KR100431620B1 - A system for analyzing dna-chips using gene ontology, and a method thereof - Google Patents

Abstract

The present invention provides a system and method for analyzing the gene expression pattern of DNA chip or microarray experiment biologically through hierarchical structure modeling of Gene Ontology (GO). It is about. A DNA chip analysis system using GO according to the present invention comprises: means for receiving a statistical clustering result of a DNA chip test result and assigning a Gene Ontology (GO) identifier to each of the genes belonging to each cluster; Means for converting each GO identifier assigned to a gene belonging to the cluster to a GO code using a GO code file; Means for obtaining an average similarity distance and a maximum similarity distance of genes using GO codes of genes belonging to the cluster; Extracts GO terms that are optimally matched using the average similarity distance and the maximum similarity distance between the genes included in the cluster and the GO nodes in the GO tree structure according to one of the basic process and the N-step selection process. Way; And means for extracting the biological meaning of the cluster using the optimally matched GO term.

Description

System and method for analyzing DNA chips using genetic lexical classification system {A SYSTEM FOR ANALYZING DNA-CHIPS USING GENE ONTOLOGY, AND A METHOD THEREOF}

The present invention relates to a system and method for analyzing DNA chips using a gene vocabulary classification system, and more specifically, to a structure of a hierarchical structure of a gene ontology (GO). The present invention relates to a system for biologically analyzing a gene expression pattern of a DNA chip or a microarray experiment through modeling, and a method of analyzing the same.

Since the double helix structure of DNA was identified by Watson and Crick in 1954, advances in restriction enzyme discovery, hybridization techniques, and polymerase chain reaction (PCR) have led to a molecular understanding of life phenomena. Contributed greatly. However, the process of understanding the function of sequencing is emerging as the need for a holistic understanding such as the Human Genomic Project (HGP) rather than a partial understanding of life phenomena with complex regulatory functions. During this process, a DNA chip was developed. Bioinformatics and functional genomics are also being actively researched to efficiently utilize the results of HGP and DNA chips.

Biochips are largely divided into microarrays and microfluidics chips, where microarrays are arranged at regular intervals, with thousands or tens of thousands of DNA or proteins, and processed to analyze the binding patterns. Refers to the chip that can be, there is the above-described DNA chip and protein chipping, and to date it can be seen that the DNA chip is the most widely used biochip. In addition, the microfluidics chip may analyze the reaction with a biological molecule or sensor integrated in the chip while flowing a small amount of analyte.

The DNA chip is a target DNA or cDNA or oligonucleotide attached to a glass plate, nitrocellulose membrane or silicon. In other words, such a DNA chip is a micro-array of cDNA or oligonucleotide probes with known sequences on a small surface of a solid surface in a predetermined position.

These DNA chips hybridize with probes labeled with fluorescent or radioactive isotopes to identify gene expression, mutations, single nucleotide polymorphism (SNP), disease diagnosis, and high-throughput screening. ; HTS) and the like. When the sample DNA fragment to be analyzed is bound to the DNA chip, the hybridization state is formed according to the complementary degree of the nucleotide sequence on the sample DNA fragment and the probe attached to the DNA chip. By observing and analyzing this, the nucleotide sequence of a sample DNA can be measured. Using these DNA chips, the expression information of a large number of genes can be easily and quickly known, and are currently being used for new drug development and medical diagnosis.

Statistical and biological methods are used to analyze DNA chip results. The degree of expression of each gene expressed through image analysis is grouped together by clustering to show common expression patterns using statistical methods. Here, by using the known function of each gene, it gives a general meaning to the cluster and at the same time biologically confirms the reliability of the cluster.

The existing biological identification process uses a method of extracting and comparing gene functions from papers or existing biological information databases. The databases used here include basic DNA information from the National Center for Biotechnology Information (NCBI), functional category information such as the MIPS (Munich Information Center for Protein Sequences) or the Cancer Genome Anatomy Project (CGAP), or Swiss-Prot. Use protein information. However, up to now, it has been made by a researcher's manual work, and there is a problem that it is difficult to perform a systematic and automated analysis due to the diversity of biological terms.

In addition, in the existing biological information database, Swiss-Prot, which is widely used as a source of protein information, categorizes the functions of proteins well by using key words, but there is a formal correlation or vertical relationship between these key words. There is no hierarchy, which impedes automation in the biological analysis of DNA chips. In addition, specialized sector-specific group information such as CGAP (Cancer Genome Anatomy Project) has limitations that apply only in that field, and because the group itself deals with functions that are too broad, it has limitations in detail. There is a problem.

