KR102388998B1 - Method and system for predicting sensitizer for overcoming cancer drug resistance - Google Patents

Abstract

The present invention uses an in silico (computer model) method rather than the conventional in vivo and in vitro methods, extracts the patient's resistance-related genes and resistance mechanisms based on a large-scale molecular network and patient's transcript data, and extracts the extracted resistance mechanism It relates to a method for predicting a sensitizer with an optimal anticancer effect that can effectively act on the drug and return it to the state before the development of resistance. When the method of the present invention is used, it is possible to select a sensitizer having an optimal anticancer effect to overcome resistance to anticancer drugs and to react sensitively again. will be able to be used.

Description

Computational method and system for predicting a sensitizer for overcoming the resistance of anticancer drugs {Method and system for predicting sensitizer for overcoming cancer drug resistance}

The present invention relates to a method (discovery of sensitizers to overcome cancer drug resistance, SOCAR) for overcoming anticancer drug resistance, and specifically, a patient's potential resistance gene using a large-scale molecular network and the patient's transcript data It relates to a method for extracting and tolerance mechanisms and predicting a sensitizer that exhibits optimal drug efficacy.

It is known that cancer is caused by abnormal proliferation and growth of cells due to various causes. These causes include exposure to chemical carcinogens, infection with oncogenic viruses, and congenital genetic abnormalities. However, fundamentally, all these causes are those that cause abnormalities in genes in cells. In normal cells, the functions of oncogenes, tumor suppressor genes, and apoptosis regulating genes are complementarily regulated to grow and maintain harmoniously. Oncogenes are normally essential for cell growth and differentiation, and contribute to cell proliferation and growth by promoting protein synthesis and transducing intracellular signals. known to contribute. Cancer is the second leading cause of death worldwide after circulatory disease, and research for the treatment of cancer has been extensively conducted. etc.

Anticancer drugs used in chemotherapy are drugs that inhibit the growth or proliferation of cancer cells. It is largely classified into cytotoxic anticancer drugs, targeted anticancer drugs, and immune anticancer drugs. Composite chemotherapy is used in which two or more drugs are used simultaneously to select anticancer drugs according to the type of cancer, the degree of progression, and the condition of the patient, and to increase the effect. . Cytotoxic anticancer drugs, known as first-generation anticancer drugs, have anticancer effects by directly attacking indiscriminate and rapidly differentiated cells. However, since normal cells with rapid differentiation characteristics (eg, bone marrow hematopoietic stem cells, hair follicle cells, intestinal mucosal cells, etc.) are also attacked by cytotoxic anticancer drugs, side effects such as leukocyte reduction, hair loss, vomiting, and diarrhea occur. There is this. Targeted anticancer drugs, known as second-generation anticancer drugs, act on specific proteins or specific gene changes that appear in cancer cells to block signal transduction involved in cancer growth and differentiation. Unlike cytotoxic anticancer drugs, it does not act on normal cells and has only a small number of side effects as it specifically acts on cancer cells.

However, one of the obstacles to the use of chemotherapy is the development of resistance to anticancer drugs. In order for chemotherapy to be successful for cancer patients, cancer cells must be killed at a blood concentration that allows normal tissues to survive. It is a case in which cancer cells do not die despite this. Anticancer drug resistance may vary from patient to patient, and may even be induced by various factors, including genetic differences between tumors derived from the same tissue. Each cancer cell derived from a single patient can acquire different genetic characteristics, and by 'mutation', not only gene expression diversity, but also activation of tumor-inducing factors and inactivation of tumor suppressors are shown. As a result, all cancers express anticancer drug resistance genes in different patterns, and cells in one cancer mass acquire diversity for drug resistance. In addition, even if tumors are not inherently resistant to specific chemotherapy, once exposed to chemotherapy, resistant cells selectively grow based on this diversity, and eventually many cancer cells quickly become resistant to chemotherapy. In this case, if a number of drugs having other intracellular target substances are used, it can be effectively treated and the cure rate can be increased. In many cases, however, cells are simultaneously resistant to structurally and functionally completely different agents. This phenomenon is called multi-drug resistance (MDR) and is caused by limiting the accumulation of intracellular drugs through limited absorption, increased release, or changes in membrane lipids such as ceramides. Multi-drug resistance inhibits apoptosis induced by most anticancer drugs, causes DNA damage repair, drug detoxication, and changes the cell cycle, thereby conferring additional resistance to anticancer drugs to cells. .

Therefore, there is a need to develop a sensitizer capable of reducing resistance to anticancer agents in order to overcome such multi-drug resistance. Anti-cancer sensitizers increase the sensitivity of cancer cells to chemotherapy to overcome drug resistance of existing anti-cancer drugs and help them kill cancer cells more efficiently. In other words, it is administered like existing anticancer drugs to overcome resistance and to react sensitively to anticancer treatment again. In previous studies related to sensitizers, breast cancer cell lines resistant to doxorubicin also showed resistance to resveratrol, and to overcome this, when resveratrol was administered together, cancer cell growth was inhibited. Confirmed to play a role (Kim, Tae Hyung, et al. "Resveratrol enhances chemosensitivity of doxorubicin in multidrug-resistant human breast cancer cells via increased cellular influx of doxorubicin." Biochimica et Biophysica Acta (BBA)-General Subjects 1840. 1 (2014): 615-625.), when Gefitinib, Erlotinib, and Trastuzumab were administered together in foretinib-resistant non-small cell carcinoma cells, it was confirmed that cancer cell growth was inhibited. (Liu, Li, et al. "Novel mechanism of lapatinib resistance in HER-2 positive breast tumor cells: activation of AXL." Cancer research 69. 17 (2009): 6871-6878.).

