KR20110054926A - System and method comprising algorithm for mode-of-action of microarray experimental data, experiment/treatment condition-specific network generation and experiment/treatment condition relation interpretation using biological network analysis, and recording media having program therefor - Google Patents

Abstract

PURPOSE: A system and method comprising algorithm for mode-of-action of microarray experimental data, experiment/treatment condition-specific network generation and experiment/treatment condition relation interpretation using biological network analysis, and recording media having program therefor are provided to grasp an unknown gene and protein. CONSTITUTION: A gene network learning unit predicts a gene network. A protein network significance inspector inspects the significance of the protein network. An activation path significance inspector inspects a significance of a searched activation signal/metabolism transfer path. An experiment condition relation reading unit interprets the relationship of the experiment/processing condition. An integrated output unit outputs a biologically integrated network through a read experiment condition relation.

Description

Systems and methods including algorithms for the mechanism of action of microarray experimental data using biological network analysis, experiment / process condition-specific network generation, and analysis of experiment / process condition relationships, and recording media having a program for performing the method { System and method comprising algorithm for mode-of-action of microarray experimental data, experiment / treatment condition-specific network generation and experiment / treatment condition relation interpretation using biological network analysis, and recording media having program therefor}

The present invention utilizes microarray experimental data to integrate genetic networks, experiment / treatment condition-specific protein networks, and experiment / treatment condition relationships. A system and method including an algorithm for analyzing a network and a recording medium having recorded thereon a computer readable program for performing the method.

It is clear that genes expressed in diseased cells are different from genes expressed in normal cells. However, since genes function more often than many other genes interact with each other than occur with a single gene change, and because related genes that interact with target genes in a disease state are simultaneously expressed, expression of these genes It is quite difficult to measure how it changes.

In 1995, a microarray technique was developed to detect forms in which a large number of genes were expressed (Schena et al. 1995. Science 270: 467-470). As the microarray technology develops, it is possible to identify in one experiment how all genes of higher organisms such as humans and mice change expression under specific experimental conditions.

Genes often function by interacting with multiple genes, which are equally true for proteins, signaling pathways, and metabolism. These complex elements work together to create a biological network, which is called a gene network, protein-protein interactions, and signal / metabolism pathways. / metabolic pathway).

A genetic network technique that teaches and teaches how genes work with each other using microarray technology was introduced by Friedman (2000. J. Comput. Biol. 7 (3-4): 601-20). This technique reveals genes that are highly related to each other by examining the expression profile of genes over time or with experiments, and creates a network of genes by linking the two genes. These networks can be used to explore the relationships between genes and interpret their function.

A protein is a molecule that works for a gene to function in life. Like a gene, two or more polypeptides are joined by an attraction force between amino acids, rather than a protein functioning alone. As a protein complex (e.g., hemoglobin: composed of four subunits, one pair of α-globin and one pair of β-globin) More often than not. The information on the connectivity between proteins is called protein-protein interaction, and this protein-protein interaction is an essential unit for understanding or interpreting the function of genes as a unit that exerts various functions for living.

Signal / metabolism pathways are collections of genes and proteins that perform specific functions, such as cell metabolism, migration, proliferation, survival, differentiation or movement of the optic nerve, and the gene networks, protein-protein interactions, or gene-proteins described above. Contains interactions. These signal / metabolism pathways are models of the binding network of genes and proteins, including transcription factor proteins that regulate gene expression.

In order to confirm the change in gene expression in response to a drug administered to a disease or disease state, a conventional state using a method such as fold-change, t-test, or Signal Analysis Analysis of Microarrays (SAM), etc. The microarray experiment using a sample of and a sample of a diseased state or a drug-treated state in a diseased state were compared with each other, and a technique of calculating the amount of gene expression change between the microarray experiments was used. However, this method can be misleading due to the error of the expression value of the gene every time the experiment is performed, and even if the change rate of the disease state and the normal state is high, whether the gene is directly related to the disease is indirectly affected. It is not possible to determine whether it is a gene or an unaffected gene. However, if such genes can be drawn in the form of a network, the effect of the gene on other genes in experimental processing such as the disease or drug target investigated can be identified.

