A method for identifying cancer driver genes based on cross entropy measurement

Abstract

The present invention discloses a method for identifying cancer driver genes based on cross entropy measurement, including the steps of: obtaining a cancer somatic mutation genomic dataset, constructing a mutation matrix based on the genomic dataset, and calculating the mutation frequency of the gene; mapping the mutation frequency of each gene to the gene function interaction graph information to obtain a gene mutation network, calculating the gene centrality score, calculating the gene improved cross entropy score, and arranging the potential scores of all mutant genes in order from high to low, and the higher the score, the more likely the gene is a driver gene. The present invention is an important supplement to the existing gene exclusion-based methods. The present invention can mine low-exclusion driver genes that are difficult to find with current methods. The present invention can also reduce the impact of the existing methods being prone to errors or incompleteness of known protein networks and biological pathways, thereby limiting their performance.

Description

Technical Field

[0001] The present invention relates to the technical field of bioinformatics, and in particular to a method for identifying cancer driver genes based on cross entropy measurement. Background Art

[0002] Cancer is caused by mutations in somatic genes, leading to malignant cell proliferation. It is essentially a hereditary disease with complex genomic changes. In the development of cancer, only a small number of mutated driver genes have selective growth effects and play an important role in cancer development, while most mutated passenger genes have no functional consequences for the development of cancer. Therefore, how to accurately identify driver genes among a large number of passenger genes is of great significance for the diagnosis, prevention and treatment of cancer and the understanding of pathological mechanisms.

[0003] The latest method for identifying driver genes based on mutual exclusivity of genes is "An Entropy-Based Methodfor Identifying Mutual Exclusive Driver Genes in Cancer, 2019" (EntroRank for short). EntroRank first downloaded the mutant genes of each cancer patient, constructed a matrix of mutant genes of the patient, and then mapped the mutant genes to the influence graph to generate a mutation graph. Secondly, the mutation graph is divided into different subgroups according to whether the genes are located in the same subcellular interval. Third, the structure of the mutant genes calculates the structural entropy of the mutant genes in the subgroup and the relative entropy between the mutant genes and the adjacent genes in the subgroup. In addition, the importance of the genes in each subgraph will be multiplied by the corresponding subcellular weight. If a mutated gene is located in more than one subcellular interval, the final score of the gene with the largest score is selected. Finally, the final scores of the driver genes according to the mutation are arranged from high to low. The scores are arranged from high to low, and the genes ranked in front are selected as predicted cancer driver genes.

[0004] At present, the mutual exclusivity-based method can only identify some highly exclusive genes, but cannot identify weakly exclusive driver genes. The exclusion-based method still belongs to the category of biological network methods, and is largely affected by the known protein networks and biological pathways that are prone to errors or incompleteness, which limits its performance.

[0005] The reason is that in biological pathways, only a few gene mutations are enough to disrupt the normal function of the corresponding entire biological process, and most mutated genes have weak exclusivity. Most exclusivity-based methods rely on the mutual exclusivity of genes. If the known protein networks and biological pathways are prone to errors or incomplete effects, the calculation of mutual exclusivity will be wrong, thus limiting its performance. Summary of the invention

[0006] The technical problem to be solved by the embodiment of the present invention is to provide a method for identifying cancer driver genes based on cross entropy measurement. The improved cross entropy is used to simultaneously measure the gene mutation frequency and mutual exclusivity, thereby achieving the complementary effect of exclusivity and mutation rate, providing a new way to identify the real driver genes.

[0007] In order to solve the above technical problems, an embodiment of the present invention provides a method for identifying cancer driver genes based on cross entropy measurement, which is characterized by comprising the following steps:

[0008] S1: Obtain cancer somatic mutation genomic dataset, download subcellular compartment information S and gene function interaction map information G , and stored in a genomics database;

[0009] S2: Construct a mutation matrix based on the genomic dataset M , calculate the mutation frequency of the gene f i ; Map the mutation frequency of each gene to the gene function interaction graph information G Gene mutation network G’ ;

[0010] S3: Calculate gene centrality scores, calculate gene improved cross entropy scores; use degree centrality to measure gene i In the mutation subnet S k Medium impact β i k , the gene-improved cross entropy score is calculated using the following formula:

[0011]

[0012]

[0013] in, H i k The crossover score of gene i in mutation network k that is identified as a driver gene, N i k is the number of neighbor nodes of gene i in mutation network k;

[0014] S4: Calculate the fraction of gene i identified as a driver gene s i k

[0015] s i k =βi k *H i k

[0016] Use the potential score of gene i s i The maximum value is taken as the final result:

[0017] s i =max(s i 1 ,s i 2 ,s i 3 ,…，s i k ), k=1,2,…,K , where k represents the daughter cell compartment.