In order to find a sensitizer with an optimal effect to overcome resistance to existing anticancer drugs, it is important to analyze the resistance mechanism of anticancer drugs. However, anticancer drug resistance mechanisms are diverse and can be observed simultaneously. That is, since patients with the same disease have different resistance mechanisms to the same anticancer agent, a sensitizer based on the resistance mechanism should be proposed in a personalized way. The wet-experiment method carried out through in vitro and in vivo experiments performed so far provides accurate and reliable results in finding a sensitizer that can overcome resistance to anticancer drugs. However, it is difficult to analyze all the various resistance mechanisms of individual cancer patients, and when selecting the most potent drug through screening, the sensitizers that can be tested in terms of labor, time, and cost are extremely limited. There is a limit in which it is difficult to detect a sensitizer.

Therefore, the present inventors used an in silico (computer model) method rather than the existing in vivo and in vitro methods to extract potential resistance genes and resistance mechanisms of patients based on large-scale molecular networks and patient transcript data, and extracted resistance A method for predicting a sensitizer with an optimal anticancer effect that can effectively act on the mechanism and return to the state prior to the occurrence of resistance has been completed.

An object of the present invention is to provide a means for efficiently predicting a sensitizer for overcoming the resistance of an anticancer drug using a computational method.

In order to achieve the above purpose,

The present invention comprises the steps of extracting differentially expressed genes (DEG) by comparing before and after the onset of resistance to cancer (step 1);

extracting the genes (activity change genes, ACG) whose protein activity changes according to the change in the expression of the DEG (step 2);

extracting potential resistance genes (PRG) by combining the DEG and ACG (step 3);

extracting resistance mechanism genes (RMG) from among the potential resistance genes (PRG) (step 4);

quantifying the effect of a candidate sensitizer on the resistance mechanism genes (RMG) (step 5); and

calculating a potential sensitizer score (PSS) of the candidate sensitizer (step 6); It provides a computational method for predicting a sensitizer for overcoming the resistance of an anticancer agent, including

In addition, the present invention provides a computer-readable medium storing a program including instructions for executing a computational method for predicting a sensitizer for overcoming resistance to an anticancer drug.

Also, the present invention

a module for extracting differentially expressed genes (DEG) by comparing before and after the onset of resistance to cancer;

a module for extracting activity change genes (ACG) whose protein activity changes according to the change in the expression of the DEG;

a module for extracting potential resistance genes (PRG) by combining the DEG and ACG;

a module for extracting resistance mechanism genes (RMG) from among the potential resistance genes (PRG);

a numerical analysis module for quantifying the effect of a candidate drug on the resistance mechanism genes (RMG); and

a module for calculating a potential sensitizer score (PSS) of the candidate drug; It provides a system for predicting a sensitizer for overcoming the resistance of an anticancer agent comprising a.

When the present invention is used, it is possible to select a sensitizer having an optimal anticancer effect in order to overcome the resistance of the anticancer agent and to react sensitively again. The sensitizer having the optimal anticancer effect identified through the present invention may be usefully utilized for customized treatment according to the advent of the precision medicine era through experimental verification.

1 is a diagram illustrating the extraction of differentially expressed genes (DEG).
2 shows differentially expressed genes (DEG) and protein activity change genes (ACG) on a molecular network by selecting the group containing the largest number of PRGs. It is a diagram extracted as a major resistance mechanism.
3A is a diagram illustrating the effect of a candidate drug on resistance mechanism genes (RMG) using the RWR algorithm.
Figure 3b is a diagram showing the RWR probability score value for the resistance mechanism gene indicated in yellow.
4 is a diagram illustrating a potential sensitizer score (PSS) derived by adding the RWR probability score of each resistance mechanism genes (RMG) to a candidate drug.
5 is a diagram showing a candidate drug used for prediction of a sensitizer.
6 is a diagram showing the AUROC values of the model (SOCAR) and the known model (KsRepo) of the present invention.
7 is a diagram showing Anatomical Therapeutic Chemical Classification (ATC) codes of the top 20 candidate sensitizers.
FIG. 8 is a diagram illustrating GO term enrichment analysis of all resistance mechanism genes (RMG), and analysis of resistance mechanism genes significantly affected by the top five candidate sensitizers.
9 is a diagram showing the sensitization mechanism of selixipag and axitinib.
10 is a diagram illustrating the cell viability of 12 candidate sensitizers measured for each experimental group.
control (control): untreated MCF/TAMR cell line
Tamoxifen 1㎛: MCF/TAMR cell line treated with tamoxifen 1㎛
Sensitizer: MCF/TAMR cell line treated with 5 μm of each candidate sensitizer
Tamoxifen 1㎛ + Sensitizer: MCF/TAMR cell line treated with tamoxifen 1㎛ and each candidate sensitizer 5㎛

Hereinafter, the present invention will be described in detail.

The present invention comprises the steps of extracting differentially expressed genes (DEG) by comparing before and after the onset of resistance to cancer (step 1);

extracting the genes (activity change genes, ACG) whose protein activity changes according to the change in the expression of the DEG (step 2);

extracting potential resistance genes (PRG) by combining the DEG and ACG (step 3);

extracting resistance mechanism genes (RMG) from among the potential resistance genes (PRG) (step 4);

quantifying the effect of a candidate sensitizer on the resistance mechanism genes (RMG) (step 5); and

calculating a potential sensitizer score (PSS) of the candidate sensitizer (step 6); It provides a computational method for predicting a sensitizer for overcoming the resistance of an anticancer agent, including

The differentially expressed genes (DEG) of step 1 may be differentially expressed genes (DEG) whose mRNA expression is changed by comparing transcript data of cancer cells sensitive to anticancer drugs and anticancer drug-resistant cells. .

