JP4578201B2 - Gene estimation apparatus, gene estimation method and program thereof - Google Patents

Description

  The present invention relates to a gene estimation apparatus, a gene estimation method and a program for estimating a gene related to the pharmacological activity of a specific compound.

  As a technique for estimating which gene a compound has pharmacological activity, a technique for estimating a gene that generates the protein from compound structure information and searching for a target protein is disclosed. As databases published by NCI (National Cancer Institute), there are A-Matrix, which is information on pharmacological activity values for compounds of cancer cells, and T-Matrix, which is information on gene expression patterns of cancer cells. Disclosure results of A-Matrix and T-Matrix clustering and techniques for finding correlation matrix (AT-Matrix) from AT-Matrix from A-Matrix and T-Matrix and finding the relationship from AT-Matrix (For example, see Non-Patent Document 1).

Uwe Scherf and 16 others, “a gene expression database for the molecular pharmacology of cancer”, Nature Genetics, Vol. 24, No. 3, March 1, 2000, p.236-244.

  However, since the above-described method requires a target protein, there is a problem that a gene cannot be estimated when the target protein is unknown. Further, the method of Non-Patent Document 1 has a problem that it is impossible to estimate the relationship between a gene with unknown pharmacological activity and a gene.

  The present invention has been made in consideration of the above-mentioned circumstances, and even if the pharmacological activity of the target protein or the target compound is unknown, a gene that can be expected to be related to the pharmacological activity of the target compound can be estimated. It is an object to provide a gene estimation device, a gene estimation method, and a program thereof.

The present invention has been made to solve the above-described problems, and in the gene estimation apparatus according to the present invention, a compound that is information on the activity value of the pharmacological activity of a plurality of types of first compounds for each of a plurality of types of cells. Compound information storage means for storing information and compound structure information that is information relating to the chemical structure of the first compound, and expression information including expression levels of a plurality of types of genes for each of the plurality of types of cells. Expression information storage means, partial path information storage means for storing information on partial paths obtained by extracting connections of some elements in various compound structures, and the compound information and partial paths referred to by the compound information storage means Based on the information on the partial path referenced from the information storage means, the first compound and the second compound whose pharmacological activity is unknown, And calculating means for calculating partial path presence / absence information indicating whether or not the partial path is included, and classifying the first compounds having similar partial path presence / absence information together as a cluster. To identify a cluster to which the first compound having the most similar chemical structure to the second compound belongs as a similar cluster based on the partial path presence / absence information, and the first cluster belonging to the identified similar cluster Classification processing means for calculating the similarity between each of the compounds and the second compound, and the activity of the pharmacological activity on the cells weighted by the similarity for all the first compounds belonging to the similar cluster An activity value estimating means for calculating a weighted average of values as an estimated activity value of the pharmacological activity of the second compound with respect to the cells, and for each of the genes, Characterized by comprising a gene estimating means for calculating an average value of the integrated value of the estimated activity value and the expression level of pharmacological activity against cells as estimated point.

  Thereby, the gene estimation apparatus by this invention can estimate the gene which can be anticipated with respect to the pharmacological activity of a target compound with respect to the target compound whose pharmacological activity is unknown.

In the gene estimation method according to the present invention , the compound information which is information on the activity value of the pharmacological activity of the plurality of types of first compounds for each of the plurality of types of cells, and the information on the chemical structure of the first compound. Compound information storage means for storing certain compound structure information, expression information storage means for storing expression information including expression levels of a plurality of types of genes for each of the plurality of types of cells, and a part of the structures of various compounds A gene using a gene estimation device comprising: partial path information storage means for storing information on partial paths from which element connections are extracted, calculation means, classification processing means, activity value estimation means, and gene estimation means In the estimation method, the calculation means refers to the compound information referred from the compound information storage means and the partial path information storage means. A calculation step of calculating partial path presence / absence information indicating whether or not the partial path is included in the first compound and the second compound whose pharmacological activity is unknown based on the information regarding the partial path; The classification processing unit collectively classifies the first compounds having similar partial path presence / absence information as clusters, and selects the second compound from the classified clusters based on the partial path presence / absence information. The cluster to which the first compound having the most similar chemical structure belongs is specified as a similar cluster, and the similarity between each of the first compounds belonging to the specified similar cluster and the second compound is calculated. The pharmacology for the cells weighted by the similarity for all the first compounds belonging to the similar cluster, wherein the classification step and the activity value estimation means An activity value estimation step for calculating a weighted average of sex activity values as an estimated activity value of the pharmacological activity of the second compound for the cells; and the gene estimation means for each gene, the pharmacological activity for each cell. And a gene estimation step of calculating an average value of integrated values of the estimated activity value and the expression level as an estimation point .