DETAILED DESCRIPTION OF THE INVENTION The present invention is derived from the above needs and explores the mechanism of action of genes in response to microarray experimental conditions in experimental treatments such as drugs administered to a disease or disease state, and biological networks such as gene networks or protein networks. To determine the causality of the expression gene by generating the, and to analyze the function of the expression gene by reading the relationship between the experiment / treatment.

In order to solve the above problems, the present invention uses a microarray experimental data, gene network, experiment / treatment condition-specific protein network (experiment / treatment condition-specific protein network), experiment / treatment conditions Systems and methods including algorithms for analyzing biologically integrated networks for relationship identification are provided.

The present invention also provides a recording medium having recorded thereon a computer readable program for performing the method.

Identifying target genes for experiments / processing, such as drugs administered to a disease or disease state, using a biological integration network from microarray data can identify genes / proteins that are not known by conventional methods, and identify gene associations. It is also possible to find gene candidates that respond to drugs and to identify the gene / protein quality easily.

In order to achieve the object of the present invention,

Microarray data from a variety of conditions, such as knockout, drug, RNAi, and overexpression in a single species, and microarrays measured to identify mechanisms of action, such as drugs administered in disease or disease states. An input unit for inputting experimental data;

A gene network learning unit for predicting a gene network from input microarray experiment data;

Using the Mode-of-action by Network Identification (MNI) technique or the Bootstrapping Regression Model to explore the mechanism of action that provides predictive ranking information of genes for action mechanisms operating under experimental conditions. A search unit composed of an algorithm;

A gene network output unit for outputting a gene network by interlocking the learned gene network with a searched action mechanism;

A search unit configured of an algorithm for searching for protein-protein interaction information by using genes or searched mechanisms significantly expressed in input microarray experiment data;

A protein network significance tester that examines the protein-protein interaction information using GNEA (Geneset Network Enrichment Analysis) to examine the significance of the protein extracted from the microarray experimental data or the protein network of the searched mechanism;

A protein network output unit for outputting an experiment / condition specific protein network according to the examined significance;

Search for activation signal / metabolism transmission pathways to determine whether genes expressed significantly in the input microarray experiment data or genes / proteins included in searched experimental / processing conditions specific protein networks exist in signal / metabolism pathways A search unit configured of an algorithm;

The list of genes included in the detected active signaling / metabolic pathway is compressed by Singular Value Decomposition (SVD) to determine whether the detected activation signal / metabolic pathway is actually significant. An activation path significance checker to check;

An activation signal / metabolism transmission path output unit displaying a significantly identified activation signal / metabolism transmission path;

Experimental condition relational readout that analyzes the relationship of the condition of experiment / process by linking genes / proteins that are significantly expressed genes or searched experimental / process condition specific protein network with activation signal / metabolism transmission pathway ; And

An integrated output unit comprising an algorithm for outputting a biological integrated network by giving the relationship of the experimental conditions read to the gene network output unit, the protein network output unit, and the activation signal / metabolism transfer path output unit

Predicting / analyzing the gene network, experiment / treatment condition specific protein network, activation signal / metabolism transmission pathway using the expression value of the gene measured through the microarray experiment including the or reading the relationship between the experiment / treatment condition It provides a system characterized in that the output to the biological integration network.

In the system according to an embodiment of the present invention, the algorithm of the search unit uses the bootstrapping regression model method by Equation 4 to learn the genetic network model from the input microarray experimental data and predict the action mechanism It could be an algorithm that explores the mechanism of action and outputs it to the genetic network:

는 유전자 i에 대해 붓스트랩 표본으로 구한 회귀계수,

는 주어진 관측치 쌍에서 복원랜덤추출한 붓스트랩 표본

을 적합시킨 회귀계수를 나타낸다).Where B represents the number of iterations of the bootstrapping process,

Is the regression coefficient obtained from the bootstrap sample for gene i ,

Is a random sample of the bootstrap extracted from a given pair of observations.

To show the regression coefficients fitted).

In a system according to an embodiment of the present invention, the algorithm for retrieving protein-protein interaction information includes a gene or protein consisting of a significant gene of input microarray experimental data and a mechanism searched from a learned gene network. It may be an algorithm that searches a list or only a mechanism of action with an experimental / processing condition specific protein network or checks for significance.

