KR20200042295A - A drug repositioning system using network-based gene set enrichment analysis method - Google Patents

Abstract

The present invention relates to a network-based gene set augmentation analysis method. According to the present invention, provided is a method for effectively identify a set of pathway genes matching a gene expression phenotype by analyzing functional networks of neighboring genes as well as single genes. In addition, provided is a method for repositioning a known drug utilizing the method of the present invention.

Description

A DRUG REPOSITIONING SYSTEM USING NETWORK-BASED GENE SET ENRICHMENT ANALYSIS METHOD}

The present invention relates to a network-based gene set augmentation analysis method and a drug re-creation method using the same.

The molecular phenotype of clinical specimens is useful for disease diagnosis, patient stratification, and drug discovery. Gene expression profiling is the most accessible strategy for molecular phenotypic analysis of clinical specimens.

DNA chip technology and RNA sequencing are used for molecular profiling of patient-derived primary cells and cell lines.

Numerous gene expression profiles of clinical specimens are freely available from public databases such as Gene Expression Omnibus (GEO) and NCI Genomic Data Commons (GDC). Functional analysis of genome-wide phenotypes can generally be interpreted as a set of annotated genes rather than individual genes.

Therefore, various algorithms for gene set analysis have recently been developed. Many of these methods classify them as over-representation approaches by measuring the statistical significance of overlap between genes and gene sets specifically expressed in clinical samples. Although the above analysis methods are reasonably reasonable, there is a problem in that genes having low importance due to the degree of specific expression in clinical samples are treated as meaningless genes. A method developed to complement this is Gene Set Enrichment Analaysis (GSEA).

However, the method also has a drawback in that it analyzes mainly genes that have a large influence on expression under the control of the causal gene rather than genes that cause the disease.

Since the present inventors are more likely to change the expression of other genes that are actually functionally related to diseases, gene set analysis based on the gene expression data is performed on all neighboring genes of each gene on the functional network of genes. It was assumed that they should be analyzed in consideration of their expression information.

Accordingly, the present inventors have developed a network-based score (Natwork-based score) NS for performing gene set augmentation analysis by integrating expression information of neighboring genes of each gene, and using the score, a network-based gene set Augmented analysis (Network-based Gene Set Enrichment Analysis) was performed.

In addition, the present inventors have developed a predictive method for re-creating a drug that can newly treat a disease among known drugs through the network-based gene set augmentation analysis.

The present invention is to solve the above-mentioned problems of the prior art, and an object of the present invention is to provide a system for improving gene set augmentation analysis through a gene network and re-creating drugs against diseases.

According to an aspect of the present invention, (a) collecting gene set information including gene expression data from a commercialized database; (b) selecting a gene set that interacts with the collected gene set; And (c) integrating functional associations between the interacting gene sets based on a network-based score (NS) metric; and network-based gene set analysis. Methods for performing enrichment analysis (NGSEA) are provided.

In one embodiment, the network-based score may be calculated by Equation 1 below.

[Equation 1]

The n _i is the number of network neighbors of the i-th gene, and x _i and x _j are the expression scores of the i and j-th genes, respectively.

In one embodiment, the information on the gene set is KEGG PATHWAY Database (https://www.genome.jp/kegg/pathway.html), Drug SIGnatures DataBase (http://tanlab.ucdenver.edu/DSigDB/ Database including DSigDBv1.0 /), Gene Ontology Consortium (http://www.geneontology.org), DisGeNET (http://www.DisGeNET.org) and Diseases (https://diseases.jensenlab.org) It can be obtained from any one or more databases selected from the group.

In one embodiment, functional association between the gene sets may be implemented in a genomic-scale functional gene network.

In one embodiment, the network-based gene set enhancement analysis may be performed through an aggregate score approach.

According to another aspect of the present invention, (a) collecting information of a disease gene expression data set for a drug from a commercialized database; And (b) evaluating the association with the disease by listing the disease gene expression data sets in order of priority according to a network-based score.