The cancer of stage 1 is breast cancer, bladder cancer, cervical cancer, bile duct cell cancer, chronic myelogenous leukemia (CML), colorectal cancer, esophageal cancer, stomach cancer, liver cancer, non-small cell lung cancer (Non-Small-Cell Lung Carcinoma, NSCLC) ), lymphoma, osteosarcoma, ovarian cancer, pancreatic cancer, prostate cancer, kidney cancer, and small cell lung cancer (Small-Cell Lung Carcinoma, SCLC) may be any one selected from the group consisting of, but is not limited thereto.

The gene (ACG) in which the activity of the protein of step 2 is changed may be extracted from a molecular network. A network is composed of a series of points (nodes) and lines (edges) connecting these points. The molecular network refers to a network including interactions between biological elements that can occur in a biological system, that is, a biological event. In a molecular network, nucleic acids, proteins, small molecules, chemicals, or metabolites constitute a node, and an edge is a directional line connecting the elements, and is a line between them. Represents an interaction.

The node may be a nucleic acid including a gene, a protein, a small molecule, a chemical, a metabolite, and the like, but is not limited thereto. The molecular network may be a physically directed interaction network between elements in a living body, and specifically may be a signaling network or an expression regulatory network, but is not limited thereto. does not

The molecular network may be extracted from a database that provides information on physically induced interaction between genes and proteins, or may be directly input through an input means. The input means may be any one selected from the group consisting of a keyboard, a scanner, a touchpad, a barcode input device, a mouse, a tablet, a trackball, an electronic pen, and a digital camera, and is not limited thereto as long as it can input data. The database is preferably a KEGG pathway, SIGNOR or TRANSFAC, but is not limited thereto.

The activity change genes (ACG) of the step 2 are genes in which the activity of the encoding protein is changed due to differentially expressed genes (DEG) upstream. That is, the change in the mRNA expression of the gene (ACG) whose protein activity is changed is not significant, so it is not extracted as the gene (DEG) whose expression changes, but the change in the mRNA expression of the gene (DEG) whose upstream expression changes is a downstream gene. It is based on the assumption that it can affect the protein activity of, and that the change in the protein activity can affect the anticancer drug resistance.

The resistance mechanism genes (RMG) of step 4 are genes present in the major resistance mechanism pathway. The main resistance mechanism maps the potential resistance genes (PRG) of step 3 in the molecular network, and after excluding uncombined components, the group containing the largest number of PRGs (largest connected component) can be extracted and determined. That is, the potential resistance gene (PRG) of step 3 forms several independent groups on the molecular network, and the major resistance mechanism is to select and extract the group containing the largest number of PRGs, Genes in the major resistance mechanism pathways are determined as resistance mechanism genes (RMG). This is based on the assumption that proteins related to the same disease interact with each other and tend to form clusters in a network, based on the assumption that they will show the same tendency not only in diseases but also in resistance mechanisms.

The quantification of the effect of the candidate sensitizer on the resistance mechanism gene of step 5 can be calculated using the RWR algorithm of Equation 1 below. Random walk with Restart (RWR) algorithm is a random walk-based ranking technique. Random walk refers to moving to a neighboring node by randomly selecting a neighboring edge from a starting node on a graph. If the random walk repeatedly moves infinitely, a probability of visiting each node is obtained. A random walker exists in the starting node, and the random walker can move to other connected nodes with a certain probability. The random walker of the present invention starts from a drug target and repeatedly moves from a current node to an adjacent node. In this process, the movement is repeated with a restart probability (r) for every movement. Therefore, by applying the RWR algorithm, it is possible to calculate the RWR probability score of each resistance mechanism gene.

[Equation 1]

: 분자 네트워크에 대한 정규화 된 인접 매트릭스,

: 재시작(restart) 확률,

: 시작 벡터,

: 확률 벡터,

:RWR 시뮬레이션의 결과)(

: normalized adjacency matrix for molecular networks,

: restart probability,

: start vector,

: probability vector,

:Result of RWR simulation)

The potential sensitizer score of step 6 can be calculated by Equation 2 below.

: 후보약물,

: 각 유전자의 RWR 확률 점수(RWR probability score),

: 내성 약물,

: 내성 기전 유전자(resistance mechanism gene)들의 집합)(

: Candidate drug,

: RWR probability score of each gene,

: resistant drugs;

: A set of resistance mechanism genes)

The RWR probability score of each resistance mechanism gene derived by the method of step 5 is calculated, and the sum of the RWR probability scores of the top 5 genes with the highest RWR probability score excluding the drug target, which is the starting node, is the potential sensitizer value. (potential sensitizer score, PSS). A candidate sensitizer with a high level of the potential sensitizer may be a potent sensitizer capable of overcoming anticancer drug resistance.

After step 6, the step of generating visualization data for the information derived through the visualization means and outputting it as visualization data through the output means may be additionally included. The output means may be any one selected from the group consisting of a monitor, a printer, and a plotter, and may be any means capable of outputting a result.

In addition, the present invention provides a computer-readable medium in which a program containing instructions for executing a computational method for predicting a sensitizer for overcoming the resistance of the anticancer agent is stored. The computer-readable recording medium includes all types of recording devices in which data readable by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc. include In addition, the computer-readable recording medium may be distributed in network-connected computer systems to store and execute computer-readable codes in a distributed manner.

In addition, the present invention provides a module for extracting differentially expressed genes (DEG) by comparing before and after the occurrence of resistance to cancer;

a module for extracting activity change genes (ACG) whose protein activity changes according to the change in the expression of the DEG;

a module for extracting potential resistance genes (PRG) by combining the DEG and ACG;

a module for extracting highly correlated resistance mechanism genes (RMG) from among the potential resistance genes (PRG);

a numerical analysis module for quantifying the effect of a candidate drug on the resistance mechanism genes (RMG); and

a module for calculating a potential sensitizer score (PSS) of the candidate drug; It provides a system for predicting a sensitizer for overcoming the resistance of an anticancer agent comprising a.

