signal acquisition pickup 42 (fluorescence intensity acquisition pickup 41) from the start position guide 13A to the end position guide 13B at a constant speed.

  In this way, the fluorescence intensity acquisition pickup 41 is moved from the start position guide 13A to the end position guide 13B in FIG. 2, and the scanning position is further changed to the start position guide 13A (end position guide 13B). ) In the direction parallel to () in the figure (x-coordinate direction) by one pitch, and similarly, from the start position guide 13A to the end position guide 13B at the new movement position. In this manner, the entire expression analysis reaction tank 101 and cell number counting reaction tank 102 are scanned, and image signals at respective coordinates are output from the fluorescence intensity acquisition pickup 41.

  Next, among the processes of the experimental process processing device 131 of FIG. 3, the process 2 of the experimental process executed by the expression level estimation unit 144, the standardization unit 145, the output unit 146, the feature value extraction unit 148, and the learning processing unit 149 is performed. Will be described with reference to the flowchart of FIG. Process 2 of the experimental process is a process in the determination mode.

  In step S <b> 11, the spot removal pattern database 92 acquires and stores spot reliability determination information obtained by a learning process described later from the SVM 91 in which the learning mode has ended or from an external input from another device. The spot reliability determination information is information used in an expression level estimation process described later, and details of the learning process will be described later with reference to the flowchart of FIG.

  In step S12, the expression level estimation unit 144 executes an expression level estimation process. The details of the expression level estimation process will be described later with reference to FIG. 6. By this process, the amount of hybridization is calculated, and the expression level is calculated.

  Next, in step S13, the standardization unit 145 (standardization unit 26) performs processing for standardizing data. Here, standardization by the expression standardization control probe 113 and standardization by the cell number counting control probe 115 are performed. Standardization by the expression standardization control probe 113 is performed as follows. That is, in FIG. 2, the expression standardization control probe 113 is shown only in one place, but actually, this expression standardization control probe 113 is a predetermined predetermined in the expression analysis reaction tank 101. It is distributed and arranged at a plurality of positions (for example, four corners of the expression analysis reaction tank 101 and approximately five at the center). Then, based on the fluorescence value of the expression standardization control probe 113 arranged at each position, a correction curved surface is calculated based on, for example, a B-spline curved surface, and each pixel is calculated based on the fluorescence value obtained by the correction curved surface. Normalization is performed by dividing the fluorescence value. By this normalization, variation in hybridization due to the position of the spot 12 in the reaction tank 101 for expression analysis is corrected.

  Further, the standardization by the cell number counting control probe 115 is performed based on the value of the amount of hybridization to the cell number counting control probe 115 (fluorescence value based on the cell number counting control 115). This is done by dividing the fluorescence value of the pixel on the spot 12. As the control probe 115 for calculating the number of cells, a repetitive sequence (for example, Alu sequence for humans) in the genome of the living body from which the expression analysis probe 112 is extracted is used. By this processing, the expression level of the acquired gene can be converted into a value per certain number of cells.

  Further, in step S14, the output unit 146 (output unit 27) outputs the expression profile data. Specifically, expression profile data (described later with reference to FIG. 8) constituted by expression profile image data and information related thereto is supplied to the storage unit 147 (expression profile data storage unit 28) and recorded.

  By such processing, the amount of hybridization is calculated, the amount of expression is calculated, the obtained data is standardized and the variation in hybridization is corrected, and the expression profile data obtained by these processing is output. Is done.

  Next, the expression level estimation process executed in step S12 of FIG. 5 will be described with reference to the flowchart of FIG.

  In step S31, the excitation light intensity estimation unit 81 inputs image information. Specifically, the excitation light intensity estimation unit 81 receives supply of fluorescence intensity data obtained by the process described with reference to FIG. 4 from the fluorescence intensity acquisition unit 22 or is stored in the expression profile data storage unit 28. Is supplied with expression profile data.

  In step S32, the excitation light intensity estimation unit 81 determines whether or not the input image information is expression profile data.

  The format of data supplied by the excitation light intensity estimation unit 81 is, for example, as shown in FIG. 7 or FIG. As shown in FIG. 7, the fluorescence intensity data supplied from the fluorescence intensity acquisition unit 22 is used as the intensity of the excitation light irradiated when the fluorescence intensity is acquired and the respective probes of the DNA chip 11. It includes at least a probe gene index that is a list of genes, the number of vertical and horizontal pixels, and a fluorescence image.

  On the other hand, the expression profile data stored in the expression profile data storage unit 28 is different from the fluorescence intensity data supplied from the fluorescence intensity acquisition unit 22 and may be obtained by another device. As shown in FIG. 8, the image data 181 and common data 182 are included. The expression profile data stored in the expression profile data storage unit 28 is subjected to processing such as estimation of hybridization, calculation of expression level, or standardization based on the fluorescence intensity data described with reference to FIG. This is data that is created, or data that is necessary for various processes executed by the biological information processing apparatus 1 is added to a fluorescent image acquired by another apparatus. That is, the image data 181 includes the number of images to be set, the excitation light intensity irradiated at the time of obtaining the fluorescence intensity, the number of vertical and horizontal pixels, and the fluorescence image. The common data 182 is data other than the fluorescence image necessary for various processes executed by the biological information processing apparatus 1, and is used, for example, to determine the spot position of the DNA chip when executing image processing. It is composed of a spot position template image, the number of spots, a probe gene index, and the like.

  That is, when the excitation light intensity estimation unit 81 receives the data shown in FIG. 8, the excitation light intensity estimation unit 81 determines that the expression profile data has been supplied, and the process proceeds to step S33. Then, when the excitation light intensity estimation unit 81 receives the supply of the fluorescence intensity data shown in FIG. 7, it determines that the expression profile data is not supplied. Since the fluorescence intensity data shown in FIG. 7 includes the zero-time high intensity information, the excitation light intensity estimation unit 81 does not need to perform the excitation light intensity estimation process. The data is supplied to the creation unit 82, and the process proceeds to step S35.

  In step S33, the excitation light intensity estimation unit 81 determines whether the image information input in step S31 includes (includes) excitation light intensity information.

  When the data supplied by the excitation light intensity estimation unit 81 is fluorescence intensity data supplied from the fluorescence intensity acquisition unit 22, the fluorescence obtained at the excitation light intensity calculated and set by the processing of the excitation light intensity calculation unit 23 Since the image is included, the fluorescence intensity data includes excitation light intensity information. On the other hand, when the expression profile data is supplied from the expression profile data storage unit 28, it is described above when the data is supplied from the fluorescence intensity acquisition unit 22 and stored after being subjected to various processes. As described above, the excitation light intensity information exists, but if it is not (for example, image data supplied from another device), the excitation light intensity information may not exist in the expression profile data.

  Therefore, in the case of at least image data supplied from the fluorescence intensity acquisition unit 22, the excitation light intensity is included in the image data 181, and therefore it is determined that there is excitation light intensity information. On the other hand, when the expression profile image data stored in the expression profile data storage unit 28 is supplied to the excitation light intensity estimation unit 81, it is an image supplied and stored from the fluorescence intensity acquisition unit 22. Sometimes, as described above, the excitation light intensity information exists, but when it is not (when the image data is supplied from another apparatus), the corresponding excitation light intensity information may not exist.

  If it is determined in step S33 that the excitation light intensity information does not exist, the excitation light intensity estimation unit 81 in step S34 stores the fluorescence intensity-hybridization stored in the fluorescence intensity-hybridization amount conversion expression storage unit 30. Using the quantity conversion equation, a process for estimating the excitation light intensity is executed, and the supplied image data is supplied to the creation unit 82. Since the creating unit 82 does not create the formula (1) (hybridize (pf)) for determining the amount of hybridization based on the estimated fluorescence intensity, the creating unit 82 supplies the supplied image data to the image processing unit 83 as it is.

  The process of estimating the excitation light intensity can be executed when the image data is measured based on at least two different intensity excitation lights. If there is no excitation light intensity information and there is no image data based on at least two different excitation light intensities, the excitation light intensity cannot be estimated. Therefore, in these cases, the process of step S34 is skipped.

If it is determined in step S33 that the excitation light intensity information exists, the excitation light intensity estimation unit 81 does not need to perform the excitation light intensity estimation process, and therefore supplies the supplied image data to the creation unit 82. Since the creating unit 82 does not need to create the formula (1) (hybridize (pf)) for determining the amount of hybridization based on the fluorescence intensity, the creating unit 82 supplies the supplied image data to the image processing unit 83, and the processing is performed. Then, the process proceeds to step S37 described later.

  If it is determined in step S32 that expression profile data has not been supplied, in step S35, the creation unit 82 determines whether the input image information is image information of an image captured with a plurality of excitation light intensities. Determine whether. If it is determined in step S35 that the image information is an image captured with a plurality of excitation light intensities, in step S36, the creating unit 82 determines the amount of hybridization based on the fluorescence intensity (1) (hybridize) (Pf)) is generated, and the supplied image data is supplied to the image processing unit 83.

  If it is determined in step S35 that the input image data is not image data of an image captured with a plurality of excitation light intensities, the process of step S36 cannot be executed and is skipped.