In one embodiment, the information on the drug is Comparative Toxicogenomics Database (http://ctdbase.org/) or PubChem database (https://ftp.ncbi.nlm.nih.gov/pubchem/Compound/Extras/) It can be obtained from the database including.

In one embodiment, the disease gene expression data set for the drug is selected from the database group comprising Bioconductor (https://bioconductor.org/packages/release/data/experiment/html/KEGGdzPathwaysGEO.html), It can be obtained from more than one database.

In one embodiment, the method for re-creating the drug may further include (c) evaluating the treatment effect by treating the drug with diseased cells.

The gene set augmentation analysis method according to an aspect of the present invention can analyze and quantify associations between neighboring gene sets, thereby effectively providing significant data for disease-related gene discovery and drug re-creation methods.

It should be understood that the effects of the present invention are not limited to the above-described effects, and include all effects that can be deduced from the configuration of the invention described in the detailed description or claims of the present invention.

1 shows a schematic diagram of a network-based gene set enhancement analysis (NGSEA) method according to the present invention.
Figure 2 shows the predictive power of the relevant KEGG pathway for the disease expression data set matched by GSEA, AE and NGSEA, (A) shows the ranking distribution consistent with the KEGG pathway term in GSEA, AE and NGSEA, ( B) shows the ranking composition consistent with the KEGG pathway term in GSEA and NGSEA, (C) shows the normalized enrichment scores (NES) distribution of Pearson's correlation coefficient (PCC) in the same or different diseases, (D) is The subnetwork of the KEGG pathway term in Alzheimer's disease (HSA05010) and Staphylococcus aureus infection (HSA05150) is shown, and (E) the subnetwork of the KEGG pathway term in acute myeloid leukemia (HSA05221) and taste transduction (HSA04742).
Figure 3 shows the known drug search results for the disease expression data matched by CMap and NGSEA, (A) is a schematic diagram of the drug search method according to NGSEA, (B) is the search ability of CMap and NGSEA AUROC The results are compared.
FIG. 4 (A) shows the results of drug search for treatment of colorectal cancer (GSE9348) in CMap and NGSEA with an ROC curve, and FIG. 4 (B) shows the top 30 chemicals predicted for colorectal cancer treatment with NGSEA. 4 (C) shows cell viability after treatment with various concentrations (0-250 μM) of budesonide in HCT-116 cell line, and FIG. 4 (D) shows budesonide of various concentrations (0-250 μM). Cell viability after treatment on HT-29 cell line.

Hereinafter, the present invention will be described with reference to the accompanying drawings. However, the present invention may be implemented in various different forms, and thus is not limited to the embodiments described herein. In addition, in order to clearly describe the present invention in the drawings, parts irrelevant to the description are omitted, and like reference numerals are assigned to similar parts throughout the specification.

When a part is said to "include" a certain component, this means that other components may be further provided instead of excluding the other component unless otherwise stated.

Unless otherwise defined, molecular biology, microbiology, protein purification, protein engineering, and DNA sequencing and routine techniques commonly used in the field of recombinant DNA within the capabilities of those skilled in the art can be performed. These techniques are known to those skilled in the art and are described in many standardized textbooks and reference books.

Unless defined otherwise herein, all technical and scientific terms used have the meaning as commonly understood by one of ordinary skill in the art.

Various scientific dictionaries including terms included herein are well known and available in the art. Although any methods and materials similar or equivalent to those described herein are found to be used in the practice or testing herein, several methods and materials are described. Depending on the context used by those skilled in the art, the present invention is not limited to specific methodologies, protocols, and reagents because it can be used in various ways.

As used herein, a singular form includes a plurality of objects unless the context clearly dictates otherwise.

Hereinafter, the present invention will be described in more detail.

According to an aspect of the present invention, (a) collecting gene set information including gene expression data from a commercialized database; (b) selecting a gene set that interacts with the collected gene set; And (c) integrating functional associations between the interacting gene sets based on a network-based score (NS) metric; network-based gene set analysis. Methods for performing enrichment analysis (NGSEA) are provided.