[0018] Each of the cancer somatic mutation genomic datasets contains somatic mutation information of different patients.

[0019] Among them, the mutation frequency fi The calculation method is

[0020] , where the rows of M represent genes and the columns represent patients. ij It represents the number of mutations of gene j in patient i, and NP represents the number of patients.

[0021] Among them, S2 also includes dividing the genes in different sub-cell compartments into different mutation sub-networks G k ' (F,E) steps.

[0022] The degree centrality is used to measure the gene i in the mutation subnetwork S k Medium impact β i k The calculation formula is as follows:

[0023]

[0024] in, d(O i k ) daughter cell compartment k The degree of gene i, NG k represents the number of nodes within daughter cell compartment k.

[0025] Wherein, the S3 also includes respectively giving SH i and RH ij Methods to assign different weights:

[0026] .

[0027] The implementation of the embodiments of the present invention has the following beneficial effects: the present invention designs a cancer driver gene prediction method based on improved cross entropy based on the repulsion and mutation rate of driver genes, which is an important supplement to the existing gene repulsion-based methods. The present invention can mine low-repulsion driver genes that are difficult to find with current methods. In addition, the present invention can also reduce the impact of existing methods on known protein networks and biological pathways that are prone to errors or incompleteness, thereby limiting their performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is an overall flow chart of the method for identifying cancer driver genes. DETAILED DESCRIPTION

[0029] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings.

[0030] from Figure 1 It can be found that a cancer driver gene driving method is mainly divided into three important step modules, namely data processing module, mutation network partition module and driver gene identification module. The following are detailed introductions to these three step modules:

[0031] Step 1: Data processing module, download cancer somatic mutation genomic datasets (Data) from the TCGA project. Each dataset contains somatic mutation information of different patients. Then, perform duplicate deletion on the Data dataset. Download subcell compartment information (S) and gene function interaction graph information G (V, E) from the COMPARTMENTS database, where V represents the set of gene nodes and E represents the set of edges connecting genes. Then, store the processed dataset Data, subcell compartment information, and gene function interaction graph information in the genomics database.

[0032] Step 2: Mutation network partitioning module, first constructs a mutation matrix M based on the genomic data set (Data) obtained by the data processing module. The rows of M represent genes and the columns represent patients. ij Indicates gene j In patients i NP represents the number of patients. Calculate the mutation frequency of the gene fi as follows:

[0033] (1)

[0034] Secondly, the mutation frequency of each gene is mapped to the gene function interaction graph information G(V,E) to obtain the gene mutation network G'(F,E), as shown in formula (2):

[0035] G(V,E)→G΄(F,E) (2)

[0036] Among them, F represents the set of gene mutation frequencies in the mutation network, E represents the set of gene connection edges, and when gene i and gene j affect each other, e ij =1, otherwise e ij =0.

[0037] Through the above method, the present invention obtains the mutation network G'(F,E). According to research, different genes play different roles in different sub-cell compartments and have different effects on cell development and differentiation. In connection with the problem dealt with in this article, genes in different sub-cell compartments can be divided into different mutation sub-networks G k ' (F,E), such as the cell compartment S k Gene i Divided into mutation subnet G k ' (F,E), S k represents the kth daughter cell compartment, as shown in formula (3).

[0038] (3)

[0039] Step 3: Driver gene identification module. The driver gene identification module is mainly divided into two important functions: calculating the gene centrality score and calculating the gene improved cross entropy score. Considering that in the gene interaction network, the greater the influence of the mutant gene in the network, the greater its ability to destroy the normal function of the corresponding entire biological process when it mutates, and its potential repulsion is also strong. Such a gene may be a driver gene. Therefore, this paper uses degree centrality to measure the gene i in the mutation subnetwork. S k Medium impact β i k , as shown in formula (4).

[0040] (4)

[0041] Among them, d(O ik ) is the degree of gene i in daughter cell compartment k, NG k represents the number of nodes within subcell compartment k.

