KR102447359B1 - Apparatus and method for predicting novel disease genes based on the integration of diverse gene-gene relations - Google Patents

Abstract

A method of operating an apparatus for discovering a disease gene, comprising: constructing a network in which genes and functions are connected by combining relationship information between genes and functional group information including genes; and an analysis cycle for selecting a new disease gene candidate from the network; repeating steps. The analysis cycle includes calculating the disease significance score of each functional group based on the statistical significance of the disease genes included in each functional group. Calculating the functional similarity disease score, network propagating the initial disease score of each gene reflecting the functional similarity disease score of each gene, and selecting a new disease gene candidate based on the disease score of genes calculated through the network propagation and adding the new disease gene candidate to disease gene information used in the next analysis cycle.

Description

Apparatus and method for predicting novel disease genes based on the integration of diverse gene-gene relations

The present invention relates to a novel disease gene prediction technology

A disease gene is a gene that can modulate a disease and is a drug target candidate. Therefore, the technology for predicting disease genes is one of the technologies most needed in the pharmaceutical industry for efficient treatment methods and drug development. With the recent development of high-throughput screening technology, a number of studies are underway to discover drug target candidates or disease genes through experiments. However, excavation through experiments consumes a lot of time and money, so the discovery results are far below the expected number of drug target candidates. Recently, a method of predicting disease genes through computational technology and experimentally verifying them to discover drug target candidates at low cost and in a short time has been proposed.

The principle of disease gene prediction techniques can be summarized as scoring similarity with previously known disease genes through various gene association relationships. These technologies are largely divided into functional similarity-based technologies and network propagation-based technologies. ToppGene and Endeavor predict new disease genes by scoring functional similarities with known disease genes, assuming that genes related to the same disease have similar functions in cells. On the other hand, PINTA and DADA are based on the assumption that genes related to the same disease have a high regulatory relationship with each other through protein interactions, etc. predict new disease genes.

PRINCE and HybridRanker predict disease genes by using the two technologies together to take advantage of the functional similarity-based technology and the network propagation-based technology. However, these techniques predict disease genes by simply summing the scores obtained after executing the functional similarity analysis process and the network propagation analysis process, respectively. Therefore, these techniques have a limitation in that they cannot derive the optimal performance that can be obtained by mutually reflecting the results from the two different methods in the calculation process of each method. In addition, these technologies are constructed based on significantly lower genetic linkage information compared to the currently increased information resources, and have a structure that limits the reasoning process using the linkage relationship to a single process. There are fundamental limitations.

The task to be solved is to provide a novel disease gene prediction device and method that calculates by linking functional group information and network characteristics of genes and recycles information discovered through an iterative process.

The task to be solved is to provide a device and method for improving the possibility of discovering disease genes by repeatedly applying the new disease-related relationships of genes discovered at each analysis cycle to the functional group analysis and network propagation process of genes.

According to an embodiment, there is provided an operating method of an apparatus for discovering a disease gene operated by at least one processor, the method comprising: constructing a network in which genes and functions are connected by combining information on a relationship between genes and information on a functional group including genes; and repeating the analysis cycle for selecting new disease gene candidates from the network. The analysis cycle includes calculating the disease significance score of each functional group based on the statistical significance of the disease genes included in each functional group. Calculating the functional similarity disease score, network propagating the initial disease score of each gene reflecting the functional similarity disease score of each gene, and selecting a new disease gene candidate based on the disease score of genes calculated through the network propagation and adding the new disease gene candidate to disease gene information used in the next analysis cycle.

By the new disease gene candidate selected in this cycle, the disease significance score of the functional groups to which the new disease gene candidate is related may vary in the next cycle.

The analysis cycle may further include scoring the disease significance of the gene with a pre-disease score for a known disease gene or a disease gene candidate selected in previous analysis cycles among genes included in the network. The initial disease score may be assigned as a weighted sum of a functional similarity disease score and a prior disease score of each gene.

In the scoring with the prior disease score, the prior disease score of the known disease gene and the prior disease score of the disease gene candidate selected in previous analysis cycles may be differentially scored.