1 shows an overall schematic diagram of a network-based gene set augmentation analysis method according to the present invention.

The network-based score can be calculated by Equation 1 below.

[Equation 1]

The n _i is the number of network neighbors of the i-th gene, and x _i and x _j are the expression scores of the i and j-th genes, respectively.

In Equation 1, by setting the score of a gene as an absolute value, all interactions with neighboring gene sets can be applied.

For example, if there are network neighbors B and C of gene set A, and B and C interact with A, respectively, to regulate the expression of A up (+) and down (-), the absolute gene score If you add without putting a value, the interaction value by B and C may be canceled, but if you put an absolute value in each value, the interaction score by B and C can be completely applied to the network-based score.

Information on the gene set can be collected from a commercialized database, and the commercialized database is a database for storing information on a gene set, KEGG PATHWAY Database (https://www.genome.jp/kegg/pathway. html), Drug SIGnatures Database (http://tanlab.ucdenver.edu/DSigDB/DSigDBv1.0/), Gene Ontology Consortium (http://www.geneontology.org), DisGeNET (http: //www.DisGeNET. org) and Diseases (https://diseases.jensenlab.org), but may be any one or more databases selected from the group.

Functional associations between the gene sets can be implemented in a genomic-scale functional gene network.

The over-representation approach that has been conventionally used in gene set expression analysis is that interaction with a less important gene set can be neglected, and information about the relative order between the differentially expressed gene sets cannot be provided. There is this.

Thus, in the present invention, in the analysis of expression of a gene set, a summation score approach may be used to assign a score of each annotated gene set based on a specific score to all gene sets neighboring the gene set.

According to another aspect of the present invention, (a) collecting information of a disease gene expression data set for a drug from a commercialized database; And (b) evaluating the association with the disease by listing the disease gene expression data sets in order of priority according to a network-based score.

Information about the drug is obtained from a database including Comparative Toxicogenomics Database (http://ctdbase.org/) or PubChem database (https://ftp.ncbi.nlm.nih.gov/pubchem/Compound/Extras/) can do.

The disease gene expression data set for the drug can be obtained from any one or more databases selected from the database group including Bioconductor (https://bioconductor.org/packages/release/data/experiment/html/KEGGdzPathwaysGEO.html). have.

The present inventors have developed a web-based gene set enrichment analysis server (www.inetbio.org/ngsea) to increase the usefulness of NGSEA.

Users can prioritize a set of functional genes representing biological and disease processes by various databases such as KEGG PATHWAY, Gene Ontology Consortium, DisGeNET and Diseases.

The user can simultaneously perform GSEA and NGSEA by submitting a gene expression phenotype.

Both Expression Matrix (.gct format) data and pre-scored gene lists (.rnk format) can be submitted as input for analysis.

Augmentation analysis on mouse genes can be performed with a genome-scale mouse functional gene network.

Users can prioritize gene sets according to ES, NES and FDR, and enrichment plots are also available.

The method for re-creating the drug may further include (c) evaluating the therapeutic effect by treating the drug with diseased cells.

Through the following examples, it will be apparent that the present invention is not limited by the following examples, although the present invention is further described.

Experimental Example 1: Gene expression profile, annotated gene set and functional human gene network

To evaluate the performance of gene set analysis for gene expression phenotypes, a standard expression data set consisting of expression profiles with KEGG pathway terms already annotated was used.

Using the GEO (KEGGdzPathwaysGEO) KEGG disease data set obtained from Bioconductor (https://bioconductor.org/packages/release/data/experiment/html/KEGGdzPathwaysGEO.html) as a standard data set for evaluating gene set enrichment analysis methods Did.

For example, the GSE21354 data set from KEGGdzPathwaysGEO annotates the KEGG pathway term 'glioma (hsa05214)' and contains microarray-based gene expression data containing 14 samples from tumor tissue and 4 normal tissues. Included.

Human KEGG pathway (https://www.genome.jp/kegg/pathway.html, June 2016) and drug target gene set from Drug Signature Database (DSigDB) (http://tanlab.ucdenver.edu/DSigDB, A version of the pathway gene was obtained from version 1).

