JP6929015B2 - Biomarker search device, biomarker search method and program - Google Patents

Description

Embodiments of the present invention relate to biomarker search devices, biomarker search methods and programs.

The human genome is composed of two sequences of about 3 billion base pairs, and is further divided into 22 types of autosomal chromosomes and X and Y sex chromosomes. Although each base pair is almost the same within the same ethnic group, there are multiple base pairs that differ from person to person. The different base pairs are called SNPs (Single-Nucleotide Polymorphisms).

It is known that some SNPs affect the phenotypic expression of diseases. However, the trait is, for example, the presence or absence of a disease. In addition, although it is difficult to recognize the relationship between individual SNPs and trait expression alone, it has been suggested that the trait may be expressed by a combination of a plurality of SNPs. Such a combination of SNPs that is found to be related to phenotypic expression as a disease is called a biomarker candidate.

Among the biomarker candidates, the true causal relationship is derived through medical verification, biological causal analysis, verification of environmental factors such as lifestyle, and the presence or absence of influence by other factors such as age. The biomarker is used as a biomarker, and is recognized as useful information applicable to actual services such as treatment using the knowledge as information related to public trait expression.

The technique for detecting such a combination of SNPs has been brought about by recent genome analysis techniques. Due to the huge number of SNP combinations, it is not easy to investigate the effect of disease on phenotypic expression for all combinations. Therefore, the actual situation is that the search must be performed by limiting the combination of SNPs. There is a method of performing a full search for combinations of up to two. Alternatively, there is also a method of creating a ranking for each SNP based on a combination of at most two types of SNPs and searching for the combination. However, when the full search is used, it is difficult to perform three or more searches due to the problem of calculation time. In addition, as medical findings, some SNPs are recognized to be highly relevant to a specific disease, and it is desirable to detect biomarker candidates in consideration of such SNP information.

However, an efficient method for searching for biomarker candidates from a huge number of combinations has not been proposed so far, taking into consideration the SNP information obtained from medical knowledge.

JP-A-2010-224815 Japanese Unexamined Patent Publication No. 2013-175135

One embodiment of the present invention provides a biomarker search device, a biomarker search method, and a program capable of efficiently searching for biomarker candidates in consideration of medical knowledge.

According to the present embodiment, a specific SNP designation unit that specifies a specific SNP presumed to be related to a specific disease from among a plurality of SNPs (Single-Nucleotide Polymorphisms) in the base sequence. When,
A candidate search unit that searches for biomarker candidates containing two or more SNPs that are presumed to be related to the specific disease based on the trait information of the specific SNP and the sample.
A biomarker candidate search device including the candidate output unit for outputting the biomarker candidate is provided.

The block diagram which shows the schematic structure of the biomarker search apparatus by one Embodiment. The figure which shows an example of the disease presence vector and the junction type matrix. The figure which shows an example of the SNP combination matrix. A flowchart showing a method of identifying the presence or absence of a disease in a target sample. A more detailed block diagram of the biomarker search device according to one embodiment. The figure which shows an example of the SNP trait DB. The flowchart which shows an example of the processing procedure of the sample information input part. The figure which shows an example of the sample information registration DB. The figure which shows an example of the SNP information registration DB. The figure which shows an example of the related SNP registration DB. The figure which shows an example of the GUI screen which serves as the sample information input part and the search condition input part. The flowchart which shows the processing procedure of the biomarker search apparatus by this embodiment. The figure which shows an example of the score of each junction element of the SNP combination matrix in step S18. 2 Three of the combination of ^V street (hereinafter, combination c1 to c3) shows each identification error of. The flowchart which shows the update method of the SNP combination matrix. The figure which shows an example of the output form of step S22. The figure which plotted the odds value or -log (P value) which shows the discrimination accuracy of each SNP combination.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a schematic configuration of a biomarker search device 1 according to an embodiment. The biomarker search device 1 of FIG. 1 includes a specific SNP designation unit 2, a candidate search unit 3, and a candidate output unit 4.

The specific SNP designation unit 2 acquires a specific SNP designated as being related to a specific disease from a plurality of SNPs (Single-Nucleotide Polymorphisms) in the genome (base sequence).