In the calculating of the disease significance score of each functional group, the disease significance may be calculated based on the ratio of the disease gene and the general gene included in each functional group.

The method may further include terminating the analysis cycle without repeating the analysis cycle when there is no gene having a significant disease score through the network propagation.

The network includes a gene network combining relationship information between genes included in a plurality of public databases, and a gene-function network in which genes related to functional groups are linked, wherein at least one known disease gene is included in at least one functional group can be related to

According to another embodiment, there is provided a method of operating an apparatus for discovering a disease gene operated by at least one processor, for each functional group to which genes are related, based on the statistical significance of disease genes related to the functional group, the disease of the corresponding functional group. Calculating a significance score, calculating a functional similarity disease score of each gene by summing the disease significance scores of functional groups to which each gene is related in a gene network constructed with relationship information between genes, genes included in the gene network Among them, for a known disease gene or a new disease gene candidate, giving a prior disease score that scores the disease significance of the gene, for the genes included in the gene network, the function similarity disease score and the prior disease score allocating the weighted sum of , as the initial disease score of the corresponding gene, and selecting a new disease gene candidate based on the network propagation disease score of each gene calculated by network propagation of the initial disease score of each gene in the gene network. includes

Calculating the function similarity disease score of each gene assigns the disease significance score of each functional group to the related genes, and normalizes the log sum of the disease significance p-values assigned to each gene to obtain the function similarity disease score can be calculated

In the assigning of the prior disease score, a differential prior disease score may be given to the known disease gene and the new disease gene candidate.

In the calculating of the disease significance score, when a gene related to a specific functional group is selected as a new disease gene candidate in the previous analysis cycle, the disease significance score of the specific functional group is calculated by reflecting the statistical significance of the new disease gene candidate. can be calculated

When the disease significance score of the specific functional group is varied, the function similarity disease score of genes related to the specific functional group may be varied.

The gene selected as a new disease gene candidate in the previous analysis cycle is different from the initial disease score of the previous analysis cycle by giving a score different from the function similarity disease score and the prior disease score calculated in the previous analysis cycle in the current analysis cycle Can spread early disease scores.

The method may further include adding the new disease gene candidate to disease gene information used in a next analysis cycle.

The selecting of the new disease gene candidate may further include terminating the discovery of the new disease gene when there is no gene having a significant network disseminated disease score by the network propagation in the gene network.

According to an embodiment, the technology of the present invention can discover new disease genes that are more accurate than the conventional individual methods or the integrated method that statistically sums the results of the two methods by calculating the gene function group analysis and the network propagation analysis with each other. .

According to an embodiment, it is possible to contribute to performance improvement by applying disease-related information obtained for each analysis cycle to functional group analysis and network propagation analysis that are iteratively interconnected.

According to an embodiment using an actual disease gene that can be used as a disease marker and an actual drug target that can be used for drug development, the disease gene prediction technology of the present invention can be used together with discovery of companion diagnostic marker candidates through disease state identification and drug reactivity prediction. It can be used to discover drug target candidates.

According to the embodiment, by accelerating the grafting of IT-based hardware such as healthcare-related medical devices and web-based health management services, and BT-based biomarkers and drug target contents related to molecular biological information such as genomics, proteomics, and epigenomics, It can contribute to the development of the medical and pharmaceutical industries.

1 is a hardware configuration diagram of an apparatus for discovering a disease gene according to an embodiment.
2 is a flowchart of a method of operating an apparatus for excavating a disease gene according to an exemplary embodiment.
3 is a view for explaining an operation of an apparatus for discovering a disease gene according to an embodiment.
4 is a view showing a result of excavating a rheumatoid arthritis disease gene according to an embodiment.
5 and 6 are graphs of the results of evaluating the performance of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art to which the present invention pertains can easily implement them. However, the present invention may be embodied in several different forms and is not limited to the embodiments described herein. And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout the specification, when a part "includes" a certain component, it means that other components may be further included, rather than excluding other components, unless otherwise stated. In addition, terms such as “…unit”, “…group”, and “module” described in the specification mean a unit that processes at least one function or operation, which may be implemented as hardware or software or a combination of hardware and software. have.

1 is a hardware configuration diagram of an apparatus for discovering a disease gene according to an embodiment.