[0042] After calculating the gene centrality score, the improved cross entropy score of the gene is calculated. The purpose of calculating the improved cross entropy score of the gene in this paper is to simultaneously measure the gene's exclusion information and mutation frequency information. The common exclusion driver gene identification method has certain defects, such as only identifying some highly exclusionary genes, but not being able to identify weakly exclusionary driver genes well, and being affected by the known protein network and biological pathways that are prone to errors or incompleteness, which limits its performance. In response to the above problems, this paper designs and uses a calculation method that can simultaneously measure the gene's exclusion information and mutation frequency information, namely the improved cross entropy, as shown in formula (5).

[0043]

[0044] (5)

[0045] in, H i k The crossover score of gene i in mutation network k that is identified as a driver gene, N i k is the number of neighbor nodes of gene i in mutation network k.

[0046] Considering that the SH i and RH ij There is a large order of magnitude, which makes SH i and RH ij There is inherent inequality in the ability to contribute to the model. For this reason, the present invention provides SH i and RH ij Different weights are assigned as shown in formula (6). The weights can be determined using the grid search method:

[0047] (6)

[0048] Through the above-mentioned method, the present invention calculates the centrality score and improved cross entropy score of each gene. By combining the centrality and cross entropy characteristics of the gene, the present invention calculates the score s of gene i being identified as a driver gene i k , that is, the potential score, as shown in formula (7):

[0049] s i k =βi k *H i k (7)

[0050] Among them, s i k represents the potential score of gene i in mutation network k, H i k The crossover score of gene i in mutation network k that is identified as a driver gene, β i k The centrality score of gene i in mutation subnetwork k.

[0051] Since there are k subcellular compartments and some genes may appear in multiple subcellular compartments, the present invention selects the potential score s of gene i calculated by equation (8) i The maximum value is taken as the final result

[0052] s i =max(s i 1 ,s i 2 ,s i 3 ,…，s i k ), k=1,2,…,K (8)

[0053] Finally, this article arranges the potential scores of all mutant genes in order from high to low. The higher the score, the more likely the gene is to be a driver gene.

[0054] The above disclosure is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for identifying cancer driver genes based on cross entropy measurement, characterized in that: The following steps are involved: S1: Obtain cancer somatic mutation genomic dataset, download subcellular compartment information S and gene function interaction map information G , and stored in a genomics database; S2: Construct a mutation matrix based on the genomic dataset M , calculate the mutation frequency of the gene f i ; Map the mutation frequency of each gene to the gene function interaction graph information G Gene mutation network G’ ; S3: Calculate gene centrality scores, calculate gene improved cross entropy scores; use degree centrality to measure gene i In the mutation subnet S k Medium impact β i k , the gene-improved cross entropy score is calculated using the following formula: in, H i k is the crossover score of gene i in mutation network k that is identified as a driver gene, N i k is the number of neighbor nodes of gene i in mutation network k; S4: Calculate the fraction of gene i identified as a driver gene s i k s i k =β i k *H i k Use the potential score of gene i s i The maximum value is taken as the final result: s i =max(s i 1 ,s i 2 ,s i 3 ,…，s i k ), k=1,2,…,K , where k represents the daughter cell compartment.

2. The method for identifying cancer driver genes based on cross entropy measurement according to claim 1, characterized in that: Each of the cancer somatic mutation genomic datasets contains somatic mutation information of different patients.

3. The method for identifying cancer driver genes based on cross entropy measurement according to claim 2, characterized in that: The mutation frequency fi The calculation method is , where the rows of M represent genes and the columns represent patients. ij It represents the number of mutations of gene j in patient i, and NP represents the number of patients.

4. The method for identifying cancer driver genes based on cross entropy measurement according to claim 3, characterized in that: The step S2 also includes dividing the genes in different daughter cell compartments into different mutation subnetworks G k ' (F,E) steps.

5. The method for identifying cancer driver genes based on cross entropy measurement according to claim 4, characterized in that: The degree centrality is used to measure the gene i in the mutation subnetwork S k Medium impact β i k The calculation formula is as follows: in, d(O i k ) daughter cell compartment k The degree of gene i, NG k represents the number of nodes within subcell compartment k.

6. The method for identifying cancer driver genes based on cross entropy measurement according to claim 5, characterized in that: The step S3 also includes giving SH i and RH ij Methods to assign different weights: 。

Images