An alternative is to use the GO terminology provided by the GON Ontology Consortium. Here, the term "ontology" simply refers to a system that classifies biological terms or vocabulary. The Genetic Vocabulary Classification Consortium was established for the integration of biological terms and provides a set of commonly used unified terms used to describe the function of genes in all species, and currently consists of more than 10,000 terms. . After all, GO means that genes or keywords implied by them become individual entities and study the relationships between them. They apply to bioinformatics.

The singularity of these GO terms is that they have a tree structure of top and bottom relations between the terms, and the whole terms are divided into three large categories. That is, about 10,000 terms with three large categories are organized in a hierarchy, like a tree structure. In order to find the biological meaning of DNA chip analysis, GO can be used to determine the function of a gene, the molecular function, ii) the biological process, and the cellular component. The categories are divided into categories and a controlled vocabulary is established for each category. These categories are not mutually exclusive, but rather a division of features for describing a single gene.

An object of the present invention for solving the above problems is to analyze DNA chips using a gene lexical classification system to systematically perform biological analysis on gene expression patterns of DNA chip experiments by modeling GO hierarchy. It is to provide a system and analysis method.

In addition, another object of the present invention is to extract the function of the most common and ideal genes of the genes belonging to the cluster (cluster) generated through the statistical clustering of the experimental results of the DNA chip using the GO term and tree structure. It is to provide a method.

1 is a block diagram of a DNA chip analysis system using a gene vocabulary classification system (Gene Ontology) according to the present invention.

2 is a diagram illustrating an example of a GO tree structure according to the present invention.

3 is a diagram illustrating an example of a modified GO tree of a text structure according to the present invention.

4 is a diagram illustrating an example of conversion of the extracted GO code according to the present invention.

5 is a view for schematically explaining the principle of obtaining the similar distance using the GO according to the present invention.

FIG. 6 illustrates an example of obtaining an optimal intersection point when there are several nodes.

7 is a flowchart illustrating a method of analyzing a DNA chip using GO according to the present invention.

As a means for achieving the above object, a system for analyzing a DNA chip using the gene vocabulary classification system according to the present invention receives a statistical clustering result of the DNA chip experiment results, genes belonging to each cluster Means for assigning a Gene Ontology (GO) identifier for each; Means for converting each GO identifier assigned to a gene belonging to the cluster to a GO code using a GO code file; Means for obtaining an average similarity distance and a maximum similarity distance of genes using GO codes of genes belonging to the cluster; Extracts GO terms that are optimally matched using the average similarity distance and the maximum similarity distance between the genes included in the cluster and the GO nodes in the GO tree structure according to one of the basic process and the N-step selection process. Way; And means for extracting the biological meaning of the cluster using the optimally matched GO term, and adding visualization means for displaying the extracted optimally matched GO term and its GO code and biological meaning. It can be included as.

In addition, the visualization means is characterized by displaying a summary of the GO term and its GO code and biological meanings that are optimally matched in the form of a table, or display a graphical result in the form of a tree structure.

Further, the similarity distance is a weight of the level of the code of the optimal branch formed by the two nodes v1 and v2 when the similarity is Pd ( v1, v2), where v1 and v2 are nodes. When v1 and v2 are the same, the Pd value is defined as 0;

In addition, when G is a combination of codes of a given cluster, each optimal intersection point is obtained using the maximum similarity distance max_pd and the average similarity distance aver_pd, where G = {v1, v2, v3, v4, , vn} in max_pd and aver_pd,

max_pd (G) = max {pd (v _i , v _j )} (where 1≤i≤j≤n)

aver_pd (G) = (sum of all pd (v _i , v _j ) in set G) / _n C ₂

= 2 × (sum of all pd (v _i , v _j ) in set G) / n (n-1)

It is defined as, and it is characterized by finally selecting max_pd and aver_pd of the lowest score among possible code combinations.

In addition, the maximum similarity distance (max_pd) is used to roughly evaluate the cluster, and the higher the optimal intersection is located at a higher level, the cluster contains bad clusters that impair the general commonality of the genes belonging. Characterized in that it means that there is a high probability.

In addition, the average similarity distance aver_pd is characterized by how well aggregated GO codes within a given cluster and how frequently similar codes are observed.