Referring to FIG. 1 , a disease gene discovery apparatus 100 is a computing device that executes a program in which an operation of the present invention is described by at least one processor.

The hardware of the disease gene discovery apparatus 100 may include at least one processor 110 , a memory 130 , a storage 150 , and a communication interface 170 , and may be connected through a bus. In addition, hardware such as an input device and an output device may be included. The disease gene discovery apparatus 100 may be equipped with various software including an operating system capable of driving a program.

The processor 110 is a device for controlling the operation of the disease gene discovery apparatus 100 , and may be various types of processors that process instructions included in a program, for example, a central processing unit (CPU), a microprocessor (MPU) processor unit), microcontroller unit (MCU), graphic processing unit (GPU), or the like. The memory 130 loads the corresponding program so that the instructions described to execute the operation of the present invention are processed by the processor 110 . The memory 130 may be, for example, read only memory (ROM), random access memory (RAM), or the like. The storage 150 stores various data, programs, etc. required to execute the operation of the present invention. The communication interface 170 is a wired/wireless communication module, and may interwork with an external database through a wired/wireless network.

The disease gene discovery apparatus 100 predicts a new disease gene by integrating the gene association relationship. The disease gene discovery apparatus 100 calculates the disease significance of the functional groups by analyzing the association between the disease gene and the functional group including the genes at every analysis cycle, and calculates the "function similarity disease score" of each gene therefrom. do. In addition, the disease gene discovery apparatus 100 calculates a "pre-disease score" of each gene. Thereafter, the disease gene discovery apparatus 100 network propagates the initial disease score of each gene in which the "function similarity disease score" and the "pre-disease score" are reflected, and analyzes the disease score finally obtained through the network propagation to obtain a new disease gene. discover candidates The disease gene discovery apparatus 100 reflects the new disease score of genes calculated in the previous analysis cycle in the calculation of the “function similarity disease score” and “pre-disease score” of the next analysis cycle, and through an iterative process of network propagation, It reflects indirect gene associations that cannot be reflected in other methods.

Hereinafter, an operation method of the disease gene discovery apparatus will be described in detail.

2 is a flowchart of a method of operating an apparatus for excavating a disease gene according to an exemplary embodiment.

Referring to FIG. 2 , the disease gene excavation apparatus 100 builds a network in which genes and functions are connected by combining intergene relation information collected from a plurality of public databases and functional group information including genes ( S110). In the network, genes corresponding to nodes are connected according to relationship information, and at least some of the genes are related to functions according to function group information. It is assumed that known disease genes are included among the genes constituting the network, and the disease genes are included in at least one functional group. The disease gene discovery apparatus 100 can build a network using various databases, for example, collects the protein-protein interaction network of HIPPIE and the signal transduction pathway of Graphite, and includes TRNASFAC, E3Net, PhosphoSitePlus and DEPOD. Relations between genes can be extended through regulatory relationship resources. In addition, the disease gene discovery apparatus 100 may collect functional groups from, for example, Molecular Signatures Database, Enrichr, and Gene Ontology. The disease gene discovery apparatus 100 may integrate similar functional groups from the collected functional groups.

The disease gene discovery apparatus 100 calculates a disease significance score of each functional group based on the statistical significance of disease genes included in each functional group (S120). The disease gene may be newly updated during the repetitive execution process of the disease gene discovery apparatus 100 , so that the disease significance of each function may be updated at each repeated analysis cycle by the disease genes newly included.

The disease gene discovery apparatus 100 calculates the "function similarity disease score" of each gene by summing the disease significance scores of all functional groups to which each gene is related ( S130 ). The disease gene discovery apparatus 100 may calculate a "function similarity disease score" of each gene by summing the disease significance scores of all functional groups associated with each gene, determining the disease score of each gene, and normalizing it.

In addition, the disease gene discovery apparatus 100 calculates a "pre-disease score" that scores a disease significance that is known for each gene or newly obtained by the apparatus, separately from the "function similarity disease score" (S140). In the initial initial disease score, the "pre-disease score" of an individual gene is determined only by whether there is a known disease significance, but the initial disease score in repeated cycles after the network propagation process is the disease obtained by network propagation. Significance is added and determined.