The system may additionally include an input module for inputting the molecular network through a computer input means, and a module for generating visualization data for information derived by the modules and an output module for outputting it as visualization data may additionally be included.

The input module may be any one selected from the group consisting of a keyboard, a scanner, a touchpad, a barcode input device, a mouse, a tablet, a trackball, an electronic pen, and a digital camera, and is not limited thereto as long as it can input data.

The output module may be any one selected from the group consisting of a monitor, a printer, and a plotter, and may be any means capable of outputting a result.

In a specific embodiment of the present invention, a gene (DEG) whose expression is changed by comparing the patient's transcript data is extracted, and a gene (ACG) whose protein activity is changed by constructing a large-scale molecular network is extracted to generate a potential resistance gene (PRG) was selected. The group (largest connected component) containing the largest number of potential resistance genes (PRGs) in the molecular network is extracted and selected as resistance mechanism genes (RMG), and the effect of candidate sensitizers on resistance mechanism genes The effect was quantified, and the sensitizer with the highest level was selected and its mechanism was analyzed. The sensitizer prediction model of the present invention has the advantage that the performance is superior to that of a known similar model. Accordingly, the method of the present invention can be effectively used for predicting and verifying a sensitizer for overcoming anticancer drug resistance for personalized drug prescription in the age of precision medicine.

Hereinafter, the present invention will be described in detail by way of Examples.

However, the following examples only illustrate the present invention, and the content of the present invention is not limited by the following examples.

< Example 1> To overcome anticancer drug resistance sensitizer (sensitizer) design of predictive model

Through the case study, ER-positive breast cancer, which accounts for 80% of all breast cancers, was selected as the target disease. Tamoxifen, the most widely used anticancer drug for the treatment of ER-positive breast cancer, has excellent anticancer effects, but is known to induce drug resistance in 40% of ER-positive breast cancer patients due to long-term administration. In the present invention, a sensitizer for overcoming tamoxifen resistance in ER-positive breast cancer was predicted by the method of Examples below.

<1-1> Extraction of differentially expressed genes (DEG)

By comparing transcriptome data of 8 types of tamoxifen-sensitive MCF7 cell line and 3 types of tamoxifen-resistant MCF7 cell line, genes with differentially expressed mRNA were compared. expressed genes, DEG) were extracted. The transcriptome data were collected from the Gene Expression Omnibus database (NCBI acc. number: GSE67916). In order to select DEGs with significantly changed expression, genes with an FDR value of less than 0.05 using the LIMMA package in R language that can analyze the change in gene expression and False Discovery Rate (FDR) threshold as a cutoff criterion. was extracted.

As a result, 801 differentially expressed genes (DEG) were extracted from tamoxifen-resistant breast cancer cells among 11,969 genes ( FIG. 1 ).

<1-2> Molecular network design

Molecular networks were extracted from three public databases (KEGG pathway, SINOR, and TRANSFAC). The molecular network was designed by dividing it into a signaling network and an expression regulatory network. The signaling network consists of physically directed interactions between the KEGG pathway and human proteins derived from the SIGNOR database, and the regulatory network consists of the KEGG pathway, the SIGNOR database, and TRANSFAC. It consists of a physically directed interaction between human proteins and human genes derived from

The KEGG pathway used to design the signal transduction network and expression control network was used after downloading the KEGG Markup Language (KGML) file corresponding to each pathway and pre-processing the data using the KEGG graph package in R studio. Among the pathways provided by KEGG, the signal transduction network was extracted using seven types (activation, inhibition, phosphorylation, dephosphorylation, glycosylation, ubiquitination, methylation), and the expression control network was divided into two types (expression, repression). extracted.

The SIGNOR database includes interactions between various types of molecules such as proteins, chemicals, and small molecules, and each molecular interaction has a respective mechanism. Therefore, an expression regulatory network was designed by extracting interactions (transcriptional regulation, transcriptional activation, and transcriptional inhibition) according to the type of mechanism. In addition, a signaling network was designed by extracting the interactions between proteins along with other mechanisms such as phosphorylation and ubiquitination. In addition, the expression control network was also extracted using TRANSFAC.

As a result, a large-scale molecular network consisting of 42344 signaling interactions and regulatory interactions between 6910 proteins or genes was constructed.

<1-3> Extraction of activity change genes (ACG)

Since all genes related to resistance cannot be considered only by extracting the gene (DEG) whose expression changes in Example 1-1, genes (activity change genes, ACG) with protein activity change were additionally extracted. .

Genes whose protein activity changes (activity change genes, ACG) mean that the change in mRNA expression of the gene is insignificant, but the activity of the protein encoded by the gene (DEG) whose expression upstream is changed is changed, It is a gene whose activity of a downstream protein is changed due to a signal transduction relationship. That is, the gene was extracted under the assumption that there is a possibility that the altered activity of the ACG-encoding protein may cause resistance.

Genes (DEGs) with 801 expression changes extracted in Example 1-1 in a signaling network consisting of 27067 signaling interactions between 4620 proteins constructed in Example 1-2 ), 211 genes (DEGs) whose expression changes were mapped. Genes (ACGs) in which the activity of proteins downstream of 211 DEGs are changed were extracted, and these genes were not extracted in Example 1-1.

As a result, activity change genes (ACG) were extracted from 670 proteins, and AKT, PI3K, EGFR and CTNNB1 genes were not extracted from genes whose expression changes (DEG), but were was extracted as a gene (ACG). Tamoxifen is known to bind to 18 targets, of which 1 is DEG and 9 is ACG, overlapping the target genes of tamoxifen. Therefore, the above results suggest the possibility that dysregulation appearing in the mechanism of action of tamoxifen is related to drug resistance (FIG. 2).