Gene sets containing less than 15 genes were excluded from the analysis to establish basic parameters of the same criteria as GSEA.

For DSigDB, drug names were mapped to compound ID (CID) provided in PubChem database (http://ftp.ncbi.nlm.nih.gov/pubchem/Compound/Extras/).

Finally, 276 KEGG pathway gene sets and 165 DSigDB gene sets were used for analysis.

Curated annotation of Gene Ontology biological process (GOBP) annotation (http://www.geneontology.org, 2018.04.04.), DisGeNET (http://www.DisGeNET.org, 2018.06.08.) For web server construction In addition, an additional gene set of diseases having three or more stars per disease gene (https://diseases.jensenlab.org) was used.

To benchmark drug discovery for disease, 17,063 links were drawn between 2,109 diseases and 1,481 chemicals based on direct evidence of an association in the 'treatment' category of the Comparative Toxicology Genomics Database (CTD) ( http://ctdbase.org/ 2018.10.04.)

Information from the drug was combined with synonyms using CID.

Network-based gene expression analysis was implemented with HumanNet-XC (www.inetbio.org/humannet), a genome-scale functional gene network.

In other words, HumanNet-XC integrates protein-protein interactions as well as functional associations between genes derived from various types of omic data through Bayesian statistics.

HumanNet-XC contains 424,501 functional links between 17,790 human genes (94.6% of the coding genome).

A functional gene network for MouseNet (www.inetbio.org/mousenet) containing 788,080 links between 17,714 mouse genes (88% of the coding genome) was used to use the mouse gene expression phenotype in the web server for NGSEA. .

Experimental Example 2: GSEA, AE and NGSEA results comparison

The javaGSEA v3.0 software from Broad Institute (http://software.broadinstitute.org/gsea/downloads.jsp) was downloaded and used for analysis and web server implementation.

The javaGSEA can analyze input data of either GSEA or GSEA-preranked

The gene expression matrix included both control and experimental groups. The 'weighted GSEA-preranked' function, which is a basic parameter, was used to improve GSEA by changing the rank of the gene.

In the conventional GSEA, genes having the highest expression were selected based on the gene expression ratio, signal-to-noise ratio (SNR), or log ₂ (Ratio) of expression ratio.

SNR is the difference in mean expression value between the experimental group and the control group divided by the sum of the standard deviations of each group.

log ₂ (Ratio) was calculated by taking the ratio of the average expression value of the sample of the experimental group to the average expression value of the control sample and taking the logarithm of 2.

NGSEA modified the conventional gene-based score to a network neighbor gene-based score.

Specifically, the absolute value of the gene-based score was integrated as an average of the absolute values of the network-based neighbor gene-based score, and the network-based score (NS) for each gene was represented by Equation 1 below.

[Equation 1]

The n _i is the number of network neighbors of the i-th gene, and x _j is the expression score of the j-th gene. In the absence of gene expression data, the gene-based score was set to 0.

As a result of experimenting with both SNR and log ₂ (Ratio), log ₂ (Ratio) generally provided better results, so all results used log ₂ (Ratio) based on gene-based scores.

Genes were listed based on the absolute value of log ₂ (Ratio) for absolute enrichment (AE) analysis.

GSEA, AE, and NGSEA used the GSEA dictionary function to perform log ₂ (Ratio) value, log ₂ (Ratio) absolute value, and NS to list the gene list, respectively, and the GSEA dictionary function enrichment scores (ES) , normalized enrichment scores (NES), P-values and false discovery rate (FDR) values for each gene set based on modified Kolmogorov Smirnov (KS) test.

To evaluate the recovery performance of the gene set, the priority of the gene set was listed as absolute NES, which equally weights the gene set with high scores for both positive and negative in GSEA.

2A, the ranking distribution of NGSEA was significantly higher compared to GSEA and AE (P = 2.35e ^-3 and P = 4.0e ^-3 , respectively, by Wilcoxon signed rank test).