Figure 9 shows the expression hybridize _e (pf) that defines the relationship between fluorescence intensity and hybridized amount of each spot. As shown in the figure, when the fluorescence intensity is given, the corresponding hybridization amount is uniquely determined based on the function (curve 191 to curve 194) (hybridization information includes the first biological substance and the first biological substance). 2 is uniquely determined based on the function from the fluorescence intensity obtained by hybridization of the two biological substances). However, as shown in the figure, the uppermost curve 191 in the figure represents the curve when the excitation light intensity level is the weakest. Hereinafter, the lower curve 192 and the curve 193 are sequentially excited. The light intensity level becomes strong, and the lowermost curve 194 represents the curve when the excitation light intensity level is strongest.

The expressions hybridize _s (pf _s ) and hybridize _w (pf _w ) in the expression (1) are respectively expressed as the hybridize _e (pf) having the stronger excitation light intensity and the weaker one of the obtained data. Represents the hybridize _e (pf).

  Each of the curves 191 to 194 is hybridized to a slight change in fluorescence intensity in the portion 191A to 194A in the left end region in the drawing and the portion 191B to 194B in the right end portion in the drawing. Since the amount of soybean is remarkably changed, it becomes difficult to uniquely determine the amount of hybridi corresponding to the fluorescence intensity in these regions. Therefore, only the central part excluding these parts 191A to 194A and parts 191B to 194B is used for the calculation of the amount of hybridization corresponding to the fluorescence intensity.

  If it is determined in step S33 that there is no fluorescence intensity information, after the process of step S34 is completed, if it is determined in step S35 that the image is not photographed with a plurality of excitation light intensities, or in step S36 After the processing is completed, in step S37, the image processing unit 83 performs image processing. The details of this image processing will be described later with reference to the flowchart of FIG. 11. By this processing, a debris region across the spot boundary is removed from the image of the DNA chip 11, and the image is decomposed into images for each spot. .

  In step S38, the verification unit 84 executes processing for verifying hybridization. Specifically, as shown in FIG. 2, a hybridization verification probe 111 is provided in the expression analysis reaction tank 101, and a hybridization verification probe 114 is provided in the cell number counting reaction tank 102, respectively. It is fixed to. As the hybridization verification probes 111 and 114, gene sequences that do not exist in the biological species to be experimented are used. For example, when the experiment target is an animal (when the expression analysis probe 112 is an animal gene), a plant chlorophyll gene is used as the hybridization verification probes 111 and 114, and the targets 111A and 114A are as follows: Its complementary sequence is used. That is, as the hybridization verification probes 111 and 114 and the targets 111A and 114A, those that reliably cause hybridization regardless of the hybridization of the expression analysis probe 112 and the target 112A are used. In addition, since a completely different species from that of the experiment target is used, when the hybridization verification probes 111 and 114 are sufficiently hybridized, in this experiment (in the measurement), the hybridization surely occurred. Can be verified. On the contrary, when the hybridization verification probes 111 and 114 are not sufficiently hybridized, there is a possibility that this measurement is an environment in which hybridization is difficult to occur for some reason. Therefore, by measuring the fluorescence value of the hybridization verification probes 111 and 114, if the fluorescence value is equal to or higher than a preset reference value, it is verified that correct hybridization processing is being performed. be able to. The verification unit 84 supplies the processed expression profile image data to the hybridization amount calculation unit 85 together with the acquired related information after the process of verifying hybridization.

  In step S39, the hybridization amount calculation unit 85 calculates the hybridization amount and the reliability. The details will be described later with reference to FIG. 22. When debris is present in the spot area by this processing, the spot area is divided into a plurality of areas, and each spot area and finally the spot. In units, the hybridization value and confidence are calculated.

  In step S40, the expression level calculation unit 25 executes a process of calculating the expression level based on the hybridization value and the reliability calculated by the hybridization level calculation unit 85 in the process of step S39. Returning to step S12 of FIG. 5, the process proceeds to step S13. Based on this processing, the expression level corresponding to the calculated (acquired) fluorescence value is calculated.

  By such processing, the hybridization value and the reliability of each spot are calculated, and the expression level is obtained based on this.

  FIG. 10 illustrates a functional configuration example of the image processing unit 83. As shown in the figure, the image processing unit 83 includes a separation unit 211, a noise removal unit 212, a region removal unit 213, and a decomposition unit 214.

  The separation unit 211 separates the background image using the template. The noise removing unit 212 removes noise using the characteristics of the background image. The area removing unit 213 removes a debris area that straddles the spot boundary. The decomposition unit 214 decomposes the input image into spot images.

  Next, the details of the image processing executed by the image processing unit 83 in FIG. 10 in step S37 in FIG. 6 will be described with reference to the flowchart in FIG.

  In step S61, the separation unit 211 separates the background image using the template. In FIG. 12, the process of separating the background is shown in principle. As shown in the figure, a template 231 is prepared for the expression profile image 221 input from the creation unit 82. The template 231 has a plurality of spot positions 232. The spot position 232 is a position corresponding to the plurality of spot positions 222 of the expression profile image 221. That is, since the spot position 222 of the expression profile image 221 is known, a template 231 having a spot position 232 corresponding to this spot position is prepared.

  As shown in the lower part of FIG. 12, the template 231 is translated, rotated, enlarged, or reduced with respect to the expression profile image 221, so that each spot position 232 becomes each spot position 222. The template 231 is arranged at a position corresponding to the expression profile image 221 so as to match, and then the background image is separated by removing the image of the spot position 232 from the expression profile image 221 (only the background image) Is extracted).

  Next, in step S62, the noise removing unit 212 removes noise using the background image feature. Specifically, as shown in FIG. 13, the noise removing unit 212 generates a frequency filter 241 and a trend filter 242, and removes noise by applying them to image data.

  The frequency filter is generated as shown in FIG. That is, in step S71, the noise removing unit 212 obtains frequency data by performing two-dimensional Fourier transform on the data of the background image 221A having the spot position 222 (the background image extracted in the process of step S61). In step S72, the noise removing unit 212 generates a filter that removes the high frequency component 261 based on the characteristics of the frequency data obtained by the Fourier transform. That is, as shown in FIG. 14, a filter having a characteristic curve 271 that removes the high-frequency component 261 is generated. Then, by applying the frequency filter 241 generated in this way to the image data, noise of the high frequency component can be removed.

  FIG. 15 schematically illustrates the basic configuration of the trend filter 242. As shown in the figure, the fluorescence intensity of the pixel at the position defined by the (x, y) coordinates is taken on the vertical axis (z axis) of the three-dimensional coordinate space, and the (x, y) coordinates are taken on the horizontal plane. Thus, the regression plane 301 is calculated based on the fluorescence intensity of each pixel. By applying the trend filter 242 to the image data so that the fluorescence intensity 302 at each coordinate (x, y) is represented by the distance from the regression plane 301, it is possible to remove low-frequency component noise.

  In step S63, the region removing unit 213 removes a debris region straddling the spot boundary. By this process, a debris boundary is extracted, and a debris region that intersects the spot boundary is removed.

  Specifically, for example, the region removing unit 213 differentiates data of the entire image, binarizes the image data using a predetermined threshold, extracts a debris boundary, and removes a debris region that intersects with the spot boundary. To do.

  For example, in the expression profile image 201 shown in FIG. 16, the boundary between the spot and the background image is a spot boundary 351, and various debris intersects each spot boundary 351 or is generated inside. Yes. For example, the debris 352 having an area intersects the spot boundary 351. The debris 353 having no area intersects the two spot boundaries 351. The debris 354 has no area and is located inside the spot boundary 351. On the other hand, the debris 355 is located inside the spot boundary 351 but has an area. Of the debris located on the background image, the debris 356 has an area, but the debris 357 does not have an area. Further, the debris 358 has an area and has a teardrop shape in which a fluorescent sample flows out from the spot. The debris 360 exists over two spots, and the spot boundaries 361 and 362 are covered with the debris 360. The area removing unit 213 executes a process of removing a debris area that intersects the spot boundary among these debris.

  FIG. 17 schematically shows an image when the image data of the image shown in FIG. 16 is differentiated and binarized by a threshold value. Spot boundaries 351A, debris 352A to 358A, 360A, and spot boundaries 361A, 362A correspond to spot boundaries 351, debris 352 to 358, 360, and spot boundaries 361, 362 in FIG. 16, respectively.

  Then, the debris boundary is extracted by correcting the image obtained by binarization as necessary, correcting the thickness of the line, or correcting the discontinuous portion to a continuous curve or straight line.

  FIG. 18 shows a state where the boundaries of the debris 352A to 358A and 360A in FIG. 17 are thickened to form debris boundaries 352B to 358B and 360B.

  As shown in FIG. 18, after the debris boundary is extracted, a process of removing a debris area that intersects with the spot area is executed. That is, by comparing the extracted debris boundary with the spot region, the intersecting debris region is removed. By this process, for example, the debris 352, 353, 360 in FIG. 16 is removed. This process will be further described with reference to FIG.