In the system according to an embodiment of the present invention, the algorithm for searching the activation signal / metabolism transmission pathway is tested for significant genes of the input microarray experiment data and the mechanism of action and significance searched from the learned gene network. Experimental / Processing Conditions An algorithm that searches for activation signal / metabolism pathways or reads the relationship of experimental / processing conditions using a list of genes or proteins consisting of two or more elements from a list of genes / proteins included in a specific protein network Can be.

The present invention also provides

a) Microarray data obtained from treatment of various conditions such as knockout, drug, RNAi, and overexpression in a single species and measured to identify mechanisms of action such as drugs administered in disease or disease states. Inputting microarray experimental data;

b) generating a genetic network using a general regression model or bootstrapping regression model using inputted microarray data and searching for a mechanism of action;

c) predicting the experimental / processing condition specific protein network using the searched mechanism of action;

d) searching for activation signal / metabolism transmission pathways using the list of genes / proteins involved in the searched mechanisms of action and experimental / processing specific protein networks; And

e) reading the relationship of the experiment / treatment conditions using the retrieved activation signal / metabolic transfer pathway;

It provides a method for building a biological integration network using the relationship between the outputted genetic network, experimental / processing conditions specific protein network, activation signal / metabolism transmission path information and the read experimental / processing conditions comprising a.

In the method according to an embodiment of the present invention, the step c) is a microarray experimental data and a disease or disease state subjected to treatment of various conditions of a single species such as knockout, drug, RNAi, and overexpression. The protein network can be derived using the identified genetic network and mechanism information using the microarray experimental data measured to identify the mechanism of action such as the drug administered in.

In the method according to an embodiment of the present invention, the steps d) and e) are microarray experimental data and disease which are subjected to the treatment of various conditions of a single species such as knockout, drug, RNAi, and overexpression. Activation signal / substrate metabolism using information from the genetic network and mechanism of action identified using microarray data measured to identify mechanisms of action, such as drugs administered in disease states, or from protein networks derived from them. You can derive the route of delivery or read the relationship of the experimental / treatment conditions.

The invention also provides a recording medium having recorded thereon a computer readable program for carrying out the method for building the biologically integrated network of the present invention. Specifically, biological data for the relationship between gene networks, experiment / treatment condition-specific protein networks, and experiment / treatment conditions can be identified using microarray data. Provided is a recording medium having recorded thereon a computer readable program for performing a method for analyzing an integrated network.

Computer-readable recording medium refers to any recording medium that can be read directly and accessed by a computer. Such recording media include magnetic recording media such as floppy disks, hard disks, and magnetic tapes, optical recording media such as CD-ROMs, CD-Rs, CDs, RWs, DVD-ROMs, DVD-RAMs, DVD-RWs, RAMs and ROMs. Electrical recording media such as and mixtures of these categories (for example, magnetic / optical recording media such as MO), but are not limited to these.

The selection of a device or apparatus for recording or inputting the above-described recording medium or a device or apparatus for reading information in the recording medium is based on the type of recording medium and the access method. Various data processor programs, software, comparators, and formats are also used to record a program for performing the method of the present invention on the medium. The information can be represented, for example, in the form of a binary file, a text file or an ASCII file formatted with commercially available software.

With reference to the accompanying drawings will be described in detail a preferred embodiment of the present invention.

Figure 1 shows the biological network for the relationship between gene network, experiment / treatment condition-specific protein network, experiment / treatment condition using microarray experimental data. A schematic diagram of a system for analyzing an integrated network is shown.

The system of the present invention comprises the input unit described above; Database; Learning unit; Search unit; Search unit; Inspection unit; An output unit; It includes a reading unit.

The input unit performs a function of inputting microarray experimental data through a treatment such as a disease or a drug. 4 shows an input unit screen. Enter the gene expression values of the microarray experimental data, which is an essential element in the input form.

In the database, the protein-protein interaction database includes experimental information of the protein-protein interaction information or includes prediction information through a text mining method. The signal / metabolism transmission pathway database contains curate information through literature information.

The learning unit functions to learn a model of a gene network using a general regression model or a bootstrapping regression model for predicting a gene network from input microarray experiment data.

The search unit functions to search the constructed database.

The searcher functions to search for a mechanism of action from the learned genetic network.

5 shows setting parameters used in the learning unit, the searching unit, and the searching unit. According to the parameters set, the learning unit, the searching unit, and the searching unit output the genetic network, the mechanism of action, the experiment / process condition specific protein network, the activation signal / metabolism transmission pathway, or the experiment / process condition according to the methodology described above. Read the relationship.