In addition, the program according to the present invention allows a computer to store compound information that is information on the activity value of the pharmacological activity of a plurality of types of first compounds for each of a plurality of types of cells, and information about the chemical structure of the first compound. Compound information storage means for storing certain compound structure information, expression information storage means for storing expression information including expression levels of a plurality of types of genes for each of the plurality of types of cells, and a part of the structures of various compounds Based on the partial path information storage means for storing information on the partial path from which the element connections are extracted, the compound information referenced from the compound information storage means, and the information on the partial path referenced from the partial path information storage means. A portion indicating whether or not the partial path is included for the first compound and the second compound whose pharmacological activity is unknown And classifying the first compounds having similar partial path presence / absence information together as a cluster, and classifying the second compound based on the partial path presence / absence information from the classified clusters. A cluster to which the first compound having the most similar chemical structure to the compound of 1 belongs is identified as a similar cluster, and the similarity between each of the first compounds belonging to the identified similar cluster and the second compound A weighting average of the activity values of the pharmacological activities for the cells weighted by the similarity for all the first compounds belonging to the similar cluster, and a classification processing means for calculating An activity value estimating means for calculating an estimated activity value of pharmacological activity against cells; and for each gene, the estimated activity value of pharmacological activity against each cell and The average value of the integrated value of the expression level is a program to function as a gene estimating means for calculating as the estimated point.

  According to the gene estimation apparatus, gene estimation method, and program thereof according to the present invention, a gene that can be expected to be related to the pharmacological activity of a target compound can be estimated for a target compound whose pharmacological activity is unknown.

Embodiments of the present invention will be described below.
The gene estimation apparatus according to one embodiment of the present invention provides a gene that can be expected to be related to a pharmacological activity of a compound with an undetermined pharmacological activity (a compound that has no effect on the cell). This is a device for performing the estimation process, and the schematic configuration will be described below. FIG. 1 is a diagram illustrating a schematic configuration of a gene estimation device according to the present embodiment.

  In FIG. 1, reference numeral 1 denotes a gene estimation device. For example, a compound whose pharmacological activity against cancer cells is undetermined is regarded as a target compound (second compound), and a gene that can be expected to be related to the pharmacological activity of the target compound (hereinafter referred to as “the target compound”). , A related gene) is estimated. Reference numeral 2 denotes a network, for example, a communication network such as the Internet. Reference numeral 3 denotes an NCI (National Cancer Institute) database, which is a database published by the NIC used in this embodiment. Specifically, the NCI database 3 is a database that stores at least expression information that is information on gene expression patterns of cancer cells and compound information that is information on pharmacological activity values of compounds against cancer cells.

  The gene estimation apparatus 1 performs processing for estimating the relevant gene of the target compound by acquiring and using the above-described expression information and compound information from the NCI database 3 via the network 2. Although not shown, the gene estimation device 1 includes an input device such as a mouse or a keyboard and a display device such as a CRT (Cathode Ray Tube) or a liquid crystal display.

  Here, the gene expression pattern of cancer cells in the expression information is information relating to the expression level (amount indicating whether or not a gene is functioning) for each of a plurality of types of cancer cells. That is, specific gene combinations (gene patterns) are expressed in specific cancer cells. In the following description, compound information is information indicating the pharmacological activity value for each of a plurality of types of compounds for each of a plurality of types of cancer cells, and the compound structure information is defined as information relating to the structure of the compound. . Specific examples of expression information and compound information will be described later.

  Next, the functional configuration of the gene estimation device 1 will be described. Reference numeral 11 denotes a control unit that controls each processing unit and data flow in the gene estimation apparatus 1. Reference numeral 12 denotes a database, an expression information database 12a for storing the above-described expression information, a compound information database 12b for storing the above-described compound information and compound structure information, and some paths (element connections) of various compounds. A partial path information database 12c for storing information obtained by assigning an ID (identifier) to the partial path from which the information is extracted, and related information that is information on the relationship between the expression pattern of the gene linked to the expression information and the compound information and the compound The related information database 12d is stored.