The candidate search unit 3 searches for biomarker candidates containing two or more SNPs that are presumed to be related to a specific disease, based on the trait information of the specific SNP and the sample. The trait information of the sample is registered in , for example, the search information registration described later. Therefore, in more detail, the candidate search unit 3 includes a combination of SNPs including one or more SNPs designated by the specific SNP designation unit 2 based on the specific SNPs designated by the specific SNP designation unit 2 and the search information registration DB. Search for K (K is an integer of 2 or more).
FIG. 2 is a diagram showing an example of registration information in the search information registration DB. As shown in the figure, a disease presence / absence vector is registered in the search information registration DB. The disease presence / absence vector of FIG. 2 records the disease presence / absence information of each sample. A value of 1 indicates that the corresponding specimen has disease, and a value of 0 indicates no disease. In addition, each SNP is composed of a combination of two bases, and A (adenine) and T (thymine) or G (guanine) and C (cytosine) are paired. Of the two types of bases that appear in each SNP, the one with the largest number is called the major allele, and the one with the smaller number is called the minor allyl. Therefore, there are three types of base combinations that make up each SNP: major allele (major homozygotes, XX), major alleles and minor alleles (heterozygotes, XY), and both minor alleles (minor homozygotes, YY). It can be classified as a conjugate. As shown in FIG. 2, a junction type matrix representing the junction type of each sample is registered in the search information registration DB.

FIG. 3 is an example of searching for two SNP combinations. In FIG. 3, a matrix in which the number of rows is the number of combinations (K) and the number of columns is the number of conjugates of SNP is called an SNP combination matrix, and the output of the SNP combination is shown. An element in the SNP combination matrix that takes a value of 1 corresponds to the one used in the corresponding SNP combination. In the case of FIG. 3, as shown by T1, the number of combinations is 2, the first combination is (SNP-00001 is XX, SNP-00002 is YY), and the second combination is (SNP-00003). Is XY, and SNP-00004 is XY). The candidate search unit 3 typically searches for a combination including two or more SNPs including a specified specific SNP. However, depending on the data set and conditions, it is assumed that a specific SNP is not included, and in that case an error is returned.

A method for identifying the presence or absence of a disease in each sample using a combination of a plurality of SNPs is represented by, for example, the flowchart of FIG. When the sample has all the SNP conjugates used in any of the K SNP combinations searched, it is identified as having a disease. In the case of FIG. 3, a sample satisfying the condition of (SNP-00001 is XX and SNP-00002 is YY) or (SNP-00003 is XY and SNP-00004 is XY) is identified as having a disease. A description of the flowchart of FIG. 4 will be described later.

As shown in FIG. 1, the candidate search unit 3 has an evaluation value calculation unit 5, an identification error calculation unit 6, and a minimum identification error selection unit 7. The evaluation value calculation unit 5 performs a process of calculating an evaluation value indicating a high possibility that each of the plurality of SNPs can be a biomarker candidate for each SNP V (V is an integer of 2 or more) times. More specifically, the evaluation value calculation unit 5 performs a process of calculating an evaluation value indicating a high possibility that each element of the SNP combination matrix of FIG. 3 is selected as a combination of SNPs at each SNP. Do this twice or more). In the present specification, the evaluation value is also referred to as a score. Hereinafter, each element of the SNP combination matrix is also referred to as a zygote element. As shown in FIG. 3, for each SNP, for example, a total of three joining elements of XX, XY, and YY are provided.

The identification error calculation unit 6 determines the high degree of relevance to a specific disease for any combination of SNPs in a collection of V SNPs corresponding to the maximum value of each evaluation value by the evaluation value calculation unit 5. Calculate the indicated identification error. More specifically, in the identification error calculation unit 6, each joint element in the collection of V joint elements corresponding to the maximum value of each evaluation value by the evaluation value calculation unit 5 is 0 in the SNP combination matrix. About 2 ^V as an all combinations taking a value of either 1, it calculates the identification error representing the degree of SNP combined matrix could be correctly identified the presence or absence of disease in each sample. In this case, when the joint element takes a value of 1, it indicates that the corresponding joint element is adopted as the SNP combination, and when it takes a value of 0, it indicates that it is not adopted as the SNP combination. The minimum identification error selection unit 7 selects the combination of SNPs having the smallest identification error in all cases where each junction element takes a value of 0 or 1 in the SNP combination matrix. The evaluation value calculation unit 5 and the identification error calculation unit 6 update the SNP combination matrix using the join element selected by the minimum identification error selection unit 7. Then, the evaluation value calculation unit 5 repeats the process U (U is an integer of 2 or more) times to update the SNP combination matrix. The SNP combination matrix, which is the output result calculated by the candidate search unit 3, typically corresponds to a biomarker candidate containing two or more SNPs.

The candidate output unit 4 outputs the biomarker candidate searched by the candidate search unit 3. More specifically, the candidate output unit 4 outputs an SNP combination matrix, which is a combination of SNPs selected by the minimum identification error selection unit 7 after U times of processing, as a biomarker candidate.

The candidate search unit 3 may have a matrix initialization unit 8 that sets an initial value of the SNP combination matrix.

