The clinical information and the Fragments PerKilobase Million (FPKM) values of mRNA expression data containing 482 colon cancer and 42 normal control samples were downloaded from The Cancer Genome Atlas (TCGA, https://cancergenome.nih.gov/). According to the clinical data, we obtained 22 CRC patients who were only treated with L-OHP and 5-FU (Table S1). GSE17536 dataset was downloaded from the GEO database . GSE17536 was performed on Affymetrix Human Genome U133 Plus 2.0 Array (Affymetrix, Santa Clara, CA, USA), including 177 colon cancer patients. Data were downloaded from the publicly available database hence it was not applicable for additional ethical approval.
Weighted gene correlation network analysis (WGCNA)
WGCNA was performed to obtain the modules which