  Reference numeral 13 denotes an information registration processing unit, which acquires expression information from the NCI database 3 via the transmission / reception processing unit 20 and the network 2 described later and registers them in the expression information database 12a, and acquires compound information from the NCI database 3. Then, a process of registering in the compound information database 12b is performed. The information registration processing unit 13 in the present embodiment acquires T-Matrix (expression information) that is information related to gene expression patterns of cancer cells from the NCI database 3, and registers necessary information in the expression information database 12a. In addition, the information registration processing unit 13 acquires A-Matrix (compound information), which is information relating to the pharmacological activity value for compounds of cancer cells, from the NCI database 3 and registers necessary information in the compound information database 12b.

  Here, a data configuration example of the expression information and the compound information stored in the expression information database 12a and the compound information database 12b described above will be described with reference to FIGS. FIG. 2 is a diagram showing a data configuration example of the expression information database 12a shown in FIG. In FIG. 2, CLID is a numerical value obtained by removing the prefix “IMAGE:” from Clone ID, and is a numerical value unique to each gene. NAME is a gene name associated with Clone ID cDNA (Type). “ME: MALME-3M” and “ME: SK-MEL-28” are names of cancer cells. Moreover, the expression level with respect to each gene is shown under the name of a cancer cell. Note that these CLIDs and NAMEs are defined by T-Matrix (expression information) referenced from the NCI database 3. In addition, the expression level of each gene shown in FIG. 2 is a value obtained by extracting seven typical cancer cells from 60 types in the NCI database 3 and normalizing them with the average and variance value.

FIG. 3 is a diagram showing a data configuration example of compound information stored in the compound information database 12b shown in FIG.
In FIG. 3, “NSC No.” is a numerical value that identifies a compound. Similarly to FIG. 2, “ME-MALME-3M”, “ME-SK-MEL-28”, and the like are names of cancer cells. Moreover, the pharmacological activity value with respect to each compound is shown under the name of a cancer cell. As for this pharmacological activity value, for example, the pharmacological activity value a (ω, δ) of the compound δ with respect to the cell ω is calculated by the following formula 1.

In the above-described formula 1, GI ₅₀ is a growth inhibitory concentration, and here means the concentration of compound δ at which the growth of cancer cells ω is inhibited with a probability of 50%. a _average and a _sd are the mean and variance of the pharmacological activity value of the cancer cell group for the specified compound, respectively. Thus, the pharmacological activity value a (ω, δ) obtained by Formula 1 means the effect of inhibiting the growth of compound δ on cancer cells ω, and is a value normalized for each compound. In the compound information database 12b, as compound structure information, information such as compound names, structure information (chemical symbols and their connection relations), structure diagrams (two-dimensional or three-dimensional molecular structure diagrams), and the like are displayed. No. "is stored in association with it.

  The information registration processing unit 13 also performs information registration processing for the partial path information database 12c. Specifically, the information registration processing unit 13 calculates, as a partial path, a path having an appearance frequency of 1.0 to 0.002 that is a connected path of 9 paths or less in a group of existing compounds. Register in the database 12c. Thus, a partial bus information database 12c having a data structure as shown in FIG. 4 is configured. FIG. 4 is a diagram showing a data configuration example of the partial bus information database 12c shown in FIG. As shown in FIG. 2, an ID (identification number) is assigned to each partial path. In the present embodiment, there are about 4000 types of compounds stored in the compound information database 12b, and information about about 10,000 partial paths is stored in the partial path information database 12c. Note that the path for hydrogen atoms is excluded when calculating the partial path in the information registration processing unit 13.

  Reference numeral 14 denotes an association analysis processing unit, which refers to expression information from the expression information database 12a and compound information from the compound information database 12b, and is information relating to a relationship between a gene expression pattern and a compound having pharmacological activity in the same cancer cell. Information is generated and registered in the related information database 12d. Specifically, the association analysis processing unit 14 associates the gene expression pattern and the pharmacological activity value with the same cancer cell name in the expression information and the cancer cell name in the compound information as a key, and stores them in the associated information database 12d. sign up. At this time, the related analysis processing unit 14 sets a lower limit value ε of the pharmacological activity value, and does not perform the linking process and the registration process for a compound having a pharmacological activity value equal to or lower than the lower limit value ε. In the present embodiment, the lower limit value ε = −10.0.