Further, the evaluation value calculation unit 5 may have a maximum joint element selection unit 9 and a plurality of joint element selection units 10. The maximum joint element selection unit 9 acquires each joint element in the SNP combination matrix, calculates an evaluation value, and selects the joint element having the maximum evaluation value. The plurality of joint element selection units 10 repeat the process of the maximum joint element selection unit 9 V times, and select a total of V different joint elements each time.

In this case, the identification error calculation unit 6 calculates the identification error for all combinations of whether or not each of the V joint elements selected by the plurality of joint element selection units 10 is selected as the combination of SNPs. do.

FIG. 5 is a more detailed block diagram of the biomarker search device 1 according to the embodiment. In addition to having each part shown in FIG. 1, the biomarker search device 1 of FIG. 5 includes an SNP trait DB 11, a sample information input unit 12, a sample information registration DB 13, an SNP information registration DB 14, and a specific SNP registration DB 15. , Related SNP registration DB 16, search range SNP selection unit (search range acquisition unit) 17, selected SNP registration DB 18, search information collation unit 19, search information registration DB 20, search condition input unit 21, and biomarker candidates. It has a registration DB 22.

The SNP trait DB 11 is a database in which a plurality of SNPs (also referred to as SNP series data) included in each sample and information on whether or not each sample has a specific trait are registered in association with each other. In this specification, "database" is abbreviated as DB.

FIG. 6 is a diagram showing an example of the SNP trait DB11. As shown in FIG. 6, in the SNP trait DB 11, the identification number of each sample, the information of SNP contained in each sample, and the information on whether or not each sample has a specific disease are registered.

Here, SNP is a base pair having different characteristics depending on an individual in the gene sequence. For example, in FIG. 6, at the gene sequence position SNP-00002, the combination of genotypes that can be taken by a plurality of samples P-001 to P-010 may be CC, CT, and TT, and differs depending on the sample. A base pair whose genotype combination differs depending on the sample is called SNP.

Further, in the SNP trait DB 11 of FIG. 6, the presence or absence of traits for two types of diseases Trait-001 and Trait-002 is represented by 0 and 1 for each sample. 0 is no trait and 1 is trait. The type and number of diseases registered in the SNP trait DB 11 are not particularly limited.

The sample information input unit 12 inputs diagnostic results such as interviews with each sample, and attribute information and trait information regarding the sample such as past medical history and morbidity history of relatives. The input attribute information is registered in the sample information DB.

FIG. 7 is a flowchart showing an example of the processing procedure of the sample information input unit 12. The sample information input unit 12 inputs the sample number (step S1), the age of the sample (step S2), the nationality (step S3), the medical history (step S4), and the body type (step S5) in this order. The input order of steps S1 to S5 is not particularly limited. The input of the sample information input unit 12 is performed using an information input device such as a keyboard.

Next, the sample information input unit 12 registers each information input in steps S1 to S5 in the sample information registration DB 13 (step S6). The process of step S6 may be performed for each step of steps S1 to S5.

Each of the information input in steps S1 to S5 is also referred to as attribute information or sample information. By registering in the sample information registration DB 13 by the process of FIG. 7, by specifying an arbitrary sample number from the sample numbers P-001 to P-010, the attribute information corresponding to the sample number is collectively collected. It can be obtained from the sample information registration DB 13.

FIG. 8 is a diagram showing an example of the sample information registration DB 13. In the example of FIG. 8, the systolic blood pressure, the diastolic blood pressure, the information on the presence or absence of a human medical history corresponding to the sample, and the information on the presence or absence of a medical history of the human relative are registered in the sample information registration DB 13 for each sample. ing. FIG. 8 is an example, and the sample information registered in the sample information registration DB 13 is not particularly limited.

The SNP information registration DB 14 registers information on SNPs constituting genotypes and information on the presence or absence of a plurality of diseases for each sample.

FIG. 9 is a diagram showing an example of the SNP information registration DB 14. In FIG. 9, each SNP is divided into three homozygotes, minor homozygotes and heterozygotes. Take a value of 1 for the applicable joint.

As shown in FIG. 9, since there are three joints for one SNP, the number of columns is three times the number of SNPs. Of these three conjugates, only one will be 1 and the other 2 will be 0.

In FIG. 9, the case where there is a trait for a certain disease is set to 1 and the case where there is no trait for a certain disease is set to 0 for each sample. Multiple specimens may be traits for the same disease.