  In the example of FIG. 19, the debris region 352D extends over the spot boundary 351 that defines the spot region 351D. A vector 355 indicated by an arrow in FIG. 19 is a vector perpendicular to the debris boundary 352B of the debris region 352D. FIG. 20 shows a graph in which the horizontal axis represents coordinates in the direction along the vector 355 and the vertical axis represents fluorescence intensity. As shown in FIG. 20, since the fluorescence intensity in the range of the debris region 352D increases with the vicinity of the debris boundary 352B as a boundary, it is determined to be debris, and this portion of data is removed in the process of step S114. The Further, for the sake of more accuracy, data in the range indicated by D in the drawing in the vicinity of the debris boundary 352B (a range of a predetermined distance from the debris boundary 352B) may be similarly removed.

  In step S64, the decomposing unit 214 performs a process of decomposing the input image into spot images. That is, the image from which the debris area has been removed in the process of step S63 is decomposed into the image of each spot. After the process of step S64 is completed, the process returns to step S37 in FIG. 6 and proceeds to step S38.

  Through such processing, after the background image is separated from the supplied expression profile image and noise and debris areas are removed, the image is decomposed into images for each spot area in order to obtain the amount of hybridization and reliability. The

  FIG. 21 illustrates a functional configuration example of the hybridization amount calculation unit 85 that performs the calculation processing of the hybridization amount and the reliability in step S39 of FIG. As shown in the figure, the hybridization amount calculation unit 85 includes a division unit 441, a reliability calculation unit 442, a selection unit 443, and an output unit 444.

  The dividing unit 441 divides the spot area. As a result, the in-spot region is divided into a plurality of divided regions composed of regions containing unnecessary substances and regions not containing them. The reliability calculation unit 442 divides the reactivity information representing the state of biological reaction between the first biological material and the second biological material in the reaction region and the reliability information representing the reliability of the reactivity information into divided regions. Calculate every time. Specifically, the reliability calculation unit 442 calculates the hybridization value and the reliability for each spot area. In addition, the reliability calculation unit 442 combines the respective divided regions of the plurality of corresponding reaction regions used for the measurement under the same conditions, and combines the combination reactivity information as the combination reactivity information and the reliability of the combination. The combination reliability information as information is calculated for each combination, and the combination reactivity information of the combination having the largest combination reliability information is set as the reactivity information of the reaction region. The selection unit 443 calculates a multiplication value of the reliability and the spot reliability in the target spot area, and is executed based on the feature vector extracted by the feature extraction unit 86 as necessary. The determination result of the reliability of the spot by the automatic learning unit 31 is acquired, and an in-spot region is selected based on these results. The output unit 444 outputs the hybridization value and reliability in spot units.

  Next, the details of the hybrid value and reliability calculation process executed in step S39 of FIG. 6 will be described with reference to the flowchart of FIG.

  In step S201, the dividing unit 441 divides the spot area. That is, the spot area is divided into an in-spot area including debris and an in-spot area not including debris. For example, in the example shown in FIG. 23, debris 464 and 465 having no area and debris 463 having an area exist in the spot region 462 inside the spot boundary 461. The debris 464, 465 that do not have an area are removed at this stage because it is clear that it is debris.

  On the other hand, the debris 463 having an area is located in the spot region 462 inside the spot boundary 461 and does not intersect the spot boundary 461, and thus is not removed in the process of step S63 described above.

  It is determined as follows whether an arbitrary determination target point in the spot area is an internal point or an external point in the debris area. In other words, if the debris boundary has one intersection with the outermost boundary line of the background image or does not have an intersection point, any point on the outermost boundary line of the background image other than the debris boundary is within the spot region to be checked. When a straight line is drawn to an arbitrary target point and the number of intersections with the debris boundary is counted, if the count value is an even number, the target point is outside the debris area, and if the count value is an odd number, The target point can be determined as a point inside the debris region.

  Therefore, the dividing unit 441 draws a straight line from a predetermined point on the outside to the target point, and recognizes that the target point is in the debris region when the number of intersections with the debris boundary becomes an odd number. However, when the straight line and the debris boundary contact each other, the contact point is not counted as an intersection.

  For example, in the example shown in FIG. 24, when a straight line 503 is drawn from a point 502 at the upper left of the outermost boundary line 501 to a point 483A in the spot region 483 to be determined, the straight line 503 is the debris boundary of the debris 482. 482A and the debris boundary 481A of the debris 481 intersect. That is, the straight line 503 intersects the debris boundary 482A of the debris 482 with four points 491, 492, 493, and 494, and intersects the boundary 481A of the debris 481 at the point 495. Therefore, in this case, the straight line 503 intersects the debris boundary at five points. Since the number of intersections is five and is an odd number, in this case, the point 483A is determined to be an inner point of the debris 481.

  FIG. 25 shows an example in which the spot area is divided. In this example, the spot area 462 having the spot boundary 461 in FIG. 23 is divided into an in-spot area 522 including debris 463 and an in-spot area 521 not including debris.

Next, in step S202, the reliability calculation unit 442 calculates a hybridization value and a reliability for each spot area. Hybridization of the spot area i soybean value ah _i ^j (reactivity information) and a spot in the area reliability ar _i ^j (reliability information) is expressed by the following equation.

  In the above formulas (4) and (5), i represents the spot number, and j represents the number of the spot in the spot. k represents the number of the pixel in the spot area.

The hybridize (pf _i ^jk ) in the equation (4) means an equation represented as curves 191 to 194 in FIG. 9, and pn _i ^j represents the number of pixels in the spot region j.

Confidence of formula (5) (pf _i ^jk) is expressed by the following equation (6).

The weighting factor w _c in equation (6), portions 191A or 194A of the curve 191 through 194 in FIG. 9 represents a weighting factor of 191B to 194B other sections (high confidence interval), the weight coefficient w _sat is 9 Represents the weights of the sections 191A to 194A and 191B to 194B (low reliability sections) of the curves 191 to 194.

When the fluorescence intensity of the corresponding pixel of the corresponding spot in the second measurement is plotted in a coordinate space in which the fluorescence intensity of the first measurement is the horizontal axis and the fluorescence intensity of the second measurement is the vertical axis, FIG. As shown, the points 701 defined by the fluorescence intensity of the two measurements are scattered in the vicinity of a straight line 702 having an inclination of 45 degrees. Ideally, the points 701 are located on the straight line 702 because the fluorescence intensities of the pixels at the corresponding positions of the corresponding spots match. However, in actuality, variations occur, and they are not completely located on the straight line 702, but are distributed in the vicinity thereof. The distribution variation tends to be large when the fluorescence intensity is relatively small and small when the fluorescence intensity is large. That is, in the range 703 fluorescence intensity is set to a small value ph _1, the dispersion is relatively large, the fluorescence intensity values ph ₁ greater than ph _2, the value of the variance is smaller than, the fluorescence intensity is further At a large value ph ₃ , the variance in the range 703 is the smallest. The variance (variance (pf)) in Equation (6) is a value calculated as the variance of the points 701 located within a predetermined range 703 set in advance. That is, the confidence (pf) (reliability information) of the equation (6) is the first obtained by hybridizing two biological materials (first biological material and second biological material) at the time of the first measurement. It is defined by the reciprocal (1 / variance (pf)) of the dispersion in a set range of the second fluorescence intensity obtained by hybridization during the second measurement.

  Accordingly, the confidence (pf) of the equation (6) can be calculated from data measured and stored in the past, and thus the fluorescence image as the past asset can be evaluated and used effectively. . This also means that the process for obtaining the fluorescence image and the process for evaluating the obtained fluorescence image can be performed by different devices at different timings. Furthermore, since it is not necessary to take a special image other than the fluorescent image to calculate this, it is possible to prevent the apparatus from becoming complicated, costly, or processing steps from being increased and complicated. The

  In step S203, the selection unit 443 performs an in-spot area selection process which will be described later with reference to the flowchart of FIG.

  In step S204, the output unit 444 in FIG. 21 executes the output processing of the hybridization value and reliability in units of spots, which will be described later using the flowchart in FIG. 35, and the processing is performed in step S39 in FIG. Return to step S40.

  By such processing, the hybridization value and reliability in spot units are calculated and output.

  In order to perform the in-spot area selection processing in step S203, the selection unit 443 in FIG. 21 has a functional configuration as shown in FIG. That is, the selection unit 443 includes a calculation unit 721, a determination unit 722, and a setting unit 723.

  The calculation unit 721 obtains the fluorescence intensity of the spot area and the fluorescence intensity of the spot peripheral area, calculates the spot reliability based on these, and calculates to correct the reliability in the in-spot area based on the spot reliability. I do. The determination unit 722 performs comparison determination between the corrected in-spot region reliability and a predetermined threshold, and is extracted by the feature amount extraction unit 86 using the learning result by the mechanical learning unit 31 as necessary. The presence / absence of spot reliability is determined based on the result of reliability determination using the feature vector. The setting unit 723 sets a flag indicating whether or not the reactivity information can be used based on the determination result by the determination unit 722. The flag is, for example, reliability information, the number of divided areas when the reaction area is divided into an area containing unnecessary substances and an area containing no unnecessary substances, or designation by the operator (input by the operator) Is input to the unit 29B).