  Reference numeral 15 denotes an FP calculation processing unit that refers to the compound structure information of all the compounds stored in the compound information database 12b and determines whether or not each partial path referred to from the partial path information database 12c is included. , FP (Finger Print), which is a sequence of numbers indicated by “0”, is calculated. Specifically, the FP calculation processing unit 15 calculates the following formulas 2 and 3 as vector variables f (δ) (= FP), which are structural features, as the FP (partial path presence / absence information) of the compound δ. Calculated by

  Here, θ, Θ, and Π (δ) shown in Expressions 2 and 3 will be described. In this embodiment, a compound is considered as an undirected graph, atoms are regarded as points, bonds are regarded as edges, and whether each element of f (δ) includes a specific path θ (hereinafter referred to as a partial path θ) or not Treat as binary. As shown in FIG. 2, the partial path θ can be expressed in a form such as “C−C = O”, for example. A set Θ of partial paths θ corresponding to each element of the vector variable f (δ) is set as a set of all partial paths stored in the partial path information database 12c. Further, if the path set included in the set Θ among all paths of the compound δ is Π (δ), the following expressions 4 and 5 are established. In this embodiment, the number of partial paths θ (about 10,000) and the number of “1” and “0” included in the vector variable f (δ) that is FP are the same.

  Reference numeral 16 denotes an FP classification processing unit, which is a clustering process of compounds stored in the compound information database 12b, a process of specifying a cluster including a compound most similar to the target compound from the clusters after clustering, and a specified cluster A process of calculating the similarity between the compound belonging to and the target compound is performed.

First, the FP classification processing unit 16 calculates the FP similarity between the compounds δ ₁ and δ ₂ stored in the compound information database 12b and performs clustering to thereby perform the FP of the compounds stored in the compound information database 12b. Classify. Specifically, the FP classification processing unit 16 calculates the similarity t (δ ₁ , δ ₂ ) between the compounds δ ₁ and δ ₂ based on the following “Tanimoto measure”, which is Equation 6.

Here, the range of the similarity t (δ ₁ , δ ₂ ) obtained by the FP classification processing unit 16 according to Equation 6 is 0 ≦ t (δ ₁ , δ ₂ ) ≦ 1, and the two compounds have the same structure on the FP. The value of similarity t is 1 when it has a feature.

Next, the FP classification processing unit 16 performs clustering based on the similarity t obtained by Expression 6. Specifically, the FP classification processing unit 16 clusters a set of compounds (hereinafter referred to as a set Ψ) based on the similarity t (δ ₁ , δ ₂ ). The algorithm used for clustering by the FP classification processing unit 16 of the present embodiment is, for example, “Hierarchical nearest neighbor logic”. In this logic, an element group having a degree of similarity equal to or less than half of the maximum distance T, which is the maximum value of the degree of similarity t between elements of the set Ψ, is treated as one cluster.

Next, assuming that the target compound is δ _target , the FP classification processing unit 16 selects a cluster ψ to which the compound δ structurally similar to the _target compound δ _target belongs among the clusters of compounds possessed by the cell ω, as shown in the following formulas 7 to 7. 9 to determine. In addition, as shown in Formula 7, the group of all compounds that are considered to be active in the cell ω is denoted by Ψ.

Next, the FP classification processing unit 16 obtains the similarity t (δ, δ _target ) with the _target compound δ _target using the above-described formula 6 for all the compounds δ belonging to the cluster ψ determined by itself.

Reference numeral 17 denotes an activity value estimation processing unit that estimates the pharmacological activity value of the target compound for the cell ω based on the pharmacological activity value of a compound similar to the target compound. Specifically, the activity value estimation processing unit 17 calculates the following equation 10 for all the compounds belonging to the cluster ψ determined by the FP classification processing unit 16 to obtain the target compound δ _target for the cell ω. The estimated activity value h (ω, δ _target ) is calculated. As shown in Expression 10, the activity value estimation processing unit 17 calculates the similarity t (δ, δ _target ) obtained by the above-described FP classification processing unit 16 and Expression 1 for all the compounds belonging to the determined cluster ψ. The estimated activity value h (ω, δ _target ) is calculated by _obtaining the integrated value of the pharmacological activity value a obtained in (1) and dividing it by the sum of the similarities t (δ, δ _target ).

  In Equation 10 described above, λ is a parameter, and when the value of λ is increased, the similarity between compounds can be evaluated more strictly. In this embodiment, λ = 4.0.