The related SNP registration DB 16 is a database in which an ID in which a certain commonality is grouped (hereinafter referred to as a group ID) and an ID of an SNP related to the commonality are registered. A certain commonality refers to, for example, an SNP group found to be related to a general disease, an SNP group that controls immunity with a chromosome, and the like. When narrowing down the search target by samples with correlated SNPs, by specifying the group ID, the ID group of the related SNPs is selected, and the search for a specific SNP group is performed only with the samples having that genotype. be able to.

FIG. 10 is a diagram showing an example of the related SNP registration DB 16. In the related SNP registration DB 16 of FIG. 10, the group ID and the ID of the related SNP are registered in association with each other. For example, in the group Chr-001, SNP-00001, SNP-00002, ..., SNP-01000 are registered as related SNP information. Chr-001 to Chr-022 are chromosome numbers, and HLA-001 to HLA-003 are HLA region numbers.

The search range SNP selection unit 17 selects a group ID registered in the related SNP registration DB 16, specifies a subset of the corresponding SNP, and registers the SNP number corresponding to this subset in the selection SNP registration DB 18. The data structure of the selected SNP registration DB 18 is the same as that of the related SNP registration DB 16, and a part of the registration data of the related SNP registration DB 16 is registered in the selected SNP registration DB 18.

The specific SNP designation unit 2 also shown in FIG. 1 designates a specific SNP from those registered in the SNP information registration DB 14 of FIG. 5 and registers it in the specific SNP registration DB 15.

The search information collation unit 19 collates the specific SNP registered in the specific SNP registration DB 15, the registration information in the SNP trait DB 11, the registration information in the sample information registration DB 13, and the registration information in the selected SNP registration DB 18. Then, the matching information is registered in the search information registration DB 20.

The specific SNP designation unit 2 and the search condition input unit 21 shown in FIG. 5 can be input on a GUI (Graphical User Interface) screen displayed on a display device (not shown). FIG. 11 is a diagram showing an example of a GUI screen that also serves as a sample information input unit 12 and a search condition input unit 21. The GUI screen of FIG. 11 has windows w1 to w4. Of these, windows w1 and w2 correspond to the specific SNP designation unit 2, and window w4 corresponds to the search condition input unit 21.

The window w1 (first window) specifies a specific SNP designated by the specific SNP designation unit 2. Window w2 (second window) lists all the specified specific SNPs. Window w3 (third window) specifies a specific disease type. The window w3 is provided with, for example, a plurality of disease names and radio buttons for selecting each disease, and the user selects a disease corresponding to the radio button by checking an arbitrary radio button. be able to. Further, in the window w4, various parameters required by the candidate search unit 3 are input. As a specific example, a parameter ε for correcting the evaluation value (score), a parameter α for correcting the identification error, a combination number K of SNPs, and a parameter U indicating the number of repetitions in the candidate search unit 3 V and so on.

When the selection and setting of the windows w1 to w4 are completed, the user presses the submit button b1 provided at the lower right of the screen. As a result, the processing of the specific SNP designation unit 2 and the search condition input unit 21 is completed.

The search condition input unit 21 may include a first correction constant input unit 21a and a second correction constant input unit 21b. The first correction constant input unit 21a inputs the first correction constant (ε) for correcting the evaluation value of a specific SNP. The evaluation value calculation unit 5 calculates the evaluation value of a specific SNP based on the first correction constant. As a result, the evaluation value of a specific SNP can be made higher than the evaluation value of other SNPs.

The second correction constant input unit 21b inputs the second correction constant (α) for correcting the identification error corresponding to the combination of SNPs including a specific SNP. The identification error calculation unit 6 calculates the identification error corresponding to the combination of SNPs including a specific SNP based on the second correction constant. Thereby, the identification error corresponding to the combination of SNPs including a specific SNP can be set small.

The search condition input unit 21 may include a K input unit 21c for inputting the number K of combinations of SNPs that are biomarker candidates.

Further, the search condition input unit 21 may include a U input unit 21d for inputting the above-mentioned U and a V input unit 21e for inputting the above-mentioned V. As described above, V is the number of joint elements selected by the evaluation value calculation unit 5 in the evaluation value calculation unit 5. Further, U is the number of times that the minimum identification error selection unit 7 performs a process of selecting the combination of SNPs having the minimum identification error.

FIG. 12 is a flowchart showing a processing procedure of the biomarker search device 1 according to the present embodiment. First, the SNP series data within the search range and the trait information of the sample are acquired from the search information registration DB 20 (step S11). In the following, the SNP sequence data within the search range may be referred to as a junction type matrix, the trait information of the sample may be referred to as a trait vector, and the disease presence / absence information of the sample may be referred to as a disease presence / absence vector.