In addition, the basic process is calculated using the maximum similarity distance (max_pd) and the average similarity distance (aver_pd) for all nodes in the GO tree structure, and the result of the basic process shows the approximate biological meaning of a given cluster. The N-step selection process is calculated using the maximum similarity distance and the average similarity distance for the GO node of a specific level selected on the GO tree structure.

In addition, an analysis method for analyzing a DNA chip using the gene vocabulary classification system according to the present invention, the method comprising the steps of: a) assigning a GO identifier for each cluster statistical clustering results of the DNA chip experiment results; b) converting GO identifiers assigned to genes belonging to the cluster into GO codes using GO code files; c) obtaining average similarity distance and maximum similarity distance of genes using GO codes of genes belonging to the cluster; d) Find the optimal GO term using the average similarity distance and the maximum similarity distance between the genes included in the cluster and the GO nodes in the GO tree structure according to one of the basic process and the N-step selection process. Extracting; And e) extracting the biological meaning of the cluster using the optimally matched GO term, further comprising displaying the extracted optimally matched GO term and its GO code and biological meaning. can do.

Hereinafter, a preferred embodiment of a system and method for analyzing a DNA chip using GO according to the present invention will be described with reference to the accompanying drawings.

1 is a block diagram of a DNA chip analysis system using GO according to the present invention, an input unit 110 for inputting a statistical clustering result of the DNA chip test result; A GO identifier allocator 130 for allocating a GO identifier for each gene belonging to each cluster with respect to the input clustering result using a GO identifier index file 120; A GO identifier / GO code conversion unit 140 for converting each GO identifier assigned to the gene into a GO code using a GO code file; An optimal intersection point extracting unit 220 for extracting an optimal intersection point by selecting a predetermined process according to the similarity distance algorithm 210 for the GO code and designating a required variable; And a biological meaning extraction unit 230 for extracting the biological meaning for each extracted optimal intersection point. In addition, each of the genes may further include a display 310 for displaying the optimal intersection, the GO code, and the biological meaning respectively extracted.

The present invention uses the GO term and tree structure to extract the most common and ideal gene function of the genes belonging to the cluster generated through the statistical clustering of the experimental results of the DNA chip.

For this purpose, the correct GO term is assigned to each gene, the optimal intersection point is extracted by efficiently using the structure of the GO hierarchical tree structure, and the optimal intersection point extraction result is displayed efficiently.

2 is a diagram illustrating an example of a GO structure according to the present invention, where the top level is the GO layer, the second layer is the molecular function, the biological process, and the cellular component described above. Hierarchical level), a tree is formed at lower levels of levels 3, 4, and 5, respectively. FIG. 3 is a diagram illustrating an example of a modified GO tree of a text structure according to the present invention. In practice, GO is not a tree structure and takes the form of a mathematical graph called an acyclic diagraph without a circuit. A similar algorithm used in the example will convert the GO structure into a GO tree structure. 3 shows an example of a slight modification of the GO tree of this text structure. In addition, Figure 4 is a view showing a conversion example of the GO code extracted in accordance with the present invention, illustrating the output of the result converted by the GO code conversion unit 140. In Fig. 4, the numbers described in the “Output” column are GO codes, and as shown in Fig. 4, the GO codes are composed of 15 number combinations. Since the GO term is a letter, it cannot determine how close it is to other GO terms in the GO tree structure. Therefore, in the present invention, the GO term is to be converted into a GO code which is a preset number combination. The GO code is 15 digits because the level of the GO hierarchy is 15 levels, where the first digit of the GO code represents the value at level 1, and the second digit of the GO code represents the value at level 2. Referring to the example of converting a GO node to a GO code as follows. The GO node of ID 400 belongs to level 1 and is the first node of level 1. At this time, the GO node of the identification code 400 is converted into a GO code of "100000000000000". Since the GO node of ID 400 is the first GO node of level 1, the value from the second to 15th digits is 0, and the value of the first digit is 1. The GO node of ID 402 is the second level. A child node of the GO node having an identification code of 400. At this time, the GO node of identification code 402 is converted into a GO code of "110000000000000". Since the GO node of the identification code 402 belongs to level 2, the value from 3 to 15 digits is 0. In addition, since it is the child node of the GO node corresponding to the identification code 400, the value of the first digit uses the value of the parent node as it is. In addition, since the GO node of the identification code 402 is the first node among the lower nodes of the node of the identification code 400 belonging to the level 2, the value of the second digit is 1. In this way, the GO node of the identification code 404 is " 120000000000000 " May be converted to a GO code. The GO node of identification 410 is the third level, the child node of the node of identification 402, and the second node of the child nodes of identification 402. . Thus, the GO node of identification 410 may be converted to a GO code of "112000000000000". In the same principle, the GO node with identifier 412 is converted to a GO code of "121000000000000". As above, the GO node is converted to a GO code, so that the GO code is the level to which the GO node belongs and the parent of the GO node. Contains information about the node.