<1-4> Extraction of potential resistance genes (PRG)

The 801 differentially expressed genes (DEG) extracted in Example 1-1 and the 670 protein activity change genes (ACG) extracted in Example 1-3 were combined to give a total 1471 potential resistance genes (PRG) were extracted (FIG. 2).

<1-5> Extraction of resistance mechanism genes (RMG)

Based on the research results that disease-related genes and proteins interact with each other and tend to form network clusters (Barabasi, A. -L., Gulbahce, N. & Loscalzo, J. Network medicine: a network-based approach) to human disease. Nature reviews genetics 12, 56 (2011)), hypothesized that the same tendency would be exhibited not only in disease but also in resistance mechanisms.

Specifically, only the potential resistance genes (PRG) of Examples 1-4 were extracted from the molecular network constructed in Examples 1-2. The extracted PRGs are connected to each other on the molecular network and exist as several independent groups. At this time, the PRGs in each group are connected to each other and the group (largest connected component) containing the largest number of PRGs was selected and extracted, and this was regarded as the major resistance mechanism. The major resistance mechanism consists of 877 PRGs, that is, 220 DEGs and 657 ACGs, and 877 PRGs were named resistance mechanism genes (RMG) (FIG. 3a).

In addition, in order to extract the biological process related to tamoxifen resistance, GO term enrichment analysis of the resistance mechanism genes (Eden, E., Navon, R., Steinfeld, I., Lipson, D., & Yakhini, Z) (2009), GOrilla: a tool for discovery and visualization of enriched GO terms in ranked gene lists. BMC Bioinformatics, 10(1), 48.) was performed. Among 4354 biological mechanisms, 1541 mechanisms showed a high degree of relationship in the resistance mechanism gene (RMG), and the top 20 mechanisms were selected (Table 1).

ranking Gene ontology term False Discovery Rate (FDR) One neurotrophin signaling pathway (GO:0038179) 1.37E-87 2 neurotrophin TRK receptor signaling pathway (GO:0048011) 3.07E-86 3 positive regulation of kinase activity (GO:0033674) 7.51E-74 4 response to peptide (GO:1901652) 7.51E-74 5 cellular response to nitrogen compound (GO:1901699) 2.24E-70 6 immune response-regulating cell surface receptor signaling pathway (GO:0002768) 1.09E-69 7 response to peptide hormone (GO:0043434) 1.09E-69 8 cellular response to organonitrogen compound (GO:0071417) 4.14E-68 9 cellular response to hormone stimulus (GO:0032870) 1.87E-66 10 fibroblast growth factor receptor signaling pathway (GO:0008543) 1.25E-64 11 response to fibroblast growth factor (GO:0071774) 1.13E-63 12 epidermal growth factor receptor signaling pathway (GO:0007173) 1.52E-63 13 positive regulation of protein kinase activity (GO:0045860) 1.92E-63 14 cellular response to fibroblast growth factor stimulus (GO:0044344) 1.92E-63 15 ERBB signaling pathway (GO:0038127) 6.35E-63 16 blood coagulation (GO:0007596) 1.46E-62 17 coagulation (GO:0050817) 1.46E-62 18 hemostasis (GO:0007599) 7.07E-62 19 cellular response to peptide (GO:1901653) 7.55E-62 20 cellular response to peptide hormone stimulus (GO:0071375) 1.38E-60

<1-6> Selection of candidate sensitizers

In order to predict a candidate sensitizer, it was hypothesized that if the candidate drug has a significant effect on the regulation of the resistance mechanism gene (RMG), it can return to the state before resistance occurs and there is a possibility of increasing the sensitivity of the anticancer agent to cancer cells.

Under this assumption, the effect of a candidate sensitizer on 877 resistance mechanism genes (RMG) in a molecular network was simulated using the Random Walk with Restart (RWR) algorithm of Equation 1 below (FIG. 3a).

[Equation 1]

: 분자 네트워크에 대한 정규화 된 인접 매트릭스,

: 재시작(restart) 확률,

: 시작 벡터,

: 확률 벡터,

:RWR 시뮬레이션의 결과)(

: normalized adjacency matrix for molecular networks,

: restart probability,

: start vector,

: probability vector,

:Result of RWR simulation)

In Equation 1, the W value is a normalized adjacency matrix for a molecular network composed of a signal transduction interaction and an expression control interaction, and the restart probability r is set to 0.7. The starting vector P0 is set to 30 for the starting node and 0 for other nodes. The probability vector Pt was iteratively updated until the difference between Pt+1 and Pt was less than 1E-05. Pt+1 is the result of RWR simulation.

Since the random walker moves from a drug target to an adjacent gene, drugs whose target is not included in the molecular network are excluded from the drug list of DrugBank. 4009 drugs were simulated with the RWR algorithm, among which 13 known sensitizers and 3996 unknown drugs were included (FIG. 5). After performing the RWR algorithm, all genes had RWR probability scores, and RWR probability scores of resistance mechanism genes were selected ( FIG. 3b ). By summing the RWR probability scores of the selected resistance mechanism genes, a potential sensitizer value (PSS) of 0-30 points was calculated by Equation 2 below (FIG. 1d). In addition, to confirm that candidate sensitizers are effective and safe for co-administration with tamoxifen, the DCDB and Twosides databases were used. The two databases provide information on side effects and drugs that have synergetic/antagonistic effects when drugs are co-administered.