The disease gene discovery apparatus 100 assigns initial disease scores of gene nodes in the gene network by weighted summing the "function similarity disease score" and the "pre-disease score" for each gene (S150).

The disease gene discovery apparatus 100 calculates a new "network propagation disease score" of genes through a network propagation method of propagating an initial disease score assigned to a gene node in a gene network (S160). A "disease score" can be simply referred to as a disease score.

The disease gene discovery apparatus 100 determines whether there is a gene having a significant "network disseminated disease score" (eg, a p-value of 0.05 or less) (S170).

The disease gene discovery apparatus 100 updates genes having a significant network disseminated disease score as new disease gene candidates, and feeds them back to the next analysis cycle ( S180 ). The disease gene discovery apparatus 100 recalculates the disease significance of the functional group, the functional similarity disease score, and the pre-disease score by using the updated disease gene candidate information in the next analysis cycle.

The disease gene discovery apparatus 100 terminates the disease gene discovery process when no new disease gene candidates are no longer discovered because there are no genes having a significant network spread disease score ( S190 ).

In this way, the disease gene discovery apparatus 100 repeatedly updates the functional similarity disease score and the network disseminated disease score by mutual reflection and calculates the gene disease score, so that the disease gene can be discovered more accurately and efficiently than in the prior art. .

Next, the operation of the disease gene discovery apparatus will be described in detail.

3 is a view for explaining an operation of an apparatus for discovering a disease gene according to an embodiment. For the sake of explanation, a connection relationship (solid line) between genes (circles) may be called a gene network 10 , and a relationship relationship (dotted line) between genes and functions (square) may be called a gene-function association network 20 , The gene network 10 and the gene-function association network 20 may be collectively referred to as a network 30 .

Referring to FIG. 3A , the apparatus 100 for discovering a disease gene constructs a gene network 10 by using inter-gene relation information included in a plurality of public databases. Various databases may be used to construct the gene network 10 . As described above, the disease gene discovery device 100 collects the protein-protein interaction network of HIPPIE and the signal transduction pathway of Graphite, and expands the relationship between genes through regulatory relationship resources such as TRNASFAC, E3Net, PhosphoSitePlus and DEPOD. can do. In this case, the disease gene excavation apparatus 100 may define the number of relational resources to which the gene pair belongs as the relational reliability of the corresponding gene pair to configure gene pairs in which the degree of association is normalized. Through this, the gene network 10 consisting of 379,730 association relationships for a total of 15,165 genes can be constructed.

In addition, the disease gene discovery apparatus 100 collects functional groups, which are sets of genes grouped with the same function, from a plurality of public databases such as Molecular Signatures Database, Enrichr, and Gene Ontology. Since functional groups collected from different databases may be duplicated, the disease gene discovery apparatus 100 needs to integrate similar functional groups based on the similarity of the collected functional groups. To this end, the disease gene discovery apparatus 100 may extract similar functional groups from functional groups classified according to the type of function, and integrate them into one functional group. For example, the collected functional groups can be categorized into three categories: cellular function, molecular relation, and co-localization. The collected functional groups may be automatically classified by the disease gene discovery apparatus 100 or may be manually classified. The disease gene discovery apparatus 100 may integrate a pair of similar functional groups into one functional group based on a similarity measure such as a Jaccard coefficient. Through this, the disease gene discovery apparatus 100 can acquire 28,465 functional groups and gene information included in them from Molecular Signatures Database, Enrichr, and Gene Ontology, and assigns genes (sources) within each functional group to the corresponding functional group. The gene-function association network 20 can be constructed by connecting to the function (square) of

Referring to (b), the disease gene discovery apparatus 100 calculates the disease significance of all functions included in the gene-function association network 20 from the disease gene information related to the functional group.

For example, the disease gene discovery apparatus 100 determines the disease association of the function 21 based on the ratio of the disease gene (11, circle-black) and the general gene (12, circle-white) related to the function 21 . can be calculated The disease gene discovery apparatus 100 calculates the statistical significance of the function 21 as a p-value from the disease association values of all functions included in the gene-function association network 20 as a p-value. Disease significance can be assigned.