In the present invention, an optimal branch refers to a node located at a lowermost level among nodes including the largest number of genes in a tree structure below, and each of the genes included therein It is a broad term that can represent all functions. In the present invention, the pseudo-distance between two nodes in the GO tree structure is obtained, and the process of finding the optimal intersection point is a preliminary step for calculating the similarity distance. The following description will be made with reference to the following. An upper node including both the node of the identification code 408 and the node of the identification code 310 includes a node of the identification code 402 and a node of the identification code 400. As described above, the lowest node among the upper nodes including both nodes is determined as an optimal intersection point, and since the node of the double identification code 402 is the lowest node, the node of the identification code 408 and the node of the identification code 410 are determined. The best intersection is the node of identification 402. With the GO code, the best intersection can be found relatively easily. In FIG. 2, the GO code of the node 408 is "111000000000000" and the GO code of the node 410 is "112000000000000". Since the two GO codes are the same to the second digit, the best intersection exists at the second level, and the first node (the second digit) of the child nodes of the first node of the first level (since the first digit is 1) It can be seen that 1) is the optimal intersection point. An example of finding the optimal intersection of several nodes is shown in FIG. The method of obtaining the similar distance using the optimum intersection is explained later in detail.

Assigning the correct GO term for each gene assigns the GO term for each gene through text mining of several biological databases. Information at the DNA or protein level, such as UniGene, LocusLink, Swiss-Prot, and MGI, is used in combination with direct identifier (ID) comparison and sequence similarity search, and the genetic identifier (ID) conversion provided by each database in the GO Consortium Use the files to assign GO terms for each gene.

Here, UniGene provides NCBI gene information at the DNA level provided by the National Center for Biotechnology Information (NCBI), and LocusLink is the NCBI's Reference Sequence Project. Swiss-Prot provides protein-level information at the Swiss Institute of Bioinformatic, and MGI provides mouse genomic information.

In the present invention, the optimal branch is obtained using the GO tree structure efficiently, and the similar distance is obtained by using the optimal intersection, and then the GO term that can represent a given cluster is found using the same. That is, the GO term that can represent the cluster is found through the similarity between the GO identifiers assigned to the genes of the cluster and the nodes of the GO tree structure. To solve this problem, we first coded each node in the GO tree structure. These codes consisted of fifteen number combinations, as described above, with each number representing positional information up to the top route. In addition, since unique codes are assigned to each node, even if the same terms are located in various places in the tree structure, they are distinguished from each other. FIG. 5 schematically illustrates a principle of obtaining similar distances using GO according to the present invention. It is a figure for explaining.

As shown in FIG. 5, an appropriate weight is assigned to each level of the GO tree structure to obtain a similar distance using these GO codes.

Pd ( v1, v2) is the weight of the level of the code of the optimal branch formed by the two nodes v1, v2. If v1 and v2 are the same, the Pd value is defined as 0. In other words,

pd (v1, v2) = weight of the best intersection code between v1 and v2 (if v1 ≠ v2)

pd (v1, v2) = 0 (if v1 = v2)

Finally, the optimal intersection point of the node of identification 500 and the node of identification 502 in FIG. 5 is the node of identification 504, and the node of identification 504 exists at the second level and is assigned to the second level. Weight is 140. Accordingly, the similarity distance between the node 500 and the node 502 is 140.

Next, when a combination of codes of a given cluster is G, a maximum similarity Pd (max_Pd) and an average similarity distance Pd (aver_Pd) are obtained.

In G = {v1, v2, v3, v4,, vn}, max_Pd and aver_Pd are defined as below, and the lowest score max_pd and aver_pd among the possible code combinations are as follows.

max_Pd (G) = max {pd (v _i , v _j )} with 1≤i≤ j ≤n

aver_Pd (G) = (sum of all pd (v _i , v _j ) in set G) / _n C ₂

= 2 * (sum of all pd (v _i , v _j ) in set G) / n (n-1)

Where max_pd is a measure that can be used to roughly evaluate the cluster. If the optimal intersection is located at a higher level, the cluster is more likely to contain bad clusters that compromise the general commonality of its genes.