[Equation 2]

: 후보약물,

: RWR 확률 점수(RWR probability score),

: 내성 약물,

: 내성 기전 유전자(resistance mechanism gene)들의 집합)(

: Candidate drug,

: RWR probability score,

: resistant drugs;

: A set of resistance mechanism genes)

As a result, the average of the potential sensitizer level (PSS) values of all 4009 genes was 7.97, and the average value of 1358 FDA-approved drugs was 7.72, and 20 candidate sensitizers with high PSS values were selected. (Fig. 4 and Table 2). As a result of analyzing the Anatomical Therapeutic Chemical Classification (ATC) codes of the selected 20 candidate sensitizers from DrugBank, it was confirmed that they belonged to 9 classifications among 14 first level ATC codes (FIG. 7). As a result of analysis using DCDB and Twosides database, there were no known side effects or synergetic/antagonistic effects when co-administered with tamoxifen among 20 candidate sensitizers. In the above results, gefitinib selected at 27th and metformin selected at 31st are known to have the potential to cause side effects such as renal tubular acidosis or carotid bruit in tamoxifen-resistant cells. They have the potential to become candidate sensitizers without side effects when administered together with existing anticancer drugs.

ranking drug name DrugBank ID One selexipag DB11362 2 Axitinib DB06626 3 Framycetin DB00452 4 Plerixafor DB06809 5 Sunitinib DB01268 6 Lenvatinib DB09078 7 Pegademase Bovine DB00061 8 Ramucirumab DB05578 9 Cinacalcet DB01012 10 Tirofiban DB00775 11 Insulin Lispro DB00046 12 Insulin Glargine DB00047 13 Anakinra DB00026 14 Eptifivatide DB00063 15 Ado-Trastuzumab Emtansine DB05773 16 Pertuzumab DB06366 17 Nintedanib DB09079 18 Chloramphenicol DB00446 19 Afatinib DB08916 20 Insulin Aspart DB01306

< Example 2> sensitizer predictive model ( SOCAR ) and comparison with existing models

<2-1> Performance evaluation of sensitizer prediction model

In order to evaluate the performance of the sensitizer prediction model of Example 1, Area under the receiver operating characteristic curve (AUROC) values were compared. The model of the present invention showed an AUROC value of 0.87 for 1358 FDA-approved drugs and an AUROC value of 0.75 for all 4009 drugs. In the present invention, since activity change genes (ACG) are defined for the first time, resistance mechanism genes that consider only genes whose expression changes (DEG) in order to measure the effect of ACG on predicting performance of a sensitizer (RMG) and AUROC values were compared in two cases of resistance mechanism gene (RMG) considering both DEG and ACG.

As a result, the AUROC value of the RMG model considering only DEG showed a value of 0.78 for FDA-approved drugs, and the AUROC value of the RMG model considering both DEG and ACG showed a value of 0.87 for FDA-approved drugs, It was confirmed that the performance of the RMG model considering both DEG and ACG was excellent ( FIG. 6 ).

<2-2> Comparison with the existing model

An in silico analysis model that predicts a sensitizer to overcome anticancer drug resistance has not been reported so far. Therefore, the performance was compared with the most comparable model, KsRepo.

Specifically, KsRepo is a drug repositioning model that outputs a candidate drug as a score when a disease-related gene is input. KsRepo can predict an effective drug based on individual disease data and has a similar purpose to the present model in terms of predicting drug effects in a personalized way. A gene (DEG) whose expression is changed in Example 1-1 was input, and a drug output as a candidate drug was determined as a candidate sensitizer.

As a result, the AUROC value of KsRepo for FDA-approved drugs was 0.61, and the AUROC value for all drugs was 0.63. The above results suggest that the SOCAR model of the present invention exhibits superior performance than the existing KsRepo model (FIG. 6).

< Example 3> Candidate of sensitizers Mechanism analysis

<3-1> Mechanism analysis of candidate sensitizers

In-depth analysis was performed on FDA-approved drugs among the drugs proposed as candidate sensitizers. As a result, according to the ATC code available from DrugBank, 9 of the top 20 drugs were anticancer drugs, 2 of them were used for breast cancer, and the remaining 7 were used for other types of cancer. The other 11 drugs were used to treat non-cancer diseases such as hypertension and diabetes mellitus.

To analyze the mechanism of action of the sensitizer, the mechanisms of the top 5 drugs (selexipag, axitinib, framycetin, plerixafor, sunitinib) among the 20 sensitizer candidates in Table 2 were analyzed by GO term enrichment analysis (Eden, E., Navon, R). ., Steinfeld, I., Lipson, D., & Yakhini, Z. (2009). GOrilla: a tool for discovery and visualization of enriched GO terms in ranked gene lists. BMC Bioinformatics, 10(1), 48.) analyzed. For each drug, the top 20 GO term biological processes were extracted from the top 50 resistance mechanism genes (RMG) with a high RWR probability score (FDR>0.001) (FIG. 8). As a result, the top 20 GO term mechanisms of each candidate drug largely overlapped with the 877 resistance mechanism genes (RMG) extracted in Examples 1-4. The above results show that the top five candidate sensitizer drugs have the potential to revert to the pre-resistance by acting on the main resistance mechanism. The top five drugs were found to have a large effect on the top two mechanisms in Table 1, the neutrophin TRK receptor signaling pathway (GO: 0048011) and the neutrophin signaling pathway (GO: 0038179). As targeting neutrophin results in inhibition of tumor cell growth and metastasis of breast cancer, the top five drug candidates may restore neutrophin-related signaling, which has been altered due to resistance, to resensitize tamoxifen to cancer cells. suggests that there is a possibility that

<3-2> Mechanism analysis of selexipag

In Table 2, the sensitization mechanism was analyzed for selexipag, a candidate sensitizer with the highest PSS.