As shown in FIG. 9, the row direction of the junction type matrix has as many rows as the number of samples, and the column direction is three types of conjugates (major homozygotes, heterozygotes, minor homozygotes) for one SNP. The total number in the column direction is 3 × the number of SNPs. The row direction of the trait vector has rows for the number of samples, and the column direction has columns for the number of diseases.

The junction type matrix has three elements for one SNP. For example, major homozygotes are represented by {1,0,0}, heterozygotes are represented by {0,1,0}, and minor homozygotes are represented by {0,0,1}. The junction type matrix is arranged by the number of SNPs in the column direction and by the number of samples in the row direction.

The junction type matrix and the disease presence / absence vector are registered in the search information registration DB 20, and in step S11, SNP sequence data within the search range and sample trait information are acquired from the search information registration DB 20.

FIG. 2 is a diagram showing an example of a disease presence / absence vector and a junction type matrix. The disease presence / absence vector specifies one column out of a plurality of columns representing the trait in the SNP trait DB 11 of FIG. Alternatively, the values obtained by performing the product or sum operation from the trait information for a plurality of columns may be replaced with 0 and 1.

As described above, in step S11 of FIG. 12, the junction type matrix and the trait vector within the search range are acquired by using the table data of FIGS. 9 and 2.

Next, the specific SNP designated by the user in the specific SNP designation unit 2 and various search conditions input by the user in the search condition input unit 21 are acquired (step S12). The search conditions acquired here include, for example, the number of SNP combinations (the number of rows in the SNP combination matrix) K input by the search condition input unit 21, and the first correction constant for correcting the evaluation value of a specific SNP. , The second correction constant for correcting the identification error corresponding to the combination of SNPs including a specific SNP, the number of times V for the evaluation value calculation unit 5 to calculate the evaluation value, and the minimum identification error selection unit 7 for the identification error. It includes the number of times U of performing the process of selecting the minimum SNP combination.

Next, the matrix initialization unit 8 initializes the SNP combination matrix (step S13). Initialize each element of the SNP combination matrix to 0 or 1. At initialization, it is arbitrary whether each junction element of the SNP combination matrix is set to 0 or 1.

FIG. 3 is a diagram showing an example of an SNP combination matrix. FIG. 3 shows an example of K = 2. In step S13 of FIG. 12, each element of the SNP combination matrix is included, and the SNP combination matrix of FIG. 3 contains 2 × 15 = 30 conjugate elements. As described above, since one SNP has three types of joints, it has three joint elements, and each joint element can take 0 or 1. Finally, each row of the SNP combination matrix corresponds to the combination of SNPs.

Next, the variable u for measuring the number of repetitions is initialized to 0 (step S14). Subsequently, the variable v for representing the v-th selected join element from the plurality of join elements included in the SNP combination matrix is initialized to 0 (step S15).

Next, the v-th junction element is acquired from the plurality of junction elements in the SNP combination matrix based on the mutual information (step S16). Next, the score, which is an evaluation value, is calculated according to the following procedure (step S17). The process of step S17 is performed by the maximum joint element selection unit 9 in the evaluation value calculation unit 5.

In the calculation of step S17, it is considered that the selection of the v-1 th element is completed and the vth element is selected. Let S (k, i) be the score which is the evaluation value of the i-th junction element of the k-th combination of the SNP combination matrix. First, let R be a set of already selected join element consisting of v-1 elements. Among them, one of the already selected conjugate elements is the j-th element of the l-th SNP combination in the SNP combination matrix. The modified mutual information RI indicating the redundancy between the i-th element of the k-th SNP combination in the SNP combination matrix and the j-th element of the l-th SNP combination in the SNP combination matrix is as follows (1). ) Defined by the formula.

Here, T _l is a set of samples that are identified as negative (not identified as negative) by K-1 SNP combinations excluding the l-th SNP combination in the SNP combination matrix. Further, T _{k and l} are a set consisting of samples of the intersection of _{T l} and T _k.

Further, I (X _{Tk, l, j} , X _{Tk, l, i} ) is a mutual information amount between the j-th junction element and the i-th junction element with respect to the sample belonging to _{T k, l.} At this time, S (k, i) is calculated by the following equation (2). However, I (Y _T , _{XT k , i} ) is a mutual information regarding the presence or absence of disease and the i-th zygote element for the sample belonging to T k.

Next, the join element having the highest score is selected from all the join elements in the SNP combination matrix (step S18). The process of step S18 is performed by the maximum joint element selection unit 9 in the evaluation value calculation unit 5.

FIG. 13 is a diagram showing an example of the score of the SNP combination matrix when it is determined that the variable v has reached V in step S18. In the example of FIG. 13, the zygote element of SNP-00003-YY is selected with k = 2 having a score of 0.9.