The disease significance p-value of each function may be calculated as in Equation 1. In this case, the p-value is adjusted by the false discovery rate (FDR).

In Equation 1, G is the total number of genes, S is the total number of disease genes, M is the number of genes in the functional group, and k is the number of disease genes in the functional group.

Referring to (c), the disease gene discovery apparatus 100 allocates an initial disease score P ₀ (g) for network propagation in the gene network 10 . The disease gene discovery apparatus 100 allocates a disease significance p-value (21, square-gray) of a function to genes linked to a corresponding function. In this case, since each gene can be associated with multiple functional groups. The "function similarity disease score" of each gene is calculated by aggregating the disease significance p-values of the functions associated with that gene. A "function similarity disease score" of a gene may be defined as the log sum of the disease significance p-values of the functions associated with the gene. The disease gene discovery apparatus 100 may adjust the functional similarity disease score of genes to a range of [0, 1] through min-max normalization. "Functional similarity disease score" S _f (g) of a gene can be expressed as Equation (2).

The conventional network propagation-based method allocates an initial disease score of each gene node by reflecting only previously known disease-significant gene information for a disease gene, and then obtains a final disease score by network propagation. On the other hand, the disease gene discovery apparatus 100 of the present invention provides a new disease gene score repeatedly obtained from the network propagation process of the device, along with a previously known disease gene score, and a novel disease gene score repeatedly obtained in the functional similarity analysis process shown above. Adjust the initial disease score to include all disease gene scores. In particular, in the present invention, a disease score (function similarity disease score) is given to a gene that is not known as a disease gene, but is functionally highly correlated with a disease, and network propagation is also possible. Through this, it is possible to overcome the limitation of the discovery of disease genes of the existing network propagation method, which is greatly affected by the initial value as well as the network structure.

는 가중치이다. The method of calculating the initial disease score of the disease gene discovery apparatus 100 may be expressed as Equation (3). The initial disease score P ₀ (g) of a gene can be defined as the weighted sum of the gene's "function similarity disease score" S _f (g) and "prior disease score" S _p (g). In Equation 3,

is the weight.

The prior disease score S _p (g) of Equation 3 is defined as Equation 4.

In Equation 4, the pre-disease score S _p (g) is 1 point for a previously known disease gene, and a range of [0, 1] for a new disease gene candidate discovered by the disease gene discovery apparatus 100 Scores within, genes that are not may be assigned 0.

For example, if the gene 11 is a known disease gene, the prior disease score of 1 is reflected in the initial disease score. The gene (12) is not a known disease gene, and if it does not receive a functional similarity score, the initial disease score is 0, but after the first discovery cycle of this device, it is selected as a candidate for a new disease gene, Disease score" can be reflected, and "function similarity disease score" can be reflected from disease significance p-values of the updated functions.

Referring to (d), the disease gene discovery apparatus 100 network propagates the initial disease score P ₀ (g) of genes (eg, 11 and 12 ) in the gene network 10 as linked genes. The gene's initial disease score P ₀ (g) is transmitted sequentially to neighboring genes at a rate of r along the network. In this case, the t-th spread disease score P ^t (g) may be defined as in Equation 5.

In Equation 5, W' is a normalized matrix of the weight adjacency matrix W, and is defined as Equation 6.

D denotes a diagonal matrix in which the diagonal element D(i, i) is the sum of the i-th row of W. The network propagation method may use other known techniques, such as an algorithm (Vanunu et al.) using the normalized weight adjacency matrix of Equation (5). It is different from the random walk with restart (RWR) method of previously developed disease gene discovery methods using network propagation in that the input/output flow of each gene is considered and normalized when the weight adjacency matrix W is normalized.

