KR101930835B1 - A method and a system for producing combinational logic network based on gene expression - Google Patents

Abstract

The present invention relates to a method and system for generating combinational logic networks based on gene expression. According to the present invention, instead of using a representative value for the expression level of each gene in a plurality of samples in biological network modeling, more accurate modeling of the target path is performed because it reflects the expression level of each gene in all samples It is possible.

Description

Technical Field [0001] The present invention relates to a method and system for generating a combinatorial logical network based on gene expression,

The present invention relates to a method and system for generating combinational logic networks based on gene expression.

Recently, the development of DNA chip or microarray technology has made it possible to simultaneously measure the expression level of thousands of genes. These microarray data can be analyzed to identify genes that affect cancer progression or periodic changes in cells.

A network that shows the correlation between genes and genes is called a gene regulation network. There is a Boolean network as a way to study gene regulation networks. This method places the activity level of a gene in two states, on and off. Use mutual information among them to determine which combination of genes determines the level of activity of the next step in a gene. When this combination is revealed, a gene regulatory network can be drawn, in which each node represents a gene, and a binary relationship such as AND, OR, or NOT serves as an edge connecting the genes. Typically, a boolean network has been used for dynamic modeling using data recording gene expression patterns over time.

Numerous cross-sectional data sets have been generated and accumulated in the field of next-generation sequencing-based genomics. There is a need for a method for more efficiently modeling gene regulation networks from such data sets.

One aspect provides a method for generating a combinatorial logical network based on gene expression.

Another aspect provides a combinatorial logic network generation system based on gene expression.

One aspect includes generating a module having one or more input genes and one output gene from data representing expression levels of two or more genes obtained from two or more samples; Generating binary code for the module by converting the expression level of a gene belonging to a module generated in the data into a binary number and generating a binary code corresponding to the partial code A corresponding to the expression level of the input gene and the expression level of the output gene And a corresponding partial code B; When two arbitrarily selected binary codes of the generated binary codes have the same partial code A and different partial codes B, the two binary codes are placed in different groups, and the binary codes included in each group are divided into partial codes A and a corresponding single partial code B; And generating a logic circuit for each of a part or all of the groups. ≪ Desc / Clms Page number 2 >

As used herein, the term "module" is a group of unit genes that constitute a biological network, and may be referred to as a community, a cluster, or a subnetwork. The module may have a logical expression determined by an expression level of an input gene and an output gene. The logical representation may comprise a single Boolean operator or a combination of Boolean operators. The logical expression may be in the form of, for example, 0, 1, And (gene X, gene Y), ~ gene X, or Or (gene X, gene Y, gene Z).

 The module may be generated considering literature, experimental measurements, transcript database, biological pathway database, gene set analysis tool, or a combination thereof. The transcript database may be a microarray database. The microarray database may be, for example, Gene Expression Omnibus (GEO) from NCBI, ArrayExpress from EBI, The Cancer Geonome Atlas (TCGA), ArrayTrack, Stanford Microarray database, or Genevestigator. The biological pathway database includes a database accessible to and accessible to ordinary engineers such as KEGG, Gene Ontology, or BioCarta. The gene set analysis tool includes an analysis tool that can be accessed and used by ordinary technicians such as DAVID, GSEA, or PATHOME.

In this method, the conversion of gene expression levels to binary numbers can be accomplished through a transfer function. The transformation function may be, for example, the R package "Binarize" (Mundus et al., 2015). The binary code for the module may be expressed, for example, as 000, 010, 011, or 101. The code consists of a partial code A corresponding to the expression level of the input gene and a partial code B corresponding to the expression level of the output gene. For example, in a module composed of an input gene X, an input gene Y, and an output gene Z, the expression level of X and Y is 0 and 1, respectively, and the expression level of Z is converted into a binary number In this case, the binary code of this module is 011, the partial code A is 01, and the partial code B is 1.

The method includes grouping the generated binary codes. In the grouping step, if two arbitrarily selected binary codes of the generated binary codes have the same partial code A and different partial codes B, the two binary codes are placed in different groups. As a result, the partial code B corresponding to one type of partial code A in the binary code included in each group has a single value.

The method includes generating a logic circuit for each or some of the generated groups. A logic circuit means a representation method in which a process of obtaining an output value from an input value is represented by a logic symbol. The logic symbol may be a symbol commonly used in the field of electronics or a modification thereof.

The method may further comprise calculating a score based on the conditional probability for each of the groups. The score can be calculated by the following equation (1).

The sample may be obtained from a gastric cancer patient.

When two or more modules are generated in the module generation step, the method may further include selecting the target module before performing the step of converting the gene expression level to binary number.