  The in-spot area selection process executed by the selection unit 443 in step S203 of FIG. 22 will be described with reference to the flowchart of FIG.

In step S251, the calculation unit 721 calculates a multiplication value of the reliability and the spot reliability in the target spot area. Confidence in the spot area is ar _i ^j represented by the formula (5), the spot reliability is a value represented by the following formula (7).

  The average value of the spot fluorescence intensity in the denominator on the right side in Expression (7) means, for example, the average value of the fluorescence values of the pixels constituting the spot 738 when the target spot is the spot 738 in FIG. To do. In addition, the average value of the fluorescence intensity around the spot of the molecule on the right side of Equation (7) is the area around the spot 738 (in this example, approximately 1 / between the spot boundary and the spot boundary of the adjacent spot. The average value of the fluorescence values of the pixels constituting the spot periphery 739, which is a range up to the position 2).

  As the average value of the spot peripheral fluorescence intensity is smaller and the average value of the spot fluorescence intensity is larger, the spot reliability of Expression (7) becomes a larger value (a value closer to 1).

Therefore, by multiplying the spot reliability to the reliability ar _i ^j in the spot area (Equation (5)), it is possible to correct the spot area reliability.

  Next, in step S252, the determination unit 722 determines whether or not the intra-spot region reliability corrected in the process of step S251 is smaller than a preset threshold value. If it is determined in step S252 that the corrected in-spot region reliability is smaller than the threshold, the process proceeds to step S256 described later.

  When it is determined in step S252 that the corrected in-spot region reliability is not smaller than the threshold value (that is, equal to or larger than the threshold value), in step S253, the feature amount extraction unit 86 performs mechanical learning. A feature amount necessary for executing the determination of the reliability used is extracted and supplied to the SVM 91 of the mechanical learning unit 31. The extracted feature quantity includes, for example, a probe-target binding intensity matrix, a fluorescence intensity of a spot area, a fluorescence intensity of a spot peripheral area, the number of pixels included in the spot area, a hybridization value, and a spot reliability. Yes, what kind of feature value is used is determined by the feature value extracted for the learning process in the learning mode, that is, the learning process described later with reference to FIG.

  In step S254, the SVM 91 of the mechanical learning unit 31 is supplied in step S253 based on the learning result (spot reliability determination information acquired in step S11 of FIG. 5) stored in the spot removal pattern database 92. The reliability of the corresponding spot is determined from the feature amount.

  SVM is a two-class pattern recognition method based on statistical learning theory. In a feature quantity space obtained by learning, SVM is based on a feature quantity vector (this is called a support vector) that is closest to another class. As described above, it is possible to set an identification boundary at a position where the Euclidean distance becomes the largest (perform linear identification using a separation hyperplane).

  Specifically, the SVM 91 determines that the feature quantity vector based on the feature quantity supplied in step S253 is reliable in the feature quantity space indicated in the spot reliability determination information acquired in step S11 of FIG. Determine which class is not reliable.

  Here, a value having a correlation with the reliability of the spot is used as the feature amount. As described above, the feature amount includes, for example, the probe-target binding intensity matrix, the fluorescence intensity of the spot area, the fluorescence intensity of the spot peripheral area, the number of pixels included in the in-spot area, the hybridization value, and the spot reliability. and so on.

  For example, when the bond strength between the probe and the target is high, the reliability of the measurement using the spot formed by the probe is considered high, but when the bond strength between the probe and the target is low, the probe is configured by the probe. It is considered that the reliability of measurement using a spot is low. In other words, the probe-target bond strength matrix is suitable for use as a feature quantity that serves as a reference for determining spot reliability. As described above, the probe-target binding strength matrix is obtained in association with the DNA chip 11 and is input from the input unit 29B or stored in the expression profile data storage unit 28 as necessary. It is described in the expression profile data.

Further, when the fluorescence intensity of the spot region is very high, the right end portion of the region of the portion 191B through 194B of formula hybridize _e (pf) defining the fluorescence intensity hybridized amount of relationship described with reference to FIG. 9 There is a possibility of corresponding to a low reliability area. Further, when the fluorescence intensity of the spot region is very low, the formula Hybridize _e left end of the area of the portion 191A to 194A of the (pf) defining the fluorescence intensity hybridized amount of relationship described with reference to FIG. 9 There is a possibility of corresponding to a low reliability area. That is, the fluorescence intensity of the spot region is suitable for use as a feature amount that is a reference for determining the spot reliability.

  When the fluorescence intensity in the spot peripheral region is high, for example, as shown in FIG. 29, the debris boundary 736 of the debris region 735 overlaps a part of the spot boundary 732 of the spot 731, and the debris region 735 is There is a possibility that the spot 731 and the entire spot 733 having the spot region 734 adjacent to the right side in the drawing with respect to the spot 731 are covered. In such a case, even if there is no debris that separates the spot areas, the reliability of the fluorescence intensity in the spot areas becomes low due to the debris covering the entire spot. That is, the fluorescent intensity in the spot peripheral region is suitable for use as a feature amount that is a reference for determining the spot reliability.

  In addition, when the number of pixels included in the spot area is small, the area of the spot area is small. Therefore, even if the spot area is debris, the debris area is removed, so that the fluorescence in the spot area is reduced. Strength reliability is not reduced. However, when the number of pixels included in the spot area is large, since the area of the spot area is large, when the spot area is a debris area, the reliability of the fluorescence intensity in the spot area decreases. For example, as shown in FIG. 30, when the solution is poorly dried and the concentration of the solution in the spot becomes non-uniform, or as shown in FIG. 31, draining during washing is insufficient. If a very large area is debris in the spot, such as when the concentration of the solution in the spot is biased, the number of debris in the spot area is small (that is, Even if the number of regions in the spot is small), the reliability of the fluorescence intensity in the spot region is low. In other words, the number of pixels included in the in-spot region or a value representing the area of the in-spot region is suitable for use as a feature amount serving as a reference for determining the spot reliability.

  Also, for example, if the hybridization value is high, there is no large debris in the spot area, or there are few factors that inhibit hybridization, so the reliability of the fluorescence intensity in the spot area is secured to some extent. Although it seems that the hybridization value is low compared to this, not only is the probe provided in the spot region and the target not hybridized, but also, for example, the spot region. It is possible to assume a state in which there is some factor that inhibits hybridization, such as large debris in the water or insufficient dripping of the solution. That is, although some factor has inhibited hybridization, there remains a possibility that the spot was originally a sufficiently hybridized spot. That is, when the hybridization value is high, it can be said that the spot reliability is high, but when the hybridization value is low, the spot reliability may be low. Thus, the hybridized value is suitable for use as a feature quantity that is a reference for determining the spot reliability.

  Needless to say, the spot reliability described using Expression (7) is also suitable as a feature amount used for learning.

  That is, in the learning mode process (described later with reference to FIG. 39), for example, as shown in FIG. 32, expression profile data 744 having link data 745 and spot data 746 is provided to the SVM 91 as feature data. Is done. Further, an image of each spot in the expression profile image 741 corresponding to the expression profile data 744 is displayed on the display unit 29A of the user interface unit 29. Then, the user operates the user interface unit 29 </ b> B to instruct each spot to be the adopted spot 742 or the non-adopted spot 743. The non-adopted spot 743 means a spot that the user visually confirms and determines that it should not be adopted because there is a large debris or the like that reduces the reliability of the measurement result using this spot. On the other hand, the adopted spot 742 means a spot where the user determines that debris or the like does not exist and effective hybridization occurs. In FIG. 32, a spot indicated with an x mark in the drawing represents a spot designated as a non-adopted spot 743, and a spot not marked with an x mark represents a spot designated as an adopted spot 742.

In the learning mode, the user, with respect SVM91, to instruct the adoption spot or Reject spot for each spot, to instruct the training data as a flag f ₁ to f _m of each spot in the expression profile data 744 means.

SVM91 learns the teacher data f ₁ to f _m which is instructed by the user, the hybridization amount ah for each spot in the expression profile data 744 as student data, the relationship between the feature amount data, such reliability ar, The learned result is stored in the spot removal pattern database 92.

  Details of the processing in the learning mode will be described later with reference to FIGS.

  As illustrated in FIG. 32, when the spot removal pattern database 92 is constructed in advance by performing learning for many spots in the learning mode, as shown in FIG. The feature amount extracted from the expression profile image 751 having the following data is provided to the SVM 91, and the SVM 91 is referred to the spot removal pattern database 92 to determine whether the data of each spot 752 of the expression profile image 751 is adopted or not. Can be made.

  That is, the SVM 91 refers to the spot removal pattern database 92, and based on the feature amount data extracted corresponding to the spot 752 of the expression profile image 751, each spot is selected as the adopted spot 752A or the non-adopted spot 752B. Judge as. In FIG. 33, spots marked with an x mark in the figure are spots determined as non-adopted spots 752B, and spots not marked with an x mark in the figure are determined as adopted spots 752A. It is.