The inspection unit checks whether the data retrieved through the search unit is statistically significant. The level of testing depends on the p-value, using 70% as the empirical significance level or 0.05 as the p-value.

The reading unit functions to read the relationship between the experiment / process conditions using the output activation signal / metabolism transfer path. The reading is due to the relationship between the signal / metabolism transmission pathways previously output, which can be represented as indicated in FIG. 6.

The integrated output unit combines the results of all the above output units into one. This may be represented as shown in FIG.

In order to generate training data, microarray data obtained by processing various conditions of the same species as that of the species of organism to be studied to obtain the mechanism of action of the drug administered to the disease or disease state were obtained and processed for the same gene. An average value of the expressions of repeated microarray experiments for various conditions is averaged to generate an N × M data matrix in a horizontal arrangement as shown in FIG. 2. In this case, the number of genes is divided into N and the number of independent microarray experiments is defined as M. In the repeated microarray experiments, the standard deviation is calculated to generate data matrices of the same type. However, the format of the data matrix for the mean expression value and the data matrix for the standard deviation should be the same.

In order to generate test data, repeated experiments of microarray experiments to determine the mechanism of action in a disease or drug are averaged to generate an N x M data matrix. The standard deviation is also calculated to produce a data matrix of the same type. N is the number of genes and M is the number of microarray experiments as in the learning data. The order of genes in the study and test data should be identical (Xing et al. 2006. Nature Protocol 1 (6): 2551-4).

The regression model is used to calculate the influence on the mechanism of action in the disease or drug using the above data matrix. Influence function f _i , assuming that the factor influencing gene i is the vector j The regression model is shown in Equation 1.

Each symbol in Regression Model 1 above is as follows. As the number of genes on the measured microarray N, and the impact is concentrated in the gene i y _i, synthesis rate (synthesis rate) influence the external influence on the network in the u _i, gene i on the gene j in accordance with the gene i If the parameter expressing n _ij and the expression reduction rate of gene i is d _i , a model of the entire gene network can be calculated (di Bernardo et al. 2005. Nature Biotechnology 23 (3): 377-). 83).

Applying the above equation to the data matrix obtained from the microarray experiment and assuming a steady-state state to predict the parameters, f _i ( y ₁ , ..., y _N , u _i ) becomes 0. Then, Equation 1 above can be transformed into Equation 2 below.

Each symbol of Formula 2 is as follows.

It performs logarithmic conversion (logarithm formation) in the x _j and p _i to fit in the formula 2 above, x _j in the same manner as in microarray observations.

In Equation 2 above, y _jb and u _jb are denominators attached to the microarray data input from the comparison (control group compared with the control group). Is the value of. In order to use the formula (2), assuming that the reaction in the particular experimental conditions, only a small number of genes in the microarray to be predicted value of a _ij and p _i data doemeuro most the p _i 0, and the p _i are assumed to be zero predict a _ij When a _ij is predicted, p _i is predicted using the predicted a _ij value and the microarray experimental data matrix. Then we predict a _ij again using the predicted p _i , and then predict p _i again. This process is repeated to find the optimal a _ij and p _i . Genetic network learning unit generates a genetic network through this process. If there is a large amount of microarray experimental data dealing with various conditions in a single species input for learning, singular value decomposition should be performed to speed up the learning. In this case, the number of singular values included through the singular value decomposition should be maintained at 70%, which is an empirically significant level of the total experiment.

Using Equations 1 and 2 above, the list of extracted genes is sorted in descending order of external influence to explore the mechanism of action. Descending order is based on z-scores or modified z-scores. Z-scores are calculated as follows:

Genes output by sorting in descending order by z-scores or modified z-scores appear in order from the most influential mechanism of action in experimental / processed microarray data such as disease or drug. Remove the rest from the list, except for genes that are most influenced by the sorted z-score or modified z-score. This is to find a mechanism of action that is expressed only in experimental / treatment conditions. Apply these results back to Equation 2 repeatedly until you get the best results. The number of repetitions is dependent on the data, so it is necessary to determine the number of repetitions in advance by testing the study data using known drug targets until the optimal results are obtained.