The Aver_pd may indicate how well aggregated GO codes are in a given cluster and how frequently similar codes are observed.

On the other hand, the process of finding GO terms that can represent a cluster can be largely divided into a basic process (N-level selective process).

The basic process is calculated using max_pd and aver_pd with the genes of the cluster for all nodes in the GO tree structure. As a result of this basic process, we can find the GO term that best matches the cluster. This gives us the approximate biological meaning of the cluster and the given cluster.

In addition, the N-step selection process does not calculate max_pd and aver_pd with the genes of the cluster for all nodes in the GO tree structure, but allows the user to designate a certain limit point. Here, the N-step selection process is calculated by specifying the level of the GO tree structure to calculate max_pd and aver_pd in advance, and it is easy to observe the GO term that is optimally matched at a specific step. Easily infer the biological meaning of In particular, the N-step selection process may indicate the next highest group of ranks in addition to the best code combination. This allows a gene to contain all of the diversity that may be involved in more than one function.

FIG. 7 is a flowchart illustrating a method for analyzing a DNA chip using GO according to the present invention. In the method for analyzing a DNA chip, the method receives a statistical clustering result of the DNA chip test result (S10). Receiving a statistical clustering result of the DNA chip test result (S10) and allocating a Gene Ontology (GO) identifier for each gene belonging to each cluster (S20); Converting each GO identifier assigned to each gene into a GO code using a GO code file (S30); Selecting a predetermined process from the basic process (S41) according to the similarity algorithm (S40) and the N-step selection process (S42) with respect to the GO code, and extracting a GO term that is optimally matched by designating a required variable (S50); Extracting the biological meaning using the optimally matched GO term (S60); And displaying an optimally matched GO term and its GO code (S70).

Referring to Figure 7, when explaining the overall method of biological analysis of the gene expression pattern of the DNA chip using the GO structure according to the present invention.

First, in the result of statistical clustering of gene expression patterns, a process of allocating a GO identifier and a code for each gene belonging to each cluster is performed.

Specifically, when the clustering result is input (S10), the GO ID is allocated to the genes in the cluster by using a file in which the GO IDs are pre-assigned through mining of several databases for each gene. (S20). Next, by using the GO code file that encodes the entire GO tree structure, the GO IDs assigned in the cluster-specific genes are converted into GO codes (S30).

Next, using a similar algorithm, an appropriate process is selected (S40) among the basic process (S41) and the N-step selection process (S42), and the required variables are designated. Then, GO terms that are optimally matched are extracted using Pd for each process (S50), and biological meanings are extracted accordingly.

Next, the optimal matching GO terms and GO codes extracted for each cluster are displayed, and the GO codes and the optimal matching GO terms for each gene in a tabular form and summary information of the biological meaning, or graphic results in the form of a tree structure are displayed. can do.

Meanwhile, the similar algorithm may be similarly applied to protein chips, which are other biochips, to analyze protein chips instead of the DNA chips shown in FIGS. 1 and 7, and to use a similar distance algorithm for proteins instead of genes. The same can be applied.

Although this invention was demonstrated concretely by the said Example, this invention is not restrict | limited by this, A deformation | transformation and improvement are possible within the range of common knowledge of a person skilled in the art.

According to the present invention, a systematic automated biological analysis of gene expression patterns of DNA chip experiments can be carried out through modeling of the GO hierarchy, and also statistical analysis of experimental results of DNA chips using GO terminology and tree structure The most common and ideal gene function of the genes belonging to the cluster generated through clustering can be extracted.