Reference numeral 18 denotes a point calculation processing unit that calculates an estimated point indicating a high _degree of relevance to the _target compound δ _target for each gene based on the estimated activity value h calculated by the activity value estimation processing unit 17. Specifically, the point calculation processing unit 18 uses the following equation 11 to calculate the estimated activity value h (ω, δ _target ) calculated by the activity value estimation processing unit 17 and the expression level (T -The average value of the integrated values of c) after normalization with Matrix) is calculated as an estimated point p (δ _target , γ) for the target compound δ _target of the gene γ.

  Reference numeral 19 denotes a gene estimation processing unit that outputs the absolute values of the estimated points p of each gene calculated by the point calculation processing unit 18 and rearranges them in descending order as a gene estimation result. That is, it is estimated that a gene having a higher estimated point p is a gene (related gene) that can be expected to be related to the pharmacological activity of the target compound. Since the estimated point p is calculated from the product of the normalized estimated activity value h and the expression level c, the absolute value of the expression level c is the target compound even if the estimated activity value h is small (that is, negatively large). If it is larger, it will greatly affect the final estimated point p.

  A transmission / reception processing unit 20 communicates with the NCI database 3 via the network 2. In addition, in the gene estimation apparatus 1 of this embodiment, in order to use the data stored in the external NCI database 3, it has the function to connect to the network 2, but it is not limited to this. For example, expression information and compound information may be registered and stored in advance in the internal database 12 from the input means. In this case, the gene estimation device 1 does not need a function for connecting to the network 2.

  Next, the process of estimating the relevant gene of the target compound in the gene estimation apparatus 1 shown in FIG. 1 will be described with a specific example. FIG. 5 is a flowchart showing a process of estimating the relevant gene of the target compound in the gene estimation apparatus 1 shown in FIG.

  As shown in FIG. 5, in step S1, the information registration processing unit 13 acquires expression information and compound information from the NCI database 3 via the network 2, and registers them in the expression information database 12a and the compound information database 12b, respectively. Specifically, the information registration processing unit 13 is a data table including the pharmacological activity values of 4463 compounds (data is present 4444) against 60 types of cancer cells as expression information from the NCI database 3. -Obtain Matrix and register it in the expression information database 12a.

  Further, the information registration processing unit 13 acquires T-Matrix, which is a data table including the expression levels of 9704 genes (9073 types for which data exists) for 60 types of cancer cells, and obtains the compound information database 12b. Register with. However, since the notation of the cell name in the same cancer cell is different between the T-Matrix and the A-Matrix, the information registration processing unit 13 performs conversion to unify either cell name (for example, ME: MALME- 3M → MEL-MALME-3M).

  Further, the information registration processing unit 13 acquires 4463 types of compound structure information from the NCI database 3 and registers them in the compound information database 12b. Further, the information registration processing unit 13 registers information related to the partial path in the partial path information database 12c.

  Next, in step S2, the association analysis processing unit 14 associates the gene expression pattern and the pharmacological activity value with the same cancer cell name in the expression information and the cancer cell name in the compound information as a key, and the associated information database. Register to 12d. At this time, the related analysis processing unit 14 sets the lower limit value ε = −10.0 of the pharmacological activity value, and does not perform the linking process and the registration process for compounds having a pharmacological activity value equal to or lower than the lower limit value ε.

Next, in step S3, the FP calculation processing unit 15 refers to the compound structure information of all the compounds stored in the compound information database 12b and includes each partial path referred to from the partial path information database 12c. Is calculated. Next, in step S3, the FP classification processing unit 16 calculates the FP similarity (between compounds) calculated by the FP calculation processing unit 15, performs clustering, and selects a cluster including a compound similar to the target compound. Identify. The FP classification processing unit 16 also calculates the similarity t (δ, δ _target ) between the compound δ belonging to the specified cluster (compound group) and the target compound δ _target .

Next, in step S4, the activity value estimation processing unit 17 sets the pharmacological activity value of the target compound with respect to the cell ω to the similarity t (δ, δ _target ) with the pharmacological activity value of the compound belonging to the cluster specified in step 3. Based on this, an estimated activity value h (ω, δ _target ) of the target compound δ _target for the cell ω is calculated. The similarity t (δ, δ _target ) is the similarity calculated by the FP classification processing unit 16 in step S3.

Next, in step S5, the point calculation processing unit 18 calculates an estimated point p indicating the _degree of relevance to the _target compound δ _target for each gene based on the estimated activity value h calculated by the activity value estimation processing unit 17. calculate. Next, in step S6, the gene estimation processing unit 19 outputs, as gene estimation results (ranking), gene groups obtained by taking the absolute values of the estimated points p of the genes calculated by the point calculation processing unit 18 and rearranging them in descending order. .