The part for searching the mechanism of action may use the Bootstrapping Regression Model without using the general regression model described above. If you use the bootstrapping regression model, you must use the bootstrapping regression model, not the general regression model, in creating the network model.

으로부터 복원랜덤추출한 확률표본을 붓스트랩 표본

이라 할 때,

으로부터 복원추출로 하나의 붓스트랩표본

을 구성한다.Step 1; Given pairs of observations

Restore random random sample from bootstrap sample

When I say

One Bootstrap Sample from Restoration Extraction from

Configure

, j=1,...N 라 하면, 주어진 관측치 쌍에서 복원랜덤추출한 붓스트랩 표본

을 식 2에 적합시키고, 회귀계수

를 계산한다. 단계 1과 단계 2의 과정을 독립적으로 B회 반복한다. 유전자 i에 대해 B회의 붓스트랩 반복으로 계산한 회귀계수

, b=1,...,B 의 평균을 계산하여 하기의 식 4와 같이 추정하고자 하는 모형계수

를 구한다.Step 2; Regression coefficients obtained from bootstrap samples for gene i

, j = 1, ... N , the bootstrap sample reconstructed randomly from a given pair of observations

Is fitted to Equation 2, and the regression coefficient

Calculate Repeat steps 1 and 2 independently of B. Regression coefficient computed from B bootstrap repetitions for gene i

The model coefficient to be estimated by calculating the average of, b = 1, ..., B

.

Experimental / Processing Conditions Search for significant genes from microarray experimental data to search for specific protein networks or map protein rank lists extracted from the above functional mechanism analysis to a protein-protein interaction database. Retrieve interaction information.

Using the retrieved protein-protein interaction information, we apply the GNEA method proposed by Liu et al. (2007, PLoS Genet. 3 (6): e96) to apply a high-value protein-protein interaction network (HSN). Extract the interaction network). The microarray experimental data to be mapped is an experiment / treatment condition-specific protein-protein interaction using the same type of test data as shown in FIG. 3 to understand the mechanism of action of the disease or drug. It should be information. In addition, the mapping is performed by the rank of each gene changed in the experimental / processing conditions.

Experimental / Processing Conditions A list of genes needed to confirm that specific protein-protein interaction information has been correctly extracted is obtained from signaling / metabolic pathway information relevant to microarray experimental conditions.

The significance of the protein-protein interaction network extracted by mapping the microarray experiment information and the protein-protein interaction information is determined using Fisher's Exact Test. Fischer's accuracy verification is performed using the following method.

If a gene predicted to act as a mechanism of action in a disease is contained within a high value protein-protein interaction network a;

If a gene predicted to act as a mechanism of action in a disease is included outside of a high value protein-protein interaction network b;

If a gene predicted not to function as a mechanism of action in a disease is contained within a high value protein-protein interaction network c;

If a gene predicted not to function as a mechanism of action in a disease is included outside of a high value protein-protein interaction network d;

And a, b, c and d are the number of genes.

Equation 5 above is used to determine whether the extracted condition-specific protein-protein interaction information is significantly extracted. The condition-specific protein-protein interaction information extracted significantly is called high-scoring protein-protein interaction network (HSN).

Map genes / signals from signal / metabolism pathway databases to genes / proteins included in experimental / treatment condition specific networks to signal / metabolism pathway information to determine conditional relationships for experiments / processes. . To determine the association between signal / metabolism pathways, we used Hu et al. (2008, Pac. Symp. Biocomput. 255-66) methodology to examine disease or drug-treated data that is relevant to the conditions of the test data. The data matrix is extracted from the training data, and the data matrix as shown in FIG. 2 is generated in the order of the information of metabolic / signaling pathway. For example, if the test data corresponds to type 2 diabetes, microarray data of related experiments such as insulin and obesity related to type 2 diabetes are extracted. We also extract a list of genes for the signal / metabolism transmission pathways associated with the extracted microarray experiments. The microarray data is filtered through a gene list extracted from a signal / metabolism transmission pathway, and in the case of the left side of FIG. 2, only a gene whose expression is included in the gene list extracted from the signal / metabolism transmission pathway is extracted, On the right side of 2, only a list of genes obtained through the above analysis, such as a disease or drug response target included in the gene list extracted from the signal / metabolism transmission pathway, is extracted.