Here, for the SNP specified in advance by the user in the specific SNP designation unit 2, S (k, i) + ε and the score value are increased by ε (ε> 0), and the score value is intentionally increased. It may be easy to be selected.

In this way, the processes of steps S17 and S18 are performed for each value of the variable v, and one junction element having the maximum score in Eq. (2) is selected. The process of step S18 is performed by the plurality of joint element selection units 10 in the evaluation value calculation unit 5.

Next, it is determined whether or not the variable v has reached the predetermined limit number V (step S19). If it has not been reached yet, the variable v is incremented by 1 (step S20), and the processes of steps S16 to S19 are repeated.

When the variable v reaches the predetermined limit number V, V joined elements are selected. ^{Therefore, the identification error is calculated for each of the 2 V} combinations in which the V junction elements of the SNP combination matrix take 0 or 1, respectively (step 21), and the combination that minimizes the identification error is selected from among them. Search and update the SNP combination matrix (step S22). However, for the junction elements other than V in the SNP combination matrix, the identification error is calculated based on the values at present. The process of step S21 is performed by the identification error calculation unit 6. The process of step S22 is performed by the minimum identification error selection unit 7.

In calculating the discrimination error, it is necessary to identify the presence or absence of disease in the target sample. FIG. 4 is a flowchart for identifying the presence or absence of a disease in the target sample. First, the ID of the sample to be identified is acquired (step S31), and then the variable Z used for identifying the presence or absence of a disease is initialized to 0 (step S32). Next, the variable k that specifies the number of rows of the SNP combination matrix is initialized to 1 (step S33). Next, it is determined whether or not the sample of the ID acquired in step S31 has all the conjugates of each SNP included in the k-th SNP combination indicated by the SNP combination matrix (step S34). If YES is determined in step S34, the variable Z is incremented by 1 (step S35).

If NO is determined in step S34, or if the process of step S35 is completed, it is determined whether or not the variable k has reached the value obtained by adding 1 to the number of rows K of the SNP combination matrix (step S36). .. If NO is determined in step S36, the variable k is incremented by 1 (step S37), and the processes after step S34 are repeated.

When step S36 is YES, if the variable Z is 1 or more, the sample with the ID acquired in step S31 is identified as having a disease, and if the variable Z = 0, it is identified as having no disease (step S38).

The discrimination error is the sum of the number of samples in which the sample identified as positive is actually negative and the number of samples in which the sample identified as negative is actually positive. However, when a specific SNP specified by the user is included, the identification error is subtracted by α (0 <α <1) times to reduce the identification error, so that the specific SNP can be easily selected.

FIG. 14 is ^{a diagram showing identification errors of three of the 2 V} combinations (hereinafter, combinations c1 to c3). The identification error of the combination c1 of FIG. 14 is 4, the identification error of the combination c2 of FIG. 14 is 3, and the identification error of the combination c3 of FIG. 14 is 2. Therefore, in step S19 of FIG. 12, the junction elements in the combination c3 of FIG. 14 having an identification error of 2 are finally selected, and a new SNP combination matrix containing these junction elements is generated. For example, in the case of the SNP combination matrix of c3 in FIG. 14 (SNP-00001 is XX and SNP-00002 is YY) or SNP-00003 is YY, the sample is identified as positive.

FIG. 15 is a flowchart showing an example of the procedure for updating the SNP combination matrix. First, the current SNP combination matrix is acquired (step S41), and the V joined elements calculated by the evaluation value calculation unit 5 are acquired (step S42).

Next, the variable i is initialized to 0 (step S43). Next, update the SNP combination matrix i-th combination among the 2 ^V-number combinations are all combinations of either zero or one each of the V of the joining element (step S44). Next, with respect to the SNP combination matrix selected in step S44, the processing of step S21 of FIG. 12 is performed to calculate the identification error (step S45).

Next, it is determined whether or not the variable i has ^{reached 2 V (step S46).} If step S46 is NO, the variable i is incremented by 1 (step S47), and the processes after step S44 are repeated. If step S46 is YES, the SNP combination matrix is updated to the smallest combination from ^{the 2 V combinations of identification errors (step S48).}

When the process of step S22 of FIG. 12 is completed, it is next determined whether or not the variable u has reached the predetermined limit number of times U (step S23). If the limit number U has not been reached yet, the variable u is incremented by 1 (step S24), and the processing after step S15 is repeated using the new SNP combination matrix generated in step S22.

As described above, in the process of FIG. 12, the process of updating the junction element is performed U times while updating the SNP combination matrix.

When it is determined in step S20 that the variable u has reached the limit number of times U, the combination of the conjugate elements finally searched in step S22 is output as a biomarker candidate (step S25).