  As described above, according to the gene estimation device 1 of the present embodiment, for example, a compound having an undetermined pharmacological activity against cancer cells is used as a target compound, and a gene that can be expected to be related to the pharmacological activity of the target compound is estimated. can do. Moreover, the information regarding a target protein is not required unlike the past.

[Demonstration experiment]
The results of an experimental simulation of the gene estimation method described above using actual data will be described below. First, the target data used will be described. In this experimental simulation, the target data are those in which the relationship between the compound and the gene is shown in advance as document information so that the result can be easily confirmed.

  FIG. 6 is a diagram showing a list of target compounds that are the target data of the experimental simulation and expected related genes. That is, if the related gene shown as related to each compound in FIG. 6 can be estimated using the compound shown in FIG. 6 as a target compound, the effectiveness of the gene estimation method in this embodiment can be shown. it can.

  In FIG. 6, the symbols (+ and −) at the right end indicate the direction of the correlation between the pharmacological activity of the compound and the expression level of the gene. If it is (+), it means that the higher the gene expression level, the higher the pharmacological activity for the compound, and the lower the gene expression level, the lower the pharmacological activity of the compound. Moreover, if it is (-), it will mean that the pharmacological activity with respect to a compound is so high that the expression level of a gene is small, and pharmacological activity is so low that the expression level of a gene is high. In addition, a blank as in () means that there is no relation between the gene expression level and the pharmacological activity of the compound.

  Although it is generally considered that there is no association between most genes and compounds, the expression level and activity of the gene TYMS that produces the protein shown on the left in the compound 5-fluorouracil shown in FIG. 6 while the related gene thymidylate synthase is the target protein Since it is pointed out that it is irrelevant, it is adopted as a target gene.

  The details of the target data used in this experiment will be described below. Among the pharmacological activity values of 4463 compounds (4444 types for which data exists) against 60 types of cancer cells published by NCI as A-Matrix, among the target compounds shown in FIG. Data excluding two compounds is used. Two of the target compounds shown in FIG. 6 are “NSC NO. 656238” and “Doxorubicin (NSC NO. 123127)”. In this case, the data for the other two compounds “5-fluorouracil” and “CPT-11” of the target compounds shown in FIG. 6 are not excluded from A-Matrix, but preferably excluded. Is good.

  In this experiment, compound structure information of 4463 kinds of compounds published by NIC is used. Hereinafter, the entire compound structure information is referred to as a compound population. Further, in the process of associating the expression information with the compound information, as described above, by setting the lower limit value ε = −10.0, all compounds having an activity value in each cell were employed as the analysis target. This is because, in gene estimation from a compound, even a compound with low activity affects the result.

  Based on the above target data, the result of estimation by the gene estimation method described above will be described below. FIG. 7 is a diagram showing the results of an experimental simulation. In FIG. 7, the rank A indicates the rank (total of 9365 species) in the entire gene set (population). Rank B indicates the rank (4845 species in total) in the gene set excluding EST (Expressed Sequence TAG). Here, the EST is a database of gene fragments whose functions are not known but whose structures are elucidated. In addition, the rank C means the rank of gene sets (4134 types in total) having expression information for 10 or more types of cancer cells, except for the record described as EST in the description of the data set.

  In FIG. 7, the values in the “estimated” column are estimated point values, and the percent values in parentheses in the ranks A to C are ratios obtained by dividing the ranking by the number of genes. As is clear from the estimated point value of FIG. 7, the absolute value of the estimated point relative to the gene is related to the gene group related to the compound of “NSC NO. 656238” and the gene related to DPYD of “5-fluorouracil”. The gene is high, and the gene appears at the top of the ranking. These gene groups related to DPYD of “NSC NO. 656238” and “5-fluorouracil” are data that are considered to be related in the papers published by NCI that publishes the data set. It can be said that the estimation points and rankings estimated by the gene estimation device 1 in the embodiment are effective.

  As shown in FIG. 7, the overall ratio is rank A≈rank B> rank C. That is, it means that the gene estimation apparatus 1 in the present embodiment selectively selects a gene whose expression information exists only in 10 or less types of cancer cells. Therefore, in rank C, genes were estimated for a gene set excluding genes whose expression information exists only in 10 or less types of cancer cells. As can be seen from the ranking of the ranking C, the data (NSC NO. The data were estimated at a fairly high ranking (the ratio was less than 3%). Thus, the gene estimation apparatus 1 of the present embodiment improves the accuracy of estimation by estimating genes for a gene set excluding genes whose expression information exists only in 10 or less types of cancer cells. be able to.