Claims (18)

In a system for analyzing a DNA chip, Means for receiving a statistical clustering result of the DNA chip test result and assigning a Gene Ontology (GO) identifier to each gene belonging to each cluster; Means for converting each GO identifier assigned to a gene belonging to the cluster to a GO code using a GO code file; Means for obtaining an average similarity distance and a maximum similarity distance of genes using GO codes of genes belonging to the cluster; And Extracts GO terms that are optimally matched using the average similarity distance and the maximum similarity distance between the genes included in the cluster and the GO nodes in the GO tree structure according to one of the basic process and the N-step selection process. Way; And Means for extracting the biological meaning of the cluster using the optimally matched GO term DNA chip analysis system using a gene vocabulary classification system (GO) comprising a. The method of claim 1, And a DNA vocabulary classification system further comprising visualization means for displaying the extracted optimally matching GO term and its GO code and biological meaning. The method of claim 2, The visualization means may display the optimally matched GO term, its GO code, and summary information of biological meanings in a tabular form, or display a graphical result in the form of a tree structure. Chip analysis system. delete The method of claim 1, The similarity distance is, If the similarity is Pd ( v1, v2), where v1 and v2 are nodes, the weight of the level of the code of the optimal branch formed by the two nodes v1 and v2, and v1 and v2 are If equal, the Pd value is defined as 0; In addition, when G is a combination of codes of a given cluster, each optimal intersection point is obtained using the maximum similarity distance max_pd and the average similarity distance aver_pd, where G = {v1, v2, v3, v4, , vn} in max_pd and aver_pd, max_pd (G) = max {pd (v _i , v _j )} (where 1≤i≤j≤n) aver_pd (G) = (sum of all pd (v _i , v _j ) in set G) / _n C ₂ = 2 × (sum of all pd (v _i , v _j ) in set G) / n (n-1) A DNA chip analysis system using a gene vocabulary classification system, characterized in that finally selecting max_pd and aver_pd of the lowest score among possible code combinations. The method of claim 5, The maximum similarity distance max_pd is used to roughly evaluate the cluster, and the higher the optimal intersection is at the higher level, the more likely that cluster contains bad clusters that compromise the general commonality of the genes involved. DNA chip analysis system using a gene vocabulary classification system, characterized in that high. The method of claim 5, Wherein said average similarity distance (aver_pd) indicates how well aggregated GO codes are in a given cluster and how frequently similar codes are observed. delete The method of claim 1, The basic process is calculated using the maximum similarity distance (max_pd) and the average similarity distance (aver_pd) for all nodes in the GO tree structure, and the result of the basic process shows the approximate biological meaning of a given cluster. DNA chip analysis system using a gene vocabulary classification system. The method of claim 1, The N-step selection process is a DNA chip analysis system using a gene lexical classification system characterized in that the calculation using the maximum similarity distance and the average similarity distance for the GO node of a specific level selected on the GO tree structure. delete delete In the method for analyzing a DNA chip, a) assigning a GO identifier to each cluster of statistical clustering results of the DNA chip experiment results; b) converting GO identifiers assigned to genes belonging to the cluster into GO codes using GO code files; c) obtaining average similarity distance and maximum similarity distance of genes using GO codes of genes belonging to the cluster; d) Find the optimal GO term using the average similarity distance and the maximum similarity distance between the genes included in the cluster and the GO nodes in the GO tree structure according to one of the basic process and the N-step selection process. Extracting; And e) extracting the biological meaning of the cluster using the optimally matched GO term DNA chip analysis method using a gene vocabulary classification system comprising a. The method of claim 13, The method of claim 1, further comprising displaying the extracted optimally matching GO term, its GO code, and biological meaning. The method of claim 13, The similarity distance is, If the similarity is Pd ( v1, v2), where v1 and v2 are nodes, the weight of the level of the code of the optimal branch formed by the two nodes v1 and v2, and v1 and v2 are If equal, the Pd value is defined as 0; In addition, when G is a combination of codes of a given cluster, each optimal intersection point is obtained using the maximum similarity distance max_pd and the average similarity distance aver_pd, where G = {v1, v2, v3, v4, , vn} in max_pd and aver_pd, max_pd (G) = max {pd (v _i , v _j )} (where 1≤i≤j≤n) aver_pd (G) = (sum of all pd (v _i , v _j ) in set G) / _n C ₂ = 2 × (sum of all pd (v _i , v _j ) in set G) / n (n-1) A method of analyzing a DNA chip using a gene lexical classification system, characterized in that finally selecting max_pd and aver_pd of the lowest score among possible code combinations. The method of claim 15, The maximum similarity distance max_pd is used to roughly evaluate the cluster, and the higher the optimal intersection is at the higher level, the more likely that cluster contains bad clusters that compromise the general commonality of the genes involved. DNA chip analysis method using a gene vocabulary classification system, characterized in that high. The method of claim 15, The average similarity distance (aver_pd) represents how well the aggregated GO codes in a given cluster, and how frequently the similar codes are observed, DNA chip analysis method using a gene lexical classification system delete