FIG. 16 is a diagram showing an example of the output form of step S22. The window w11 (first window) of FIG. 16 specifies a specific SNP designated by the specific SNP designation unit 2. Window w12 (second window) lists all the specified specific SNPs. Window w13 (third window) specifies a specific disease type. Window W14 (fourth window) displays each SNP in the biomarker candidate. After confirming the biomarker candidates in the window w14, the user can specify the specific SNP again in the window w11, press the resubmit button b2, and search for the biomarker candidates in FIG. 12 again. be.

The output form of step S22 is not limited to the screen display example shown in FIG. For example, FIG. 17 is a plot of odds values or -logs (P values) indicating the identification accuracy of each SNP combination. The horizontal axis is the number including a specific SNP, and the vertical axis is the odds value or -log (P value). The broken line in FIG. 17 is the availability identification threshold. This availability identification threshold value is generated by adding the region value input by the user in the search condition input unit 21 and the average value or standard deviation value of the odds value or −log (P value). Among the biomarker candidates, when at least one specific SNP specified by the user is included, the specific SNP is displayed as available, and in other cases, the error is displayed as the specific SNP unavailable 23.

Based on the result of FIG. 17, the user can re-search the biomarker candidate by changing the search condition such as a specific SNP or the search condition by using FIG.

As described above, in the present embodiment, a specific SNP presumed to be related to a specific disease is input in advance from a plurality of SNPs in the genome, and based on the input specific SNP and the trait information of the sample. Therefore, biomarker candidates containing one or more SNPs that are presumed to be related to a specific disease are searched for. Thereby, when the information that a specific SNP is related to a specific disease is known from the knowledge of a doctor, the biomarker candidate can be searched by taking the information into consideration.

Further, according to the present embodiment, since the biomarker candidate including two or more SNPs can be searched, the combination of SNPs can be accurately searched for the disease caused by the combination of a plurality of SNPs.

Further, according to the present embodiment, whether or not the process of selecting the joint element having the maximum evaluation value from the joint elements is performed V times and each of the selected V joint elements is selected as the SNP. calculating an identification error for Kano 2 ^V street, because the identification error is to select a minimum SNP combinations as final biomarker candidates, without omission combinations of SNP applicable from a large SNP information, a short time You can select with.

At least a part of the biomarker search apparatus described in the above-described embodiment may be configured by hardware or software. When configured by software, a program that realizes at least a part of the functions of the biomarker search device may be stored in a recording medium such as a flexible disk or a CD-ROM, read by a computer, and executed. The recording medium is not limited to a removable one such as a magnetic disk or an optical disk, and may be a fixed recording medium such as a hard disk device or a memory.

Further, a program that realizes at least a part of the functions of the biomarker search device may be distributed via a communication line (including wireless communication) such as the Internet. Further, the program may be encrypted, modulated, compressed, and distributed via a wired line or wireless line such as the Internet, or stored in a recording medium.

Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other embodiments, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are also included in the scope of the invention described in the claims and the equivalent scope thereof.

1 Biomarker search device, 2 Specific SNP designation unit, 3 Candidate search unit, 4 Candidate output unit, 5 Evaluation value calculation unit, 6 Identification error calculation unit, 7 Minimum identification error selection unit, 11 SNP trait DB, 12 Specimen information input Part, 13 Specimen information registration DB, 14 SNP information registration DB, 8 Matrix initialization part, 9 Maximum junction element selection unit, 10 Multiple junction element selection unit, 11 SNP trait DB, 12 Construction information input unit, 13 Samples Information registration DB, 14 SNP information registration DB, 15 specific SNP registration DB, 16 related SNP registration DB, 17 search range SNP selection unit, 18 selection SNP registration DB, 19 search information collation unit, 20 search information registration DB, 21 search conditions Input section, 22 Biomarker candidate registration DB

Claims (12)