The singular value decomposition is performed using the extracted data matrix A and decomposed into an eigenarray matrix U , a matrix Σ consisting of singular values and an eigengene matrix V ^T.

In this case, we need to determine the k ' singular values in the ∑ matrix. In this case, we choose the (70 / Z )% that we have used empirically. Z is the number of experiments in the extracted matrix A.

The pathway activity level of the signal / metabolism transmission pathway is determined in the microarray experiment extracted above (Tomfohr et al. 2005. BMC Bioinformatics 6: 225). It is possible to determine the experiments in which the genes involved in the signal / metabolism transmission pathways were significant. To determine the activity of the signal / metabolism pathway ( l _j ) for k ' significant genes in each experiment, use Equation 7 below.

j represents a microarray experiment;

VT is the eigengene matrix V ^T calculated from singular value decomposition;

k ' is the number of singular values determined in the Σ matrix.

Calculating the activity of signal / metabolism transmission pathways generated by singular value decomposition identifies the signal / metabolism transmission pathways that are significant from the microarray data. For this purpose, Signal Analysis of Microarrays (SAM) is used using the activity of signal / metabolism transfer pathways (Tusher et al. 2001. Proc Natl. Acad. Sci. USA 98 (9): 5116-21).

In order to confirm the connectivity of the signal / metabolism transmission pathways, the relationship between the signal / metabolism transmission pathways was calculated using Spearman's rank correlation for the activity of the signal / metabolism transmission pathways of microarray experiments. .

In Equation 8, R ( x _i ) and R ( y _i ) are ascending ordered values of activity of signal / metabolism transmission pathways from the results of microarray experiments to calculate Spearman's rank correlation. When calculating Spearman's rank correlation, if the value is a positive relationship of 0.6 or more, or a negative relationship of less than -0.6, then two microarray experiments will be performed. The processed condition can be judged to have a relationship, which means that there is a relationship between the signal / metabolism transmission pathways processed in both experiments. The absolute value of 0.6 for the relationship is an empirical basis for the statistical method of examining the relationship. Empirical evidence states that correlations are weak relations with absolute values greater than 0.4 and strong relations with absolute values greater than 0.7.

If the Spearman rank relation is greater than 0.6, the signal / metabolism transmission paths that are significant in the microarray test data, which are input data, are connected to each other, or the signal / metabolism transmission paths are If you include genes of one or more common elements, you can determine that spheres have a positive / negative relationship as the value of the rank relationship. Use this to establish connectivity of signal / metabolism transmission pathways and to read the relationship of experimental / process conditions.

The biological integrated network is output by integrating the relationship between the genetic network, experimental / processing condition specific network, activation signal / metabolism transmission pathway, and read experimental / processing condition.

As such, those skilled in the art to which the present invention pertains will understand that the present invention may be implemented in other specific forms without changing the technical spirit or essential features. Therefore, the above-described embodiments are to be understood as illustrative in all respects and not as restrictive. The scope of the present invention is shown by the following claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

1 is a microintegrated network for identifying genetic networks, experiment / treatment-specific protein networks, and experiment / treatment condition relationships using microarray data. Represents a schematic of a system for analyzing.

Figure 2 shows the form of the learning material of the microarray experiment processed in the input unit.

Figure 3 shows the form of the test data of the microarray experiment processed in the input unit.

4 shows an input unit screen.

5 illustrates the types of parameters used in the learning unit, the searching unit, and the searching unit.

6 shows a form in which the relationship between the experiment / process conditions is read using the relationship between the activation signal / metabolism transfer path of the experiment / process condition relationship readout unit.

7 illustrates a form of a biological integration network expressed by unifying the results of all output units output as one.

Claims (8)