In addition, when applying the method to two or more contiguous modules in which the output gene in one module is the input gene of another module, a wider portion of the biological network can be represented by the logic circuit. At this time, among the groups generated from each module, a logic circuit for the group having the highest score based on the conditional probability can be connected to represent the combinational logic network.

1 is a flowchart illustrating a combinational logical network generation method according to the present invention.

In another aspect, a module generation unit (110, 210) for generating a module having one or more input genes and one output gene from data indicating the expression levels of two or more genes obtained from two or more samples; A binary code conversion unit for converting the expression level of a gene belonging to a module generated in the data into a binary number to generate a binary code for the module, the code comprising a partial code A corresponding to the expression level of the input gene, And a partial code B corresponding to the level of the partial code; When two arbitrarily selected binary codes of the generated binary codes have the same partial code A and different partial codes B, the two binary codes are placed in different groups, and the binary codes included in each group are divided into partial codes A and a single partial code B corresponding thereto; And a logic circuit generation unit (140, 240) for generating logic circuits for all or a part of the groups, respectively.

The term "part" or the like refers to a unit for processing at least one function or operation, which may be implemented by hardware, software, or a combination of hardware and software.

The module generators 110 and 210 may generate modules considering literature, experimental measurements, transcript database, biological path database, gene set analysis tool, or a combination thereof. Samples, modules, transcript databases, biological pathway databases, and gene set analysis tools are described above.

The binary conversion units 120 and 220 may generate a binary code for the module by converting the gene expression level into a binary number through a conversion function. The conversion function and the binary code are as described above.

The system 100, 200 includes a group generator 130, 230 for grouping binary codes. When two arbitrarily selected binary codes have the same partial code A and different partial code B, the group generating units 130 and 230 respectively place the two types of binary codes into different groups, The binary code can be made up of the partial code A and the corresponding single partial code B,

The system 100, 200 includes logic circuit generators 140, 240 for generating logic circuits for some or all of the groups generated. The logic circuit is as described above.

The system 200 may further include a score calculation unit 235 for calculating a score based on the conditional probability for each group generated. The score calculation unit 235 may calculate scores for each group using the above equation.

When two or more modules are generated in the module generators 110 and 210, the system may further include a module selector for selecting the target module before performing the step of converting the gene expression level to binary numbers.

2A and 2B are diagrams illustrating a combinational logical network generation system according to the present invention.

According to the present invention, instead of using a representative value for the expression level of each gene in a plurality of samples in biological network modeling, more accurate modeling of the target path is performed because it reflects the expression level of each gene in all samples It is possible.

1 is a flowchart illustrating a combinational logical network generation method according to the present invention.
2A and 2B are diagrams illustrating a combinational logical network generation system according to the present invention.
Fig. 3 shows a graphical representation of the module VANGL2 using PRICKLE1 and PRICKLE2 as the input gene and VANGL2 as the output gene.
Figure 4 shows the gene expression levels and the values converted to binary numbers in the gastric cancer transcript data set GEO GSE15459.
Figure 5 shows an observation table of the module VANGL2.
Figure 6 shows four possible combinatorial logical networks for module VANGL2, with truth tables, scores, and logical representations.
FIGS. 7A and 7B show the results of applying the network model of the present invention to the lower path having WNT 5A and WNT 9A as start points and VANGL1 and VANGL2 as end points in the WNT signal propagation path.

Hereinafter, the present invention will be described in more detail with reference to the following examples. However, these embodiments are provided to aid understanding of the present invention, and the scope of the present invention is not limited thereto in any sense.

Example: Generation of a combinatorial logical network for gastric cancer gene expression

MRNA expression data of patients with intestinal type gastric cancer according to the Lauren classification (race, comparative group, profiling platform requirement) were obtained from the GEO accession GSE15459 data set. R package "Binarize" (Mundus et al., 2015, "Binarize: Binarization of One-Dimensional Data." Available from https://cran.r-project.org/package=Binarize) (TCBB) 9 (2): 487-498), which was used to generate the binary data.

Fig. 3 shows a graphical representation of the module VANGL2 using PRICKLE1 and PRICKLE2 as the input gene and VANGL2 as the output gene.

Figure 4 shows the gene expression levels and the values converted to binary numbers in the gastric cancer transcript data set GEO GSE15459. 4 (a) is data showing the expression levels of genes belonging to the module VANGL2 in various samples. Figure 4 (b) shows the values of each expression level shown in (a) converted to binary numbers by the BASC algorithm. In Fig. 4 (b), colorless indicates that the expression level is low or not expressed (binary 0), and black indicates that the expression level is high or expressed (binary 1).