Specifically, selexipag is a cellular response to glucagon stimulus in addition to the neutrophin TRK receptor signaling pathway (GO: 0048011) and neutrophin signaling pathway (GO: 0038179), which are the mechanisms most affected by the top five candidate sensitizers in Example 3-1. (GO:0071377) and response to glucagon (GO:0033762) have a significant effect. Glucagon activates GCCR as a peptide hormone and induces the activities of GNAS, ADCY, cAMP, and PKA proteins at the bottom. Among them, the PKA protein plays a role in inhibiting cell growth by activating the extrinsic apoptosis pathway induced outside the MCF7 cell. In the tamoxifen-resistant condition, genes encoding GNAS, ADCY, and PKA proteins were extracted as genes (ACGs) in which the activity of the proteins was changed. Since the expression of two proteins (RAMP1, IGF1R) among the three upstream proteins (RAMP1, IGF1R, LPAR6) that activate GNAS in tamoxifen-resistant MCF7 cells was decreased, it is presumed that the expression of GNAS was decreased, and the activity signal intensity of GNAS was decreased. is weakened, reducing PKA activity, resulting in increased cell growth. As an agonist of PTGIR, selexipag increases GNAS activity and inhibits cell growth, thereby reversing the pre-tamoxifen resistance ( FIG. 9 ).

<3-3> Mechanism analysis of axitinib

In Table 2, the sensitization mechanism was analyzed for axitinib, a candidate sensitizer with the second highest sensitizer level.

Specifically, axitinib was used in the neutrophin TRK receptor signaling pathway (GO: 0048011) and neutrophin signaling pathway (GO: 0038179), which are the mechanisms most affected by the top five candidate sensitizers in Example 3-1, as well as the EGFR signaling pathway (GO). :0007173) and the ERBB signaling pathway (GO:0038127). EGFR (HER1) and ERBB (HER2) promote cell growth, proliferation, angiogenesis and cell survival through mediating proteins such as MAPK and BAD (FIG. 9). In the tamoxifen-resistant condition, the downstream proteins of HER1 and HER2 were extracted as a gene (ACG) with a change in protein activity. The expression of four genes (PTPN21, PIK3R1, ITGA6, FGFR2) among the upstream genes (IGF1R, PTPN21, PIK3R1, ITGA6, FGFR2) of SRC, the upstream modulator farthest from the mediator proteins, was increased, PI3K is the expression of seven genes (IRS2, LYN, LIFR, FGFR2, NRAS, FYN, GAB1) out of ten upstream genes (PGR, ERBB4, IGF1R, IRS2, LYN, LIFR, FGFR2, NRAS, FYN, GAB1). As it increased, it is speculated that the expression of SRC and PI3K was increased. MAPK and BAD activation induced by increased expression of SRC and PI3K increases cell growth, proliferation, angiogenesis and cell survival and induces tamoxifen resistance. Axitinib acts as an antagonist of VEGFR and reduces the activity of top modulators such as SRC and PI3K, thereby returning abnormal cell behavior to the state before resistance was developed (FIG. 9).

< Experimental example 1> of sensitizers Verification of the effect of overcoming anticancer drug resistance

In order to verify the effect of overcoming anticancer drug resistance of the drug derived as a candidate sensitizer using the sensitizer prediction method of Example 1, the following experiment was performed.

Specifically, the tamoxifen-resistant MCF7 breast cancer cell line (hereinafter referred to as MCF/TAMR) (AX4011, AXOL Bioscience, Cambridge, UK) was treated with 10% fetal bovine serum (FBS, Cat # 16000044, Gibco, California, USA) and insulin at 6 ng/ml. (Cat # 41400045, Gibco, California, USA) was added and cultured in DMEM/F-12 GlutaMAX supplemented medium without phenol red (Cat # 10565018, Gibco, California, USA). The cultured MCF/TAMR cell line was dispensed in a 96-well plate at 4×10 ³ cells/well, and 1 μm of tamoxifen and/or 12 sensitizers were treated with 5 μm. After 48 hours of treatment, cell viability was measured through WST-8 assay (CCK-8, Dojindo, Kumamoto, Japan). After adding WST-8 reagent and incubating at 37°C for 4 hours, absorbance was measured at a wavelength of 450 nm using an xMark Microplate Absorbance Spectrophotometer (Bio-Rad, California, USA). The survival ratio of cells treated with tamoxifen and/or a sensitizer with respect to control cells that were not treated with anything was derived by Equation 1 below. In order to increase the accuracy, three or more independent experiments were performed.

<Formula 1>

Cell viability (%) = (absorbance of wells treated with tamoxifen and/or sensitizer-absorbance of blank wells)/(absorbance of control wells-absorbance of blank wells)

In order to analyze the difference in cell viability for each experimental group derived by Equation 1 above, a student t-test with an assumed equal variance was performed using R version 3.2.0 (R Foundation for Statistical Computing, Austria). .

As a result, the cell viability (98.22±1.102%) of MCF/TAMR cells resistant to tamoxifen hardly decreased when only tamoxifen was treated. In addition, the cell viability of only six drugs, selexipag, plerixafor, sunitinib, lenvatinib, eptifibatide and chloramphenicol, showed a value of more than 90% when a single treatment was performed, indicating multi-drug resistance to other drugs other than tamoxifen. All candidate sensitizers used in the experiment showed reduced cell viability compared to the tamoxifen-treated group and the single sensitizer-treated group, so it was observed that the effect of the combined administration was shown, compared to the control group not treated with anything. Several candidate sensitizers were shown to significantly inhibit cell growth (p<0.0005). In addition, the combination of pertuzumab and tamoxifen showed the most significantly improved anticancer drug resistance overcoming effect compared to the control group (p <0.0001) and the group treated with pertuzumab alone (p=0.0251) (Table 3 and FIG. 10). The above results show that the candidate sensitizer derived by the sensitizer prediction method according to the present invention actually has the effect of overcoming anticancer drug resistance.