Referring to Figure 2B, the ranking of matching KEGG pathway conditions was improved by NGSEA compared to GSEA in 18 of the 24 (75%) tested disease expression data sets.

For example, the KEGG term 'Glioma' was searched 131 times in GSEA, but the gene expression data set derived from glioma samples (GSE21354) was searched 18 times in NGSEA.

On the other hand, the performance of AE was not significantly improved from GSEA (P = 0.11 by Wilcoxon signed rank test).

The results suggest that the main factor of improvement observed in NGSEA is by network-based analysis of gene expression data.

The robustness of the three enrichment assays was confirmed by comparing the assigned scores of the KEGG pathway between different expression profiles for the same disease.

Referring to FIG. 2C, all three concentration analysis results showed that the path score between the same diseases was significantly correlated with the path score between different diseases.

In particular, NGSEA improved the significance of correlation differences between the same and different disease groups compared to GSEA (P = 2.72e ^-6 and P = 3.44e ^-5 , Wilcoxon rank sum test, respectively).

The results suggest that NGSEA enrichment analysis has less effect on variability between expression profiles for the same disease process.

For example, referring to FIG. 2D, in the case of gene expression data for Alzheimer's disease (GSE5281_VCX), the network-based score measurement method put the KEGG term 'Alzheimer's disease' from 17th to 5th, and the majority of pathway genes Although highly rated in NGSEA (red), the KEGG term 'Staphylococcus aureus infection' went down from 6th to 267th, and the majority of pathway genes were highly rated in GSEA.

Referring to FIG. 2E, in the case of acute myeloid leukemia, it was confirmed that the ranking similarly changed between related and unrelated pathways.

The above results suggest that network-based scoring increased the ranking of true functional genes by further increasing the scores assigned to the true related gene set in the basic biological process of the sorted gene list for enrichment analysis.

Experimental Example 3: Drug repositioning using Connectivity Map (CMap)

The 24 KEGG disease gene expression data sets for FDA approved drugs retrieved from the CMap web server (https://portals.broadinstitute.org/cmap) were listed in order of priority.

Since CMap requires a list of up and down tags (Affymetrix HG-U133a probe ID) as input data, 50 up and down regulated probe IDs were selected from each of the 24 disease expression data sets.

If the input gene was not based on the Affymetrix HG-U133a probe ID, it was converted to the AffyMetrix HG-U133a probe ID to perform CMap analysis.

Referring to FIG. 3A, target genes for each FDA-approved drug were listed as a gene list according to a network-based score, and used as a functional gene set for testing association with disease.

Based on the active bioassay of DSigDB, a set of target genes for drugs was collected from drug-target links.

The drug was prioritized by NGSEA with a set of 24 gene expression data for 12 diseases of KEGGdzPathwaysGEO and a target gene set for 165 FDA-approved drugs from DSigDB with more than 15 targets.

The ability to search for known drugs for each of the 24 disease-related gene expression data sets of CMap and NGSEA was compared.

For benchmarking, 17,063 associations between 2,109 diseases and 1,481 chemicals were identified in the 'Treatment' category of the Comparative Toxicogenomics Database (CTD).

The performance of known drug recovery was benchmarked with an area under the receiver operating characteristic curve (AUROC).

AUROC analysis was performed with drugs included in both CMap and NGSEA to prevent biased evaluation by test drug differences.

Referring to Figure 3B, NGSEA AUROC for drug treatment was significantly improved compared to the CMap (P = 9.62e ^-4 , Wilcoxon signed rank test).

Specifically, the recovery of known drugs for a consistent disease gene expression data set in NGSEA improved from 16 to 24 compared to CMap.

NGSEA was particularly effective in anticancer drug screening.

Improved performance was observed in 14 of the 16 cancer-related expression data by NGSEA (87.5%), suggesting that NGSEA with medicinal targeting information may be an effective approach in reconditioning anticancer drugs.