  On the other hand, as shown in FIG. 7, the combination of a compound and a gene that is considered to be related in a paper unrelated to the NCI data set can be estimated to a certain extent (less than 15% in proportion). (A ratio of 15% or more).

  Further, in the above experiment, when the gene estimation device 1 searches for a paper that is not shown in FIG. 6 and describes a relation between a gene ranked higher and a compound, the compound “5-fluorouracil” is obtained. Articles and documents describing that there is a relationship between genes ranked 35 (0.8%) in rank C and the compound "5-fluorouracil" and genes ranked 42 (1.0%) in rank C. One by one is found. Similarly, three papers and literatures describing that there is a relation between the compound “CTP-11” and the gene ranked 75 (1.8%) in rank C have been found. Furthermore, four papers and literatures have been found that state that there is a relationship between the compound “CTP-11” and the gene of ranking 82 (2.0%) in rank C.

  As described above, the discovery of papers and documents that support the relationship between the related gene and the compound estimated by the gene estimation device 1 of the present embodiment as the highest ranking is also the gene in the gene estimation device 1 of the present embodiment. It can be said that this estimation method is effective.

  Further, in the above-described embodiment, each processing unit of the gene estimation device 1 shown in FIG. 1 includes a memory and a CPU (central processing unit) as hardware, and a program for realizing the functions of each processing unit Is loaded into the memory and executed by the CPU to realize its function. In addition, the present invention is not limited to this, and part or all of the processing of each processing unit may be realized by dedicated hardware.

  The memory includes a nonvolatile memory such as a hard disk device, a magneto-optical disk device, and a flash memory, a recording medium such as a CD-ROM that can only be read, and a volatile memory such as a RAM (Random Access Memory). Or a computer-readable / writable recording medium based on a combination thereof.

  Further, each processing unit of the gene estimation apparatus 1 shown in FIG. 1 is realized by the computer executing the program as described above, but means for supplying the program to the computer, for example, such a program is provided. A recorded computer-readable recording medium or a transmission medium for transmitting such a program can also be applied as an embodiment of the present invention. A program product such as a computer-readable recording medium in which the above program is recorded can also be applied as an embodiment of the present invention. The above program, recording medium, transmission medium, and program product are included in the scope of the present invention.

  The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Further, the “computer-readable recording medium” refers to a volatile memory (RAM) in a computer system serving as a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those holding a program for a certain period of time are also included.

  The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.

  The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.

It is a figure which shows schematic structure of the gene estimation apparatus in this embodiment. It is a figure which shows the data structural example of the expression information database 12a shown in FIG. It is a figure which shows the data structural example of the compound information stored in the compound information database 12b shown in FIG. It is a figure which shows the data structural example of the partial bus information database 12c shown in FIG. It is a flowchart which shows the process which estimates the related gene of the target compound in the gene estimation apparatus 1 shown in FIG. It is a figure which shows the list | wrist of the target gene used as the object data of experiment simulation, and the related gene expected. It is a figure which shows the result of experiment simulation.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Gene estimation apparatus 2 Network 3 NCI database 11 Control part 12 Database 12a Expression information database 12b Compound information database 12c Partial path information database 12d Related information database 13 Information registration process part 14 Related analysis process part 15 FP calculation process part 16 FP classification process Unit 17 activity value estimation processing unit 18 point calculation processing unit 19 gene estimation processing unit 20 transmission / reception processing unit

Claims (6)

Compound information storage means for storing compound information, which is information regarding the activity value of the pharmacological activity of the plurality of types of first compounds for each of a plurality of types of cells, and compound structure information, which is information about the chemical structure of the first compound. When,
Expression information storage means for storing expression information including expression levels of each of a plurality of types of genes for each of the plurality of types of cells;
Partial path information storage means for storing information on partial paths extracted from the connection of some elements in the structures of various compounds;
Based on the compound information referenced from the compound information storage means and the information on the partial path referenced from the partial path information storage means, the first compound and the second compound whose pharmacological activity is unknown Calculating means for calculating partial path presence / absence information indicating whether or not the partial path is included;
The first compounds having similar partial path presence / absence information are collectively classified as clusters, and the chemical structure of the second compound is most similar to the second compound based on the partial path presence / absence information from the classified clusters. Classification processing means for identifying a cluster to which the first compound belongs as a similar cluster, and calculating a similarity between each of the first compounds belonging to the identified similar cluster and the second compound;
For all the first compounds belonging to the similar cluster, the weighted average of the activity values of the pharmacological activity for the cells weighted by the similarity is calculated as the estimated activity value of the pharmacological activity of the second compound for the cells. Activity value estimation means to calculate as
A gene estimation device comprising: a gene estimation unit that calculates an average value of an integrated value of the estimated activity value of the pharmacological activity for each cell and the expression level for each gene as an estimated point . When the first compound is δ ₁ , the second compound is δ _2, and the partial path presence / absence information for the compound δ is a vector variable f (δ), The gene estimation apparatus according to claim 1, wherein similarity is calculated .