  The SVM is described in detail in Nello Cristianini, John Shawe-Taylor, An Introduction to Support Vector Machines and other kernel-based learning methods, Cambridge.

  In step S <b> 255, the determination unit 722 determines whether the in-spot area has been adopted in the reliability determination by mechanical learning.

  When it is determined in step S252 that the corrected in-spot area reliability is smaller than the threshold value, or in step S255, it is determined that the in-spot area reliability is not adopted in the reliability determination by mechanical learning. In the case where the discard flag based on reliability is set for the data in the spot area, in step S256, the setting unit 723 sets the discard flag based on reliability in the area data in the spot. Returning to step S203 of FIG. 22, the process proceeds to step S204. That is, in this case, the data in the spot area that is currently processed is not reliable. The data in the spot area is basically not used thereafter.

  On the other hand, when it is determined in step S255 that the target spot area is adopted in the reliability determination by mechanical learning (when the discard flag based on reliability is not set), the data of the spot area is According to the machine learning, since there is reliability, the process of step S256 is skipped. That is, the discard flag based on reliability is not set for the in-spot region, and the process returns to step S203 in FIG. 22 and proceeds to step S204.

  In other words, even when the corrected in-spot reliability is lower than a predetermined threshold, if it is determined by the SVM 91 that there is a spot reliability, the reliability discard flag is not set in the spot area data. .

  As the mechanical learning, a neural network or the like can be employed in addition to SVM.

  In this way, instead of actually destroying the data itself, it is possible to set the discard flag for each spot area so that the data in the spot area is not used as necessary. Of course, conversely, this data can be used by intentionally ignoring this flag.

  It should be noted that the discard flag may be set by the user's manual operation (designated by the operator) in the spot area selection processing in step S203. For example, in the example of FIG. 29, the debris boundary 736 of the debris region 735 overlaps with a part of the spot boundary 732 of the spot 731, and the debris region 735 is adjacent to the spot 731 and the in-spot region 734 adjacent to the right side in FIG. It covers the entire spot 733 having In a state in which such an image is displayed on the display unit 29A, the user operates the input unit 29B of the user interface unit 29, so that the in-spot areas included in the spots 731 and 733 are discarded due to reliability. A flag can be set.

  As described above, after the in-spot region selection process in step S203 is performed, in step S204, the output unit 444 in FIG. 21 performs the hybridized value for each spot, which will be described later using the flowchart in FIG. Execute reliability output processing. In order to execute this processing, the output unit 444 in FIG. 21 has a functional configuration as shown in FIG. 34, for example. As shown in the figure, the output unit 444 includes a determination unit 751, a generation unit 752, a calculation unit 753, a selection unit 754, and a setting unit 755.

  The determination unit 751 performs a comparison determination between the number of divided spots and a threshold value. The generation unit 752 generates a combination of in-spot areas. The calculation unit 753 calculates the amount of hybridization and the reliability in spot units. The selection unit 754 selects a combination having the maximum reliability. The setting unit 755 sets various flags in the expression profile data.

  Next, with reference to the flowchart of FIG. 35, the output processing of the hybridization value and reliability in units of spots executed by the output unit 444 will be described.

  In step S301, the determination unit 751 determines whether the number of divided spots is smaller than a predetermined threshold value set in advance. That is, among the plurality of spots to which the corresponding probes are fixed, the number of spots divided into the spot-internal regions is counted by the process in step S201 in FIG.

If it is determined in step S301 that the number of divided spots is smaller than the threshold value, in step S302, the generation unit 752 generates a combination of in-spot areas. FIG. 36 is a diagram for explaining this combination. As shown in the figure, in this example, n experiments were performed, and in each experiment, one substrate was measured. That is, FIG. 36 shows _n substrates such as the substrate 11A ₁ in the first experiment, the substrate 11A ₂ in the second experiment, and the substrate 11A _{n in the nth} experiment. . A corresponding probe is fixed to a spot at a corresponding position on each substrate. Actually, the corresponding spot does not need to be a spot in a different experiment, and may be a different substrate in the same experiment, or a spot at a different position on the same substrate.

For example, the spot 771 ₁ in the upper right in the figure in the substrate 11A _1, top right spot 771 ₂ of the substrate 11A _2, and the upper right spot 771 _n of the substrate 11A _n is a corresponding spot, respectively. The spot 771 ₁ of the substrate 11A ₁ is divided into three regions. In FIG. 36, each spot area is indicated by the letter R. The subscript number of the letter R represents the substrate (experiment number), and the superscript number represents the number of the in-spot region (divided region) at that spot. For example, the spot 771 ₁ is divided into three spots in the region R ₁ ^1, R ₁ ^2, R ₁ ^3. The hybridizing amount of the in-spot region R ₁ ¹ (represented by the formula (4)) is ah ₁ ¹ , the hybridizing amount of the in-spot region R ₁ ² is ah ₁ ^2, and the in-spot region R _The amount of hybridization of ₁ ³ is ah ₁ ³ . The area within the spot of the spot 771 ₂ of the substrate 11A ₂ is one R ₂ ^{1 and} the amount of hybridization is ah ₂ ¹ . Further, the spot 771 _n of the substrate 11A _{n in} the n-th experiment is divided into two regions R _n ¹ and R _n ^{2 in} the spot, and the amount of each hybridization is ah _n ¹ or ah _n ^2. Yes.

When combining the in-spot regions, one in-spot region is selected from the spots on each substrate. For example, as shown in FIG. 37, the spot 771 ₁ includes ^three regions, that is, the in-spot regions R ₁ ¹ , R ₁ ² , and R ₁ ³ , and one of them is selected. Since only one in-spot region R ₂ ¹ exists in the spot 771 ₂ , only this region is selected in this spot. In the spot 771 _n , since there are two areas of the spot areas R _n ¹ and R _n ² , one of them is selected. As a result, for example, when n = 3, 6 (= 3 × 1 × 2) combinations are possible.

Specifically, R ₁ ¹ -R ₂ ¹ -R ₃ ¹ , R ₁ ² -R ₂ ¹ -R ₃ ¹ , R ₁ ³ -R ₂ ¹ -R ₃ ¹ , R ₁ ¹ -R are used as the in-spot regions. Six combinations of ₂ ¹ -R ₃ ² , R ₁ ² -R ₂ ¹ -R ₃ ² , R ₁ ³ -R ₂ ¹ -R ₃ ² are conceivable.

  In step S303, the calculation unit 753 calculates the amount of hybridization and the reliability in units of spots. Specifically, the following equations (8) and (9) are calculated.

  In Expressions (8) to (10), i represents the number of experiments, and represents the experiment number in FIG. 36 (i = 1, 2,..., N).

pn _i represents the number of pixels in the spot area in the i-th experiment in the spot area constituting the combination. Sn represents the total number of pixels in the spot area constituting the combination. For example, it represents the sum of the number of pixels of the in-spot regions R ₁ ¹ , R ₂ ¹ , R ₃ ¹ constituting one combination in FIG.

Ah _i of equation (8) represents a hybridized amount of the i-th spot area within the spot region constituting the combination, specifically, the formula (4). Moreover, ar _i of formula (9) represents the reliability of the i-th spot area within the spot region constituting the combination is specifically represented by the formula (5).

  That is, Sh in Expression (8) represents the amount of hybridization of the combination in the spot area, and Sr in Expression (9) represents the reliability of the combination in the spot area.

  Next, in step S304, the selection unit 754 selects a combination having the maximum reliability (minimum variance). That is, the maximum one is selected from the reliability Sr of each combination calculated in step S303, and the process returns to step S204 in FIG. 22 and proceeds to step S40 in FIG.

For example, in the example of FIG. 37, since combinations of six in-spot regions are conceivable, combinations of in-spot regions R ₁ ¹ -R ₂ ¹ -R ₃ ^{1 to} R ₁ ³ -R ₂ ¹ -R ₃ Among the reliability Sr _{1 to} Sr ₆ in each of the ^two , the combination of the maximum values is obtained. For example, when Sr ₁ is the highest among the reliability levels Sr _{1 to} Sr ₆ , the combination of the in-spot regions R ₁ ¹ -R ₂ ¹ -R ₃ ¹ is selected.

  If it is determined in step S301 that the number of divided spots is equal to or greater than the threshold value, the processes in steps S302 to S304 are skipped. In step S305, the setting unit 755 sets a reliability discard flag, and the process returns to step S204 in FIG. 22 and proceeds to step S40 in FIG.

That is, for example, when n ₂ spots larger than the threshold n ₁ among the corresponding n spots 771 _{1 to} 771 _n shown in FIG. There are many spots, and there may be some problem in the measurement environment itself. Therefore, in such a case, it is considered that the reliability of the data of the spot itself has already been impaired, so a discard flag based on the reliability is set for such a spot.