Experimental Example 4: Analysis of anticancer effects by drug treatment

Cell viability was measured after drug treatment through MTS (3- (4,5-dimethylthiazol-2-yl) -5- (3-carboxymethoxyphenyl) -2- (4-sulfophenyl) -2H-tetrazolium) analysis.

The drug candidates selected in Experimental Example 3 were treated with colon cancer cell lines HCT116 or HT-29 at concentrations of 50 to 250 μM for 24, 48 and 72 hours, and MTS reagent was added. Cell viability was calculated by measuring absorbance at 490 nm on an ELISA microplate reader (Molecular Devices, USA). All experiments were repeated 6 times.

Referring to Figure 4A, the recovery performance of known anticancer agents is the greatest improvement in colorectal cancer (GSE9348) by NGSEA. AUROC values were determined to be 0.488 and 0.775 in CMAP and NGSEA, respectively.

Referring to FIG. 4B, 6 chemicals out of 30 predictions for the treatment of colorectal cancer by NGSEA were drugs currently used for colorectal cancer, 3 of which were treated for colorectal cancer (https://clinicaltrials.gov /) Has been clinically tested.

For the subsequent experimental verification, the anticancer effect of dobutamin (5th place) and budesonide (17th place) of colon cancer among the remaining candidates known to have no evidence of anticancer effect on colon cancer was confirmed. The dobutamine and budesonide were purchased from Sigma.

4C and 4D, in the cell viability analysis using the colon cancer cell lines HCT116 and HT-29, treatment with budesonide significantly inhibited cancer cell growth.

The above description of the present invention is for illustration only, and a person having ordinary knowledge in the technical field to which the present invention pertains can understand that it can be easily modified into other specific forms without changing the technical spirit or essential features of the present invention. will be. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined form.

The scope of the present invention is indicated by the following claims, and all modifications or variations derived from the meaning and scope of the claims and their equivalent concepts should be interpreted to be included in the scope of the present invention.

Claims (10)

(a) collecting gene set information including gene expression data from a commercialized database;
(b) selecting a gene set that interacts with the collected gene set; And
(c) integrating functional associations between the interacting gene sets based on a network-based score (NS) metric; network-based gene set enrichment analysis analysis; NGSEA). According to claim 1,
The network-based score is calculated by Equation 1 below.
[Equation 1]

The n _i is the number of network neighbors of the i-th gene, and x _i and x _j are the expression scores of the I and j-th genes, respectively. According to claim 1,
Information about the gene set is KEGG PATHWAY Database (https://www.genome.jp/kegg/pathway.html), Drug SIGnatures DataBase (http://tanlab.ucdenver.edu/DSigDB/DSigDBv1.0/), Any one selected from the database group including Gene Ontology Consortium (http://www.geneontology.org), DisGeNET (http://www.DisGeNET.org) and Diseases (https://diseases.jensenlab.org) The method obtained from the above database. According to claim 1,
A method in which functional linkages between the gene sets are implemented with a genomic-scale functional gene network. According to claim 4,
The genome-scale functional gene network is obtained from a database of HumanNet (www.inetbio.org/humannet) or MouseNet (www.inetbio.org/mousenet). According to claim 1,
The network-based gene set enhancement analysis is performed through an aggregate score approach. (a) collecting information of a disease gene expression data set for a drug from a commercialized database; And
(b) evaluating the association with the disease by listing the disease gene expression data sets in order of priority based on a network-based score. The method of claim 7,
Information about the drug is obtained from a database including Comparative Toxicogenomics Database (http://ctdbase.org/) or PubChem database (https://ftp.ncbi.nlm.nih.gov/pubchem/Compound/Extras/) How to re-create the drug. The method of claim 7,
The disease gene expression data set for the drug is obtained from any one or more databases selected from the database group comprising Bioconductor (https://bioconductor.org/packages/release/data/experiment/html/KEGGdzPathwaysGEO.html), , Drug re-creation method. The method of claim 7,
(c) evaluating the therapeutic effect by treating the drug with diseased cells; further comprising, a method for predicting drug re-creation.

Images