A specific SNP designation unit that pre-designates a specific SNP known to be related to a specific disease from a plurality of SNPs (Single-Nucleotide Polymorphisms) in the genome.
Each conjugate element is selected as a combination of SNPs for each SNP based on the specific SNP , disease presence / absence information of each sample, and search information in which a junction type matrix representing the junction type of the SNP is registered. The process of calculating the evaluation value indicating the high possibility is performed V (V is an integer of 2 or more) times, and the V joint elements corresponding to the maximum value of the evaluation value each time are collected. For any combination of the conjugate elements, an identification error representing the degree to which the presence or absence of a disease in each sample could be correctly identified was calculated, the combination of the conjugate elements having the smallest identification error was selected, and the selection was made. By updating the biomarker candidates to the conjugate element and repeating the update U (U is an integer of 2 or more) times, a biomarker candidate containing two or more SNPs presumed to be related to a specific disease is searched for. Candidate search department and
A biomarker search device including a candidate output unit that outputs the biomarker candidate. The biomarker search device according to claim 1, wherein the candidate search unit searches for the biomarker candidate containing two or more SNPs known to be related to the specific disease. The biomarker search device according to claim 2, wherein the candidate output unit uses the combination of the joined elements selected with the minimum identification error as the biomarker candidate after the U times of processing. The biomarker search device according to any one of claims 1 to 3, further comprising a search condition input unit for inputting the value of V and the value of U. The candidate search unit initializes an SNP combination matrix in which the number of combinations of SNPs that can be biomarker candidates is the number of rows and the number of conjugate elements for a plurality of SNPs that can be biomarker candidates is the number of columns. Has a chemical unit,
A maximum joint element selection unit that acquires each joint element in the SNP combination matrix, calculates the evaluation value, and selects the joint element having the maximum evaluation value.
A plurality of joint element selection units that repeat the process of the maximum joint element selection unit V times and select a total of V different joint elements for each time of the processing.
Any one of claims 1 to 4 , further comprising an identification error calculation unit for calculating the identification error for all combinations of the V joint elements selected by the plurality of joint element selection units. The biomarker search device according to the section. The maximum junction element selection unit is for each junction element in the k (k is an integer of 1 or more) rows in the SNP combination matrix, and each junction in rows other than the k row in the SNP combination matrix. The biomarker search device according to claim 5 , wherein the evaluation value is calculated based on mutual information with a body element. The candidate output unit displays on a two-dimensional plane the correspondence between the number of the specific SNPs included in the biomarker candidate and the value representing the significance of the corresponding SNP for each SNP in the biomarker candidate. The biomarker search apparatus according to any one of claims 1 to 6. The biomarker search apparatus according to claim 7 , wherein the value representing the significance includes at least one of a P value and an odds ratio. The first window that specifies the specific SNP specified by the specific SNP designation unit, and
A second window listing all of the specified SNPs,
A third window that specifies the type of the particular disease,
A fourth window for specifying the conditions of the candidate search unit, and
The biomarker search device according to any one of claims 1 to 8 , further comprising a display control unit for displaying the above on the display screen of the display device. The first window that specifies the specific SNP specified by the specific SNP designation unit, and
A second window listing all of the specified SNPs,
A third window that specifies the type of the particular disease,
A fourth window for displaying each SNP in the biomarker candidate and a display control unit for displaying on the display screen of the display device are provided.
The biomarker search according to any one of claims 1 to 8 , wherein the display control unit highlights the SNP included in the biomarker candidate among the specific SNPs designated in the first window. Device. From multiple SNPs (Single-Nucleotide Polymorphisms) in the genome, a specific SNP known to be related to a specific disease is specified in advance.
Each conjugate element is selected as a combination of SNPs for each SNP based on the specific SNP , disease presence / absence information of each sample, and search information in which a junction type matrix representing the junction type of the SNP is registered. The process of calculating the evaluation value indicating the high possibility is performed V (V is an integer of 2 or more) times, and the V joint elements corresponding to the maximum value of the evaluation value of each time are collected. For any combination of the conjugate elements, an identification error representing the degree to which the presence or absence of a disease in each sample could be correctly identified was calculated, the combination of the conjugate elements having the smallest identification error was selected, and the selection was made. By updating the biomarker candidates to the conjugate element and repeating the update U (U is an integer of 2 or more) times, a biomarker candidate containing two or more SNPs presumed to be related to a specific disease is searched for. death,
A biomarker search method that outputs the biomarker candidate. A procedure for pre-designating a specific SNP known to be related to a specific disease from a plurality of SNPs (Single-Nucleotide Polymorphisms) in the genome.
Each conjugate element is selected as a combination of SNPs for each SNP based on the specific SNP , disease presence / absence information of each sample, and search information in which a junction type matrix representing the junction type of the SNP is registered. The process of calculating the evaluation value indicating the high possibility is performed V (V is an integer of 2 or more) times, and the V joint elements corresponding to the maximum value of the evaluation value of each time are collected. For any combination of the conjugate elements, an identification error representing the degree to which the presence or absence of a disease in each sample could be correctly identified was calculated, the combination of the conjugate elements having the smallest identification error was selected, and the selection was made. By updating the biomarker candidates to the conjugate element and repeating the update U (U is an integer of 2 or more) times, a biomarker candidate containing two or more SNPs presumed to be related to a specific disease is searched for. And the procedure to do
A program for causing a computer to execute the procedure for outputting the biomarker candidate.

Images