The activity value estimating means sets the cell to ω, sets the pharmacological activity value of the _first compound δ ₁ for the cell ω to a (ω, δ ₁ ), and sets the first compound δ ₁ and the second compound 3. The estimated activity value is calculated by Equation 2 when the similarity with the compound δ ₂ is t (δ ₁ , δ ₂ ), the cluster is ψ, and λ is a parameter. The gene estimation apparatus according to 1.

The gene estimation means sets the gene as γ, and the second compound δ for the cell ω. ₂ The estimated activity value of h (ω, δ ₂ The gene estimation apparatus according to claim 3, wherein the estimation point is calculated by Equation 3 when the expression level of the gene γ is cγ and the set of cells is Ω.

Compound information storage means for storing compound information, which is information regarding the activity value of the pharmacological activity of the plurality of types of first compounds for each of a plurality of types of cells, and compound structure information, which is information about the chemical structure of the first compound. And expression information storage means for storing expression information including expression levels of each of the plurality of types of genes for each of the plurality of types of cells, and information on partial paths from which connections of some elements in the structures of various compounds are extracted A gene estimation method using a gene estimation apparatus comprising a partial path information storage means, a calculation means, a classification processing means, an activity value estimation means, and a gene estimation means,
A second pharmacological activity that is unknown to the first compound based on the compound information referenced from the compound information storage means and the information on the partial path referenced from the partial path information storage means; A calculation step of calculating partial path presence / absence information indicating whether the partial path is included with respect to the compound of
The classification processing unit collectively classifies the first compounds having similar partial path presence / absence information as clusters, and chemically classifies the second compounds from the classified clusters based on the partial path presence / absence information. Classification that specifies a cluster to which the first compound having the most similar structure belongs as a similar cluster, and calculates a similarity between each of the first compounds belonging to the specified similar cluster and the second compound Processing steps;
The activity value estimation means calculates the weighted average of the activity values of the pharmacological activity for the cells weighted by the similarity for all the first compounds belonging to the similar cluster, and the cells of the second compound An activity value estimation step for calculating an estimated activity value of pharmacological activity against
The gene estimation means includes a gene estimation step for calculating an average value of an integrated value of the estimated activity value of the pharmacological activity for each cell and the expression level for each gene as an estimated point. Gene estimation method. Computer
Compound information storage means for storing compound information, which is information regarding the activity value of the pharmacological activity of the plurality of types of first compounds for each of a plurality of types of cells, and compound structure information, which is information about the chemical structure of the first compound. When,
Expression information storage means for storing expression information including expression levels of each of a plurality of types of genes for each of the plurality of types of cells;
Partial path information storage means for storing information on partial paths extracted from the connection of some elements in the structures of various compounds;
Based on the compound information referenced from the compound information storage means and the information on the partial path referenced from the partial path information storage means, the first compound and the second compound whose pharmacological activity is unknown Calculating means for calculating partial path presence / absence information indicating whether or not the partial path is included;
The first compounds having similar partial path presence / absence information are collectively classified as clusters, and the chemical structure of the second compound is most similar to the second compound based on the partial path presence / absence information from the classified clusters. Classification processing means for identifying a cluster to which the first compound belongs as a similar cluster, and calculating a similarity between each of the first compounds belonging to the identified similar cluster and the second compound;
For all the first compounds belonging to the similar cluster, the weighted average of the activity values of the pharmacological activity for the cells weighted by the similarity is calculated as the estimated activity value of the pharmacological activity of the second compound for the cells. Activity value estimation means to calculate as
A program for causing each gene to function as gene estimation means for calculating an average value of integrated values of the estimated activity value and the expression level of pharmacological activity for each cell as an estimated point.

Images