Microarray data from a variety of conditions, such as knockout, drug, RNAi, and overexpression in a single species, and microarrays measured to identify mechanisms of action, such as drugs administered in disease or disease states. An input unit for inputting experimental data; A gene network learning unit for predicting a gene network from input microarray experiment data; Using the Mode-of-action by Network Identification (MNI) technique or the Bootstrapping Regression Model to explore the mechanism of action that provides predictive ranking information of genes for action mechanisms operating under experimental conditions. A search unit composed of an algorithm; A gene network output unit for outputting a gene network by interlocking the learned gene network with a searched action mechanism; A search unit configured of an algorithm for searching for protein-protein interaction information by using genes or searched mechanisms significantly expressed in input microarray experiment data; A protein network significance tester that examines the protein-protein interaction information by using GNEA (Geneset Network Enrichment Analysis) technique to examine the significance of a significant gene extracted from microarray experimental data or a protein network formed by a searched mechanism; A protein network output unit for outputting an experiment / condition specific protein network according to the examined significance; Search for activation signal / metabolism transmission pathways to determine whether genes expressed significantly in the input microarray experiment data or genes / proteins included in searched experimental / processing conditions specific protein networks exist in signal / metabolism pathways A search unit configured of an algorithm; The list of genes included in the detected active signaling / metabolic pathway is compressed by Singular Value Decomposition (SVD) to determine whether the detected activation signal / metabolic pathway is actually significant. An activation path significance checker to check; An activation signal / metabolism transmission path output unit displaying a significantly identified activation signal / metabolism transmission path; Experimental condition relational readout that analyzes the relationship of the condition of experiment / process by linking genes / proteins that are significantly expressed genes or searched experimental / process condition specific protein network with activation signal / metabolism transmission pathway ; And An integrated output unit comprising an algorithm for outputting a biological integrated network by giving the relationship of the experimental conditions read to the gene network output unit, the protein network output unit, and the activation signal / metabolism transfer path output unit Predicting / analyzing the gene network, experimental / treatment condition specific protein network, activation signal / metabolism transmission pathway using the expression value of the gene measured through the microarray experiment including the or reading the relationship between the experiment / treatment condition And output to the biological integration network. The algorithm of claim 1, wherein the algorithm of the searcher searches for a working mechanism using a bootstrapping regression model method by Equation 4 to learn a genetic network model from the input microarray experimental data and to predict the working mechanism. System characterized in that the output to the genetic network:

Where B represents the number of iterations of the bootstrapping process,

Is the regression coefficient obtained from the bootstrap sample for gene i ,

Is a random sample of the bootstrap extracted from a given pair of observations.

To show the regression coefficients fitted). The method of claim 1, wherein the algorithm for retrieving protein-protein interaction information comprises a list of genes or proteins consisting of significant genes of input microarray experimental data and a mechanism searched from a learned gene network. Experimental / Processing Conditions A system characterized in that it is an algorithm for searching or checking for significance with a specific protein network. The method of claim 1, wherein the algorithm for searching the activation signal / metabolism transmission pathway is characterized by a significant gene of input microarray experimental data and a mechanism of action searched from a learned gene network, and an experimental / processing condition whose significance is examined. A system for searching for activation signal / metabolism transmission pathways or reading the relationship between experimental / processing conditions using a list of genes or proteins composed of two or more elements from the list of genes / proteins included in the protein network . a) Microarray data obtained from treatment of various conditions such as knockout, drug, RNAi, and overexpression in a single species and measured to identify mechanisms of action such as drugs administered in disease or disease states. Inputting microarray experimental data; b) generating a genetic network using a general regression model or bootstrapping regression model using inputted microarray data and searching for a mechanism of action; c) predicting the experimental / processing condition specific protein network using the searched mechanism of action; d) searching for activation signal / metabolism transmission pathways using the list of genes / proteins involved in the searched mechanisms of action and experimental / processing specific protein networks; And e) reading the relationship of the experiment / treatment conditions using the retrieved activation signal / metabolic transfer pathway; Method for constructing a biological integration network using the relationship between the outputted genetic network, experimental / processing condition specific protein network, activation signal / metabolism transmission path information and the read experimental / processing conditions comprising a. 6. The method of claim 5, wherein step c) comprises microarray data obtained by treatment of various conditions of a single species such as knockout, drug, RNAi, and overexpression, and drug administered in a disease or disease state. Method for deriving a protein network using the identified genetic network and mechanism information using the measured microarray data to confirm the same mechanism of action. 6. The method of claim 5, wherein the steps d) and e) are performed in a microarray experimental data and a disease or disease state, which are treated with various conditions of a single species such as knockout, drug, RNAi, and overexpression. Derivation of the activation signal / metabolism transmission pathway using the gene network and mechanism information identified using the measured microarray data to confirm the mechanism of action such as the drug, The relationship between the experiment / treatment conditions is read. A recording medium having recorded thereon a computer readable program for performing the method of any one of claims 5 to 7.

Images