The disease gene discovery apparatus 100 Z-scores the disease score of the gene acquired through network propagation in the last stage of each analysis cycle, and then converts the gene having a significantly high score (p-value 0.05 or less) to the new disease gene. to be selected as

The disease gene discovery device 100 adds the selected new disease gene to the disease gene information of the device. Then, the disease gene discovery apparatus 100 recalculates the disease significance score of the functional groups and the “function similarity disease score” of each gene obtained therefrom using the updated disease gene, and the recalculated “function similarity disease score” Repeat the procedure for network propagation of the initial disease score of each gene that reflects

The disease gene discovery apparatus 100 repeats this procedure until no additional new disease genes are selected to discover disease gene candidates by linking all known disease significance of genes, correlation between genes and cellular functions, and correlation between genes. can

4 is a view showing a result of excavating a rheumatoid arthritis disease gene according to an embodiment.

Referring to FIG. 4 , the disease gene discovery apparatus 100 calculates the "function similarity disease score" of each gene, and network propagates the initial disease score reflecting the "function similarity disease score" and "pre-disease score" of the gene. Repeat the analysis cycle to discover new disease gene candidates.

In FIG. 4 , red is a test set selected from known disease genes, green is the rest of the known disease genes excluding the test set, blue is a disease gene candidate selected in each analysis cycle, and black is the rest of the genes.

The test set is 11 kinds of genes that are actually targets of drugs among rheumatoid arthritis, and the ranking change of these 11 kinds of genes is observed when the present invention is performed by inputting the remaining known disease genes.

As a result of the analysis, it can be seen that the score after network propagation increases in each analysis cycle for 11 test genes known as drug targets. In particular, in the fifth analysis cycle, it can be confirmed that the score of the test gene is as high as the existing known disease gene. In addition, it is confirmed that 10 genes out of 11 test genes are located in the 350th place out of 15,000 genes, and 6 of them are located in the 100th place. From this result, it can be confirmed that the device of the present invention accumulates and reflects the indirect association between the test gene and the existing disease genes through an inference process through repeated cycles, thereby enabling the discovery of actual drug targets excluding direct disease significance information.

5 and 6 are graphs of the results of evaluating the performance of the present invention.

In order to evaluate the performance of the present invention and to explain the differences from existing methods, we leave-one-out test known drug targets for modularized methods according to the working principle.

First, we collect known drug targets and disease gene information related to rheumatoid arthritis, cancer, and type 2 diabetes from DrugCentral, a drug target database, and DisGeNet, a known disease gene database. After selecting a sample from the collected drug targets and disease genes, a leave-one-out test that predicts the disease significance of the test gene is performed using the remaining genes except for one test gene. The performance is analyzed by expressing the leave-one-out test results for all known disease genes or drug targets as ROC curves.

Referring to FIG. 5 , as a result of comparing the results of performing the function group analysis-based prediction method and the network propagation-based prediction method respectively and the method in which the two methods are linked, the linked linkage developed in the present invention in all three diseases It can be seen that the performance of the method is excellent.

In particular, as a result of comparing the AUC of the ROC curve for cancer with the most existing disease gene information, it can be seen that the performance is significantly improved with a p-value of 0.05 or less.

Referring to FIG. 6 , it is a performance comparison result with the existing methods ToppGene, Random with Restart, and PRINCE. Each method in turn represents a functional similarity-based method, a network propagation-based method, and an existing integration method of the two methods. As a result of the performance comparison, it can be seen that the disease gene prediction ability of the present invention is the best, and a very significant performance improvement can be confirmed, particularly in cancer, with a p-value of 0.05 or less.

According to an embodiment, it can be confirmed that the performance of the method of the present invention is improved compared to the existing method of statistically integrating the results of the two methods by calculating the functional group analysis and the network propagation analysis in conjunction with each other. In addition, it can be confirmed that the process of repeatedly applying the disease-related information obtained through the cross-linkage calculation to the calculation also contributes to the performance improvement.

According to an embodiment using an actual disease gene that can be used as a disease marker and an actual drug target that can be used for drug development, the disease gene prediction technology of the present invention can be used together with discovery of companion diagnostic marker candidates through disease state identification and drug reactivity prediction. It can be used to discover drug target candidates.

In addition, according to the embodiment, by accelerating the grafting of IT-based hardware such as healthcare-related medical devices and web-based health management services and BT-based biomarkers and drug target contents related to molecular biological information such as genomics, proteomics, and epigenomics, , can contribute to the development of health care and pharmaceutical industries.