FIG. 38 shows the format of the expression profile data supplied and stored in step S14 of FIG. 5 from the output unit 146 of FIG. 3 to the storage unit 147 (expression profile data storage unit 28 of FIG. 1). As shown in the figure, the expression profile data 801 includes the number of spots N, the number n of repeated experiments, and data of each combination. When there is one spot of one probe in one experiment, the number of spots N is equal to the number n of repeated experiments. The data of each of the m combinations includes the amount of hybridization Sh _{1 to} Sh _m , the reliability Sr _{1 to} Sr _m , the flags f _{1 to} f _{m of each of} the _{first to mth} combinations, and the spot data constituting the combination The pointers P _{1 to} P _m representing the positions of For example, the pointer P ₁ is a pointer of the first combination and represents the storage position of the link data 802 of the first combination. The link data 802 includes the number of spots N and pointers P _{11 to} P _1N to the first to Nth spot data.

For example, the pointer P ₁₁ represents the storage position of the spot data of the first spot of the first combination.

  The spot data 803 has the number of spot areas an constituting the spot an, the selected spot area number selected-no, and the spot reliability. This spot reliability is a value represented by Expression (7).

The spot data 803 further includes a hybridization amount ah, a reliability ar, a pixel number pn, and a flag f for each in-spot region constituting the spot. FIG. 38 shows the amounts of hybridization ah ₁ ¹ , ah ₁ ² , ah ₁ ³ , reliability ar ₁ ¹ , ar ₁ ² , ar ₁ ³ , pixels in the _{first to} ^third spot regions of the first spot. Numbers pn ₁ ¹ , pn ₁ ² , pn ₁ ³ , and flags f ₁ ¹ , f ₁ ² , and f ₁ ³ are shown.

  Next, the learning process for obtaining the spot reliability determination information acquired in step S11 of FIG. 5 will be described. The learning process can be executed in the learning mode of the biological information processing apparatus 1 described with reference to FIG.

  The biological information processing apparatus that performs quantitative measurement of the gene expression level by performing determination of the reliability of the spot using the biological information processing apparatus that has performed the learning process and the spot reliability determination information that is the learning result thereof The devices may be the same device or different devices.

  Further, the collection of the feature amount data and the teacher data can also be executed by a device different from the device that performs the learning process. That is, more accurate determination is performed by collecting feature amount data used for learning and teacher data in a plurality of devices, and performing learning in the SVM 91 of one device based on a large amount of sample data. Can be obtained.

  Further, in the biological information processing apparatus 1 different from the biological information processing apparatus 1 including the SVM 91 for which the spot reliability determination information is obtained, the spot reliability determination information is stored in the spot removal pattern database 92, and in the determination mode, the SVM 91 By making the spot reliability determination based on the stored spot reliability determination information, a common learning result can be used in many devices. In comparison with this, it is possible to prevent the occurrence of variations in the inspection results by the inspector.

  The learning process will be described with reference to the flowchart of FIG.

  In step S351, the hybridization amount estimation unit 24 acquires image data. That is, the target used for the learning process is adjusted and hybridized by the same process as the process 1 of the experimental process described with reference to FIG. 4, and is included in the fluorescence intensity data described with reference to FIG. Expression profile image data is acquired, or expression profile image data included in the expression profile data described with reference to FIG. 8 stored in the expression profile data storage unit 28 is acquired.

  In step S352, a feature amount extraction process to be described later with reference to FIG. 40 is executed.

  In step S353, the user interface unit 29 displays the acquired expression profile image on the display unit 29A. In step S354, the input unit 29B receives the presence / absence of reliability for each spot area from the user, specifically, the non-adopted spot 743 indicated by a cross in the drawing described with reference to FIG. An instruction input is received and supplied to the SVM 91.

  In step S355, the SVM 91 performs a learning process using the presence / absence of reliability for each spot area input in step S373 as teacher data.

  SVM is a pattern recognition method of two classes based on statistical learning theory, and supports each class based on a feature vector (this is called a support vector) that is closest to another class in the learning data. An identification boundary is set at a position where the Euclidean distance from the vector is the largest, and pattern recognition is performed based on the identification boundary. Specifically, the SVM 91 is based on the feature amount vector obtained in the learning mode and the presence or absence of reliability as a teacher signal (for example, information on the adopted spot 742 or the non-adopted spot 743 described with reference to FIG. 32). By performing the learning process, it is possible to set an identification boundary that divides each class with reliability and without reliability in the feature amount space. In the determination mode, as described above, the SVM 91 determines the reliability of the corresponding spot based on the spot reliability determination information as the learning result, that is, based on the identification boundary in the feature amount space and the supplied feature amount vector. Presence / absence can be determined.

  In step S356, the SVM 91 supplies the learning result to the spot removal pattern database 92 for storage, and the process is terminated.

  By such processing, learning processing using SVM is performed, and spot reliability determination information as a learning result, that is, class identification boundaries (boundaries with or without reliability) in the feature amount space are obtained and stored. be able to. In the determination mode, based on the spot reliability determination information, the presence / absence of reliability is determined based on the feature vector corresponding to the spot (the supplied feature vector is determined with respect to the obtained identification boundary). It is determined whether the class is classified as reliable or not reliable).

  Next, the feature quantity extraction process executed in step S352 of FIG. 39 will be described with reference to the flowchart of FIG.

  In step S381, the excitation light intensity estimation unit 81 inputs image information. Specifically, the excitation light intensity estimation unit 81 receives supply of fluorescence intensity data obtained by the process described with reference to FIG. 4 from the fluorescence intensity acquisition unit 22 or is stored in the expression profile data storage unit 28. Is supplied with expression profile data. That is, the learning process can be performed using the above-described experimental process processing device 131 by acquiring the fluorescence intensity of the hybridized DNA chip 11 with the target adjusted in the experimental process 1. It is also possible to use expression profile data obtained by other devices.

  In steps S382 to S387, basically the same processing as in steps S32 to S37 of FIG. 6 is executed.

  That is, when the excitation light intensity estimation unit 81 is supplied with the expression profile data stored in the expression profile data storage unit 28 described with reference to FIG. 8, the input image information includes excitation light intensity information. It is determined whether or not. When it is determined that the excitation light intensity information does not exist, the excitation light intensity estimation unit 81 uses the fluorescence intensity-hybridization amount conversion expression stored in the fluorescence intensity-hybridization amount conversion expression storage unit 30 to perform excitation. The light intensity is estimated.

  When it is determined that the expression profile data is not supplied, that is, when the fluorescence intensity data supplied from the light intensity acquisition unit 22 described with reference to FIG. It is determined whether the input image information is image information of an image captured with a plurality of excitation light intensities. When it is determined that the image information is an image captured with a plurality of excitation light intensities, the creation unit 82 determines the amount of hybridization based on the fluorescence intensity (hybridize (pf)) (the above-described equation (1)). Create

  Since the image processing unit 83 performs the image processing described with reference to the flowchart of FIG. 11, the debris region straddling the spot boundary is removed from the image of the DNA chip 11, and the image is decomposed into images for each spot. .

  In step S388, the verification unit 84 executes a process for verifying hybridization in the same manner as the process in step S35 of FIG. Specifically, the fluorescence values of the hybridization verification probes 111 and 114 are measured, and if the fluorescence values are equal to or higher than a preset reference value, it is verified that correct hybridization processing is performed. be able to.

  In step S389, the dividing unit 441 of the hybridization amount calculating unit 85 divides the spot area, similarly to the process in step S201 of FIG. For example, as described with reference to FIG. 23, the spot region is divided into an in-spot region including debris and an in-spot region not including debris, and debris 464 and 465 having no area are debris. It is removed because it is clear.

In step S390, the reliability calculation unit 442 of the hybridization amount calculation unit 85 calculates the hybridization value and the reliability for each spot area, as in the process of step S202 of FIG. Hybridization of the spot area i soybean value ah _i ^j (reactivity information) and a spot in the area reliability ar _i ^j (reliability information) is represented by the above-mentioned equations (4) Equation (5).

  By this processing, for example, the spot reliability and the amount of hybridization (total value of the amount of hybridization in the spot area) are extracted as feature amounts. Since these feature quantities have a correlation with the reliability of the spot, they are suitable as the feature quantities used for the reliability determination.

  In step S391, the image processing unit 83 acquires the fluorescence intensity of the spot area. Specifically, the image processing unit 388 acquires the fluorescence intensity of each spot such as the spot 731, the spot 733, and the spot 738 described with reference to FIG. 29 and holds them together with the position information of each spot.

If the fluorescence intensity of the spot region is very high, for example, due for excitation light intensity is too strong, the right end of the formula hybridize _e (pf) defining a hybridized amount of relationship described with reference to FIG. 9 There is a high possibility of corresponding to a low reliability state corresponding to the region portions 191B to 194B. On the other hand, when the fluorescence intensity of the spot area is very dark, for example, the hybridization described with reference to FIG. 9 is performed because the excitation light intensity is too weak, the solution is not uniformly dropped, or is not washed uniformly. It is likely to correspond to the state of low reliability corresponding to the portion 191A or 194A of the area of the left end of the formula Hybridize _e defines the relation (pf) amounts. That is, the fluorescence intensity of the spot region is suitable as a feature amount used for reliability determination.

  In step S392, the image processing unit 83 acquires the fluorescence intensity of the spot peripheral area. Specifically, the image processing unit 388 acquires the fluorescence intensity of each spot peripheral region such as the spot periphery 737 and the spot periphery 739 described with reference to FIG. 29, and stores the fluorescence intensity together with the position information of each spot periphery region. .