Figure 5 shows an observation table of the module VANGL2. In Fig. 5, X denotes a "Don't-care term ", meaning that the output value is independent of the input value or that the input value is not observed in the patient.

As shown in FIG. 5, this module has binary codes 000, 001, 01X, 100, 101, and 110. The partial codes 00 and 10 for the input genes PRICKLE1 and PRICKLE2 can have both 0 and 1 as partial codes for the output gene VANGL2 and the partial codes 01 for PRICKLE1 and PRICKLE2 have X as the partial code for VANGL2 , Partial code 11 for PRICKLE1 and PRICKLE2 has 0 as a partial code for VANGL2.

The partial codes 00 and 10 for PRICKLE1 and PRICKLE2 have 0 or 1 as a partial code for VANGL2 so that each group has a single value for the partial gene for the output gene, Were divided into four different truth tables.

For each truth table, use the python library "PyEDA" (Drake, C, 2016, Retrieved September 19, 2016, Avaiable from https://github.com/cjdrake/pyeda) Respectively. Also, in order to distinguish predominant logical expressions with more samples, a score based on the conditional probability for each truth table was calculated according to Equation 1 below.

[Equation 1]

Figure 6 shows four possible combinatorial logical networks for module VANGL2, with truth tables, scores, and logical representations. As shown in Figure 6, the logical expression with the highest score 2.75 is zero. It can be interpreted that VANGL2 is 0 regardless of PRICKLE1 and PRICKLE2.

FIGS. 7A and 7B show the results of applying the network model of the present invention to the lower path having WNT 5A and WNT 9A as start points and VANGL1 and VANGL2 as end points in the WNT signal propagation path. 7a shows the lower path of the WNT signaling pathway (Nam, Chang et al., Oncogene 33 (41): 4941-4951, 2014) and FIG. 7b shows a network structure corresponding to the logical expression having the highest score Represents a combinatorial logical network structure. The remaining logical representations for each module and their scores are shown in Table 1 below.

Thus, when using the method of the present invention, the signaling path of interest can be represented using a Boolean expression, which can reflect the expression level in all samples, instead of using the representative value at the node.

100, 200: Combinational Logical Network Generation System
110 and 210:
120, and 220:
130, and 230:
235: score calculation unit
140, 240: a logic circuit generating section

Claims (10)

Generating a module having one or more input genes and one output gene from the data indicating the expression levels of two or more genes obtained from two or more samples;
Generating binary code for the module by converting the expression level of a gene belonging to a module generated in the data into a binary number and generating a binary code corresponding to the partial code A corresponding to the expression level of the input gene and the expression level of the output gene And a corresponding partial code B;
When two arbitrarily selected binary codes of the generated binary codes have the same partial code A and different partial codes B, the two binary codes are placed in different groups, and the binary codes included in each group are divided into partial codes A and a corresponding single partial code B; And
Generating a logic circuit for each of the groups, and generating logic circuits for some or all of the groups. 2. The method of claim 1, wherein said module generation step generates a module taking into account the literature, experimental measurements, transcript database, biological path database, gene set analysis tool, or a combination thereof. 2. The method of claim 1, wherein the step of converting the binary number converts the gene expression level into a binary number through a transfer function. 2. The method of claim 1, further comprising calculating scores for each group based on conditional probabilities. 5. The method of claim 4, wherein the score is calculated by:

In the above equation,

Is the gene expression binary value of the output portion of the module,

Is the binary sequence of one or more gene expression in the input portion of the module,

Is the probability of a particular output value under the condition of a particular input sequence,

Is the number of times observed in each group. 2. The method of claim 1, wherein the sample is obtained from a gastric cancer patient. A module generating unit for generating a module having one or more input genes and one output gene from the data indicating the expression levels of two or more genes obtained from two or more samples;
A binary code conversion unit for converting the expression level of a gene belonging to a module generated in the data into a binary number to generate a binary code for the module, the code comprising a partial code A corresponding to the expression level of the input gene, A partial code B corresponding to the level of the partial code;
When two arbitrarily selected binary codes of the generated binary codes have the same partial code A and different partial codes B, the two binary codes are placed in different groups, and the binary codes included in each group are divided into partial codes A and a corresponding single partial code B; And
And a logic circuit generation unit for generating logic circuits for a part or all of the groups, respectively. 8. The system of claim 7, wherein the module generating unit generates a module with consideration of literature, experimental measurements, transcript database, biological path database, gene set analysis tool, or a combination thereof. 8. The system of claim 7, wherein the binary conversion unit converts the gene expression level to a binary number through a conversion function. 8. The system of claim 7, further comprising a score calculation unit for calculating a score based on a conditional probability for each group generated.

Images