The embodiment of the present invention described above is not implemented only through the apparatus and method, but may be implemented through a program for realizing a function corresponding to the configuration of the embodiment of the present invention or a recording medium in which the program is recorded.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto. is within the scope of the right.

Claims (15)

As a method of operating a disease gene discovery apparatus operated by at least one processor,
A step of constructing a network in which genes and functions are connected by combining information on relationship between genes and information on functional groups including genes; and
Repeating the analysis cycle to select a new disease gene candidate in the network,
The analysis cycle is
calculating a disease significance score for each functional group based on the statistical significance of the disease genes included in each functional group;
In the network, calculating the function similarity disease score of each gene by summing the disease significance scores of the functional groups to which each gene is associated;
Network propagation of the initial disease score of each gene reflecting the functional similarity disease score of each gene;
selecting a new disease gene candidate based on the disease score of the genes calculated through the network propagation; and
adding the new disease gene candidate to disease gene information used in the next analysis cycle
comprising, a method of operation. In claim 1,
According to the new disease gene candidate selected in this cycle, the disease significance score of the functional groups associated with the new disease gene candidate in a next cycle is variable. In claim 1,
The analysis cycle is
Among the genes included in the network, for a known disease gene or a disease gene candidate selected in previous analysis cycles, the step of scoring the disease significance of the gene with a prior disease score,
The initial disease score is
Method of operation, assigned as the weighted sum of the functional similarity disease score and pre-disease score of each gene. In claim 3,
The step of scoring with the prior disease score is
A method of operating, differentially scoring the prior disease score of the known disease gene and the prior disease score of the disease gene candidate selected in previous analysis cycles. In claim 1,
The step of calculating the disease significance score for each functional group is
A method for calculating disease significance based on the ratio of disease genes and general genes included in each functional group. In claim 1,
If there is no gene having a significant disease score through the network propagation, terminating the analysis cycle without repeating
Further comprising, the method of operation. In claim 1,
the network is
A gene network combining relationship information between genes included in a plurality of public databases, and a gene-function network in which genes related to functional groups are connected,
wherein at least one known disease gene is associated with at least one functional group. As a method of operating a disease gene discovery apparatus operated by at least one processor,
For each functional group to which the genes are related, calculating a disease significance score for the functional group based on the statistical significance of the disease genes associated with the functional group;
calculating the function similarity disease score of each gene by summing the disease significance scores of the functional groups to which each gene is related in a gene network constructed with inter-gene relationship information;
Among the genes included in the gene network, for a known disease gene or a new disease gene candidate, giving a prior disease score that scores the disease significance of the gene;
allocating a weighted sum of the function similarity disease score and the prior disease score to the genes included in the gene network as an initial disease score of the gene; and
Selecting a new disease gene candidate based on the network propagation disease score of each gene calculated by network propagation of the initial disease score of each gene in the gene network
comprising, a method of operation. In claim 8,
The step of calculating the functional similarity disease score of each gene is
and assigning a disease significance score of each functional group to associated genes, and calculating the function similarity disease score by normalizing the log sum of disease significance p-values assigned to each gene. In claim 8,
The step of giving the prior disease score is
For the known disease gene and the new disease gene candidate, a differential prior disease score is given. In claim 8,
The step of calculating the disease significance score
When a gene related to a specific functional group is selected as a new disease gene candidate in a previous analysis cycle, the disease significance score of the specific functional group is calculated by reflecting the statistical significance of the new disease gene candidate. In claim 11,
When the disease significance score of the specific functional group is varied, the function similarity disease score of genes related to the specific functional group is varied. In claim 11,
The gene selected as a new disease gene candidate in the previous analysis cycle is different from the initial disease score of the previous analysis cycle by giving a score different from the function similarity disease score and the prior disease score calculated in the previous analysis cycle in the current analysis cycle Propagating early disease scores, a method of operation. In claim 8,
The method further comprising the step of adding the new disease gene candidate to disease gene information used in the next analysis cycle. In claim 8,
The step of selecting the new disease gene candidate
In the gene network, when there is no gene having a significant network spread disease score by the network propagation, terminating the discovery of a new disease gene
Further comprising, the method of operation.

Images