  When the fluorescence intensity in the spot peripheral region is very high, for example, as shown in FIG. 29, the solution may flow outside the spot range, and the debris region may cover the entire spot or spots. There is sex. This state can be easily determined if the reliability is low when visually determined, but if the reliability is automatically determined by detecting debris, the density of the spot area is uniform. Therefore, there is a possibility that the fluorescence intensity is erroneously detected in a state where there is no debris. However, it is possible to accurately determine the spot reliability in the case as shown in FIG. 29 by using the fluorescence intensity of the spot peripheral region as a feature amount (learning as a teacher data when determined visually). It is made like that. Thus, the fluorescence intensity in the spot peripheral region is suitable as a feature amount used for reliability determination.

  In step S393, the image processing unit 83 acquires the number of pixels included in each spot area.

  In a certain spot, when the number of pixels contained in the spot area is large (the area of the spot area is large) even if the number of areas in the spot is small, the spot is shown in FIG. 30 or FIG. 31, for example. The probability of large debris is high, so it is considered that the reliability of the spot is low. On the other hand, in a certain spot, the number of pixels included in the spot area is small (the area of the spot area is small). When there are many, it is thought that the debris which exists in the spot is only small. Even if there is a certain number of small debris in the spot area, it is considered that the reliability of this spot is high by executing the debris removal process. As described above, information on the number of pixels included in the spot area is suitable as a feature amount used for reliability determination.

  In step S394, the feature amount extraction unit 86 receives information indicating the feature amount extracted in the above processing, and is input by the input unit 29B or stored in the expression profile data storage unit 28 as necessary. The acquired probe-target bond strength matrix is generated to generate feature data, and the process returns to step S352 in FIG. 39 and proceeds to step S353.

  Through such processing, feature amounts are extracted and feature amount data is generated. Learning for determining the reliability for each spot is performed using the feature amount data.

  In addition, here, pairs of expression profile image data obtained using different excitation light intensities are used to estimate the excitation light intensity, or to determine the hybridize (pf) that uniquely determines the amount of hybridization from the fluorescence intensity. However, it is possible to increase the feature amount for the teacher data by using the expression profile image data obtained by using different excitation light intensities for learning.

  That is, as shown in FIG. 41, the expression profile image data 741-1-1 and 741-2-1 obtained when the DNA chip 11 in the same hybridized state is irradiated with a plurality of excitation lights. , Each feature amount data is extracted, and expression profile data 744-1 and 744-2 including spot data 746 in which they are described are generated. In many cases, the feature amounts described in the expression profile data 744-1 and 744-2 obtained when different excitation light intensities are irradiated have different values.

  However, since the expression profile image data 741-1-1 and 741-2-1 indicate the fluorescence intensity with respect to the DNA chip 11 in the same hybridized state, the hybridization distribution within the spot at the same spot position. The amount is the same. That is, the adopted spot 742 or the non-adopted spot 743 as teacher data is similarly given to the same spot in the expression profile image data 741-1-1 and 741-2-1.

  That is, by using pairs of expression profile image data obtained by using different excitation light intensities, it becomes possible to associate many feature vectors with teacher data of one spot, thereby improving the learning effect. be able to.

  In this way, learning can be effectively performed by changing the acquisition conditions of the expression profile image data so that different feature amount data can be obtained in the same hybridizing distribution amount.

  In addition, here, the case where the expression profile image data is obtained using two different excitation light intensities has been described, but also when the expression profile is obtained with two or more different excitation light intensities, Similarly, it goes without saying that it is possible to effectively learn by obtaining a large amount of feature data.

  As described above, the embodiment in the case of measuring the hybridization of the DNA chip has been described. However, the present invention is not limited to the DNA chip. In the case of measuring whether various biological materials are biologically bound to other predetermined biological materials. It is possible to apply to.

  The series of processes described above can be executed by hardware or can be executed by software. In this case, for example, the biological information processing apparatus 1 includes a personal computer 901 as shown in FIG.

  42, a CPU (Central Processing Unit) 921 executes various processes according to a program stored in a ROM (Read Only Memory) 922 or a program loaded from a storage unit 928 to a RAM (Random Access Memory) 923. To do. The RAM 923 also appropriately stores data necessary for the CPU 921 to execute various processes.

  The CPU 921, ROM 922, and RAM 923 are connected to each other via a bus 924. An input / output interface 925 is also connected to the bus 924.

  The input / output interface 925 includes an input unit 926 including a keyboard and a mouse, a display including a CRT (Cathode Ray Tube) and an LCD (Liquid Crystal display), an output unit 927 including a speaker, and a hard disk. A communication unit 929 including a storage unit 928 and a modem is connected. The communication unit 929 performs communication processing via a network including the Internet.

  A drive 930 is also connected to the input / output interface 925 as necessary, and a removable medium 931 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory is appropriately attached, and a computer program read from them is loaded. It is installed in the storage unit 928 as necessary.

  When a series of processing is executed by software, a program constituting the software executes various functions by installing a computer incorporated in dedicated hardware or various programs. For example, a general-purpose personal computer is installed from a network or a recording medium.

  As shown in FIG. 42, this recording medium is distributed to provide a program to the user separately from the apparatus main body, and includes a magnetic disk (including a floppy disk) on which the program is recorded, an optical disk (CD- It is not only composed of removable media 931 consisting of ROM (compact disk-read only memory), DVD (including digital versatile disk), magneto-optical disk (including MD (mini-disk)), or semiconductor memory. The program is configured by a ROM 922 in which a program is recorded and a hard disk included in the storage unit 928, which is provided to the user in a state of being incorporated in the apparatus main body in advance.

  In the present specification, the step of describing the program recorded on the recording medium is not limited to the processing performed in chronological order according to the described order, but is not necessarily performed in chronological order. It also includes processes that are executed individually.

  In addition, in this specification, the system means a logical collection of a plurality of devices (or functional modules that realize a specific function), and whether each device or functional module is in a single casing. It doesn't matter whether or not.

  The embodiment of the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the gist of the present invention.

It is a block diagram showing the example of a structure of the biometric information processing apparatus as embodiment of this invention. It is a perspective view showing the structural example of a DNA chip. It is a block diagram showing the structural example of an experimental process processing apparatus. It is a flowchart explaining the process 1 of an experiment process. It is a flowchart explaining the process 2 of an experiment process. It is a flowchart explaining an expression level estimation process. It is a figure which shows the example of an input image format. It is a figure which shows the example of an input image format. It is a figure showing the relationship between fluorescence intensity and the amount of hybridization. It is a block diagram showing the example of a structure of an image process part. It is a flowchart explaining an image process. It is a figure explaining extraction of the expression profile picture by a template. It is a figure explaining the process of a filter. It is a figure explaining the production | generation process of a filter. It is a figure explaining the process by a trend filter. The It is a figure which shows the example of debris It is a figure showing the example of the binarized image. It is a figure showing the state which made the debris boundary thick. It is a figure explaining the debris area | region which cross | intersects a spot boundary. It is a figure explaining the boundary of a debris area | region and the area | region in a spot. It is a block diagram showing the example of a structure of the amount calculation part of hybridization. It is a flowchart explaining the calculation process of the amount of hybridization and reliability. It is a figure showing the example of the image of the spot containing debris. It is a figure showing the principle of determination of being inside a debris area | region. It is a figure explaining the area | region in a spot. It is a figure explaining reliability confidence. It is a block diagram showing the example of a structure of a selection part. It is a flowchart explaining the area | region selection process in a spot. It is a figure explaining a spot periphery. It is a figure explaining the example of the debris of a big area | region. It is a figure explaining the example of the debris of a big area | region. It is a figure explaining learning of SVM. It is a figure explaining determination of SVM. It is a block diagram showing the example of a structure of an output part. It is a flowchart explaining the output process of the hybridization value and reliability in a spot unit. It is a figure explaining the structure of the spot by experiment of multiple times. It is a figure which shows the example of the combination of the area | region in a spot by a multiple times experiment. It is a figure showing the structural example of expression profile data. It is a flowchart explaining a learning process. It is a flowchart explaining a feature-value extraction process. It is a figure explaining extraction of the feature-value using the expression profile image data obtained by different excitation light intensity | strength. And FIG. 11 is a block diagram illustrating a configuration example of a personal computer.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Biological information processing apparatus, 11 DNA chip, 21 Pickup part, 22 Fluorescence intensity acquisition part, 23 Excitation light intensity calculation part, 24 Hybridization amount estimation part, 25 Expression amount calculation part, 28 Expression profile data storage part, 29 User interface Unit, 29A display unit, 29B input unit, 30 fluorescence intensity-hybridization amount conversion type storage unit, 31 mechanical learning unit, 81 excitation light intensity estimation unit, 82 creation unit, 83 image processing unit, 84 verification unit, 85 hybrid Soybean amount calculator, 91 SVM, 92 spot removal pattern database

Claims (24)