sensitizer name process Tamoxifen Throughput sensitizer throughput cell
viability (%) ± standard error of measurement (SEM) p-value (versus control) p-value (compared to the sensitizer-only group) selexipag control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 90.52 1.178 0.056 Tamoxifen + sensitizer 1 uM 5 uM 83.65 0.8484 0.0002 0.0009 Axitinib control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 86.78 1.196 < 0.0001 Tamoxifen + sensitizer 1 uM 5 uM 79.2 1.378 < 0.0001 0.0006 Plerixafor control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 90.44 1.878 0.0062 Tamoxifen + sensitizer 1 uM 5 uM 78.62 1.144 < 0.0001 < 0.0001 Sunitinib control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 90.41 0.7089 0.0075 Tamoxifen + sensitizer 1 uM 5 uM 82.82 1.748 < 0.0001 0.0008 Lenvatinib control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 95.78 2.45 0.1779 Tamoxifen + sensitizer 1 uM 5 uM 86.52 2.368 0.0051 0.0278 Cinacalcet control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 87.23 3.026 0.0007 Tamoxifen + sensitizer 1 uM 5 uM 74.04 2.828 < 0.0001 0.0052 Tirofiban control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 89.52 1.602 0.0118 Tamoxifen + sensitizer 1 uM 5 uM 77 3.084 < 0.0001 0.0136 Eptifivatide control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 93.54 2.343 0.2536 Tamoxifen + sensitizer 1 uM 5 uM 82.07 3.349 0.0009 0.0125 Pertuzumab control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 1 uM 88.51 2.296 0.0234 Tamoxifen + sensitizer 1 uM 1 uM 77.64 2.858 < 0.0001 0.0251 Nintedanib control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 70.28 1.32 < 0.0001 Tamoxifen + sensitizer 1 uM 5 uM 62.4 1.967 < 0.0001 0.0007 Chloramphenicol control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 91.09 2.664 0.0338 Tamoxifen + sensitizer 1 uM 5 uM 81.45 3.356 < 0.0001 0.021 Afatinib control - - 100 1.549 - Tamoxifen 1 uM - 98.22 1.102 0.6065 sensitizer - 5 uM 88.19 2.633 0.0057 Tamoxifen + sensitizer 1 uM 5 uM 77.65 1.66 < 0.0001 0.0105

Claims (8)

extracting differentially expressed genes (DEG) by comparing before and after the onset of resistance to cancer (step 1);
extracting the genes (activity change genes, ACG) whose protein activity changes according to the change in the expression of the DEG (step 2);
extracting potential resistance genes (PRG) by combining the DEG and ACG (step 3);
extracting resistance mechanism genes (RMG) from among the potential resistance genes (PRG) (step 4);
quantifying the effect of a candidate sensitizer on the resistance mechanism genes (RMG) (step 5); and
calculating a potential sensitizer score (PSS) of the candidate sensitizer (step 6); A computational method for predicting a sensitizer for overcoming the resistance of an anticancer drug performed in a computing device, comprising a.
The method of claim 1, wherein the cancer is breast cancer, bladder cancer, cervical cancer, bile duct cell cancer, chronic myelogenous leukemia (CML), colorectal cancer, esophageal cancer, stomach cancer, liver cancer, non-small cell lung cancer (Non-Small-Cell Lung) Carcinoma, NSCLC), lymphoma, osteosarcoma, ovarian cancer, pancreatic cancer, prostate cancer, kidney cancer, and small-cell lung cancer (Small-Cell Lung Carcinoma, SCLC) characterized in that any one selected from the group consisting of, performed on a computing device A computational method for predicting a sensitizer to overcome resistance to anticancer drugs.
According to claim 1, wherein the gene (ACG) in which the activity of the protein of step 2 is changed is extracted from a molecular network, a sensitive agent ( A computational method to predict the sensitizer.
The method according to claim 1, wherein the activity change genes (ACG) of the step 2 change the activity of the protein due to differentially expressed genes (DEG) upstream. A computational method for predicting a sensitizer for overcoming the resistance of an anticancer drug performed in a computing device, characterized in that.
The computing device according to claim 1, wherein the quantification of the effect of a candidate drug on the resistance mechanism genes (RMG) of step 5 is simulated through a Random walk with restart (RWR) algorithm. A computational method for predicting a sensitizer to overcome resistance to anticancer drugs.
According to claim 1, wherein the potential sensitizer score of step 6 (potential sensitizer score, PSS) is characterized in that calculated by the formula expressed by the following Equation 2, Overcoming the resistance of anticancer drugs performed in a computing device A computational method for predicting the sensitizer for
[Equation 2]

(Sj: candidate drug, Pk: RWR probability score,
Dj: resistance drug, r(Dj): a set of resistance mechanism genes)
A computer-readable medium storing a program comprising instructions for executing a method for predicting a sensitizer for overcoming resistance to an anticancer drug according to any one of claims 1 to 6.
a module for extracting differentially expressed genes (DEG) by comparing before and after the onset of resistance to cancer;
a module for extracting activity change genes (ACG) whose protein activity changes according to the change in the expression of the DEG;
a module for extracting potential resistance genes (PRG) by combining the DEG and ACG;
a module for extracting highly correlated resistance mechanism genes (RMG) from among the potential resistance genes (PRG);
a numerical analysis module for quantifying the effect of a candidate drug on the resistance mechanism genes (RMG); and
a module for calculating a potential sensitizer score (PSS) of the candidate drug; A system for predicting a sensitizer for overcoming the resistance of an anticancer agent comprising a.

Images