In a biological information processing apparatus for measuring a state of a biological reaction between a first biological material fixed in a reaction region provided on a substrate and a second biological material that performs a biological reaction with respect to the first biological material ,
Image input means for receiving input of image information of the reaction region;
Feature amount extraction means for extracting the feature amount of the image information of the reaction region input by the image input means;
Storage means for storing determination criterion information that is a determination criterion for the presence or absence of reliability when the image information of the reaction region is used for measurement;
Judgment to determine the presence or absence of reliability when the image information of the reaction region having the feature quantity extracted by the feature quantity extraction means is used for measurement based on the judgment criterion information stored by the storage means Means,
Based on a determination result by the determination means, and a calculation means for calculating reactivity information representing a state of a biological reaction between the first biological material and the second biological material in the reaction region,
The determination criterion information includes a correlation between information indicating presence / absence of reliability when the image information of the reaction region is used for measurement using image information for learning and a feature amount of the image information of the reaction region. It was obtained by learning,
The feature amount is a value having a correlation with the reliability of the image information of the reaction region. The biological information processing apparatus according to claim 1, wherein the determination criterion information is obtained by SVM. The biological information processing apparatus according to claim 1, wherein the feature amount extracted by the feature amount extraction unit includes a fluorescence intensity obtained when the reaction region is irradiated with excitation light. The biological information processing apparatus according to claim 1, wherein the feature amount extracted by the feature amount extraction unit includes a fluorescence intensity obtained when the region around the reaction region is irradiated with excitation light. The living body according to claim 1, wherein the feature amount extracted by the feature amount extraction unit includes a value corresponding to an area of an image corresponding to a substance that is an obstacle to perform measurement included in the reaction region. Information processing device. The biometric information according to claim 1, wherein the feature quantity extracted by the feature quantity extraction unit includes a value corresponding to a bioreaction quantity of the first biological material and the second biological material in the reaction region. Processing equipment. The feature amount extracted by the feature amount extraction unit includes information indicating ease of biological reaction between the first biological material and the second biological material in the reaction region. Biological information processing device. Display means for displaying the image information for learning input by the image input means;
Teacher information input means for receiving, as teacher information, information indicating the presence or absence of reliability of the image information of the reaction region by a user who refers to the image information displayed by the display means;
Learning means for obtaining the determination criterion information by learning the correlation between the teacher information input by the teacher information input means and the feature quantity extracted by the feature quantity extraction means;
The biological information processing apparatus according to claim 1, wherein the storage unit stores the determination criterion information obtained by the learning unit. The first biological substance and the second biological substance are genes having base sequences complementary to each other or substances derived therefrom,
The biological information processing apparatus according to claim 1, wherein the reactivity information is information on hybridization between the first biological material and the second biological material. The biological information according to claim 9, wherein the hybridization information is information that is uniquely determined based on a function from a fluorescence intensity obtained by hybridizing the first biological material and the second biological material. Information processing device. A biological information processing apparatus for measuring a state of a biological reaction between a first biological material fixed in a reaction region provided on a substrate and a second biological material that is biologically reactive with the first biological material. In the biological information processing method,
An image input control step for controlling input of image information of the reaction region;
A feature amount extraction step of extracting a feature amount of the image information of the reaction region whose input is controlled by the processing of the image input control step;
An acquisition control step for controlling acquisition of determination criterion information that is a determination criterion for the presence or absence of reliability when the image information of the reaction region is used for measurement;
Reliability in the case where image information of the reaction region having the feature amount extracted by the feature amount extraction step is used for measurement based on the determination criterion information whose acquisition is controlled by the acquisition control step processing A determination step for determining the presence or absence of a degree;
A calculation step for calculating reactivity information representing a state of biological reaction between the first biological material and the second biological material in the reaction region based on a determination result obtained by the determination step;
The determination criterion information includes a correlation between information indicating presence / absence of reliability when the image information of the reaction region is used for measurement using image information for learning and a feature amount of the image information of the reaction region. It was obtained by learning,
The feature amount is a value having a correlation with the reliability of the image information of the reaction region. A computer executes a process of measuring a state of a biological reaction between a first biological material fixed in a reaction region provided on a substrate and a second biological material that reacts with the first biological material. In the program to let
An image input control step for controlling input of image information of the reaction region;
A feature amount extraction step of extracting a feature amount of the image information of the reaction region whose input is controlled by the processing of the image input control step;
An acquisition control step for controlling acquisition of determination criterion information that is a determination criterion for the presence or absence of reliability when the image information of the reaction region is used for measurement;
Reliability in the case where image information of the reaction region having the feature amount extracted by the feature amount extraction step is used for measurement based on the determination criterion information whose acquisition is controlled by the acquisition control step processing A determination step for determining the presence or absence of a degree;
A calculation step for calculating reactivity information representing a state of biological reaction between the first biological material and the second biological material in the reaction region based on a determination result obtained by the determination step;
The determination criterion information includes a correlation between information indicating presence / absence of reliability when the image information of the reaction region is used for measurement using image information for learning and a feature amount of the image information of the reaction region. It was obtained by learning,
The feature amount is a value having a correlation with reliability of image information in the reaction area.   A recording medium on which the program according to claim 12 is recorded. In order to measure the state of the biological reaction between the first biological material fixed in the reaction region provided on the substrate and the second biological material that reacts with the first biological material, the reaction In a learning device that executes a learning process necessary for determining the presence or absence of reliability of a region,
Image input means for receiving input of image information of the reaction region;
Display means for displaying the image information input by the image input means;
Teacher information input means for receiving, as teacher information, information indicating the presence or absence of reliability of the image information of the reaction region by a user who refers to the image information displayed by the display means;
Feature amount extraction means for extracting a feature amount having a correlation with the reliability of the image information of the reaction region for each reaction region;
A learning device comprising: learning means for learning a correlation of the feature quantity extracted by the feature quantity extraction means with respect to the information indicating the reliability of the reaction region input by the teacher information input means. The learning apparatus according to claim 14, wherein the feature amount extracted by the feature amount extraction unit includes a fluorescence intensity obtained when the reaction region is irradiated with excitation light. The learning apparatus according to claim 14, wherein the feature amount extracted by the feature amount extraction unit includes a fluorescence intensity obtained when the region around the reaction region is irradiated with excitation light. 15. The learning according to claim 14, wherein the feature amount extracted by the feature amount extraction unit includes a value corresponding to an area of an image corresponding to a substance that is an obstacle to perform the measurement, which is included in the reaction region. apparatus. The learning device according to claim 14, wherein the feature amount extracted by the feature amount extraction unit includes a value corresponding to a biological reaction amount of the first biological material and the second biological material in the reaction region. . The feature amount extracted by the feature amount extraction unit includes information indicating ease of biological reaction between the first biological material and the second biological material in the reaction region. Learning device. The learning device according to claim 14, wherein the learning unit performs a learning process using SVM. The image input means inputs a plurality of pieces of image information on the same substrate obtained when the intensity of excitation light irradiated on the substrate is different in order to obtain the image information,
The feature amount extraction means extracts the feature amount for each of the reaction regions of each of a plurality of pieces of image information obtained when the intensity of excitation light is different,
The teacher information input means receives an input of one class of teacher information for one reaction region by a user referring to the image information displayed by the display means,
The learning unit is configured to obtain each of the feature amounts obtained when the intensity of excitation light extracted by the feature amount extraction unit is different from the one class of teacher information input by the teacher information input unit. The learning apparatus according to claim 14, wherein correlation is learned. In order to measure the state of the biological reaction between the first biological material fixed in the reaction region provided on the substrate and the second biological material that reacts with the first biological material, the reaction In the learning control method of the learning device that executes the learning process necessary for determining the presence or absence of the reliability of the region,
An image input control step for controlling input of image information of the reaction region;
A display control step for controlling display of the image information whose input is controlled by the processing of the image input control step;
Teacher information input control step for controlling input of information indicating the reliability of the image information of the reaction region as teacher information by a user who refers to the image information whose display is controlled by the processing of the display control step. When,
For each reaction region, a feature amount extraction step for extracting a feature amount having a correlation with the reliability of the image information of the reaction region;
A learning step for learning a correlation of the feature amount extracted by the feature amount extraction step with respect to the information indicating the reliability of the reaction region whose input is controlled by the teacher information input control step; Learning control method. In order to measure the state of the biological reaction between the first biological material fixed in the reaction region provided on the substrate and the second biological material that reacts with the first biological material, the reaction A program for causing a computer to execute a learning process necessary for determining whether or not a region has reliability,
An image input control step for controlling input of image information of the reaction region;
A display control step for controlling display of the image information whose input is controlled by the processing of the image input control step;
Teacher information input control step for controlling input of information indicating the reliability of the image information of the reaction region as teacher information by a user who refers to the image information whose display is controlled by the processing of the display control step. When,
For each reaction region, a feature amount extraction step for extracting a feature amount having a correlation with the reliability of the image information of the reaction region;
A learning step for learning a correlation of the feature amount extracted by the feature amount extraction step with respect to the information indicating the reliability of the reaction region whose input is controlled by the teacher information input control step; Including programs.   A recording medium on which the program according to claim 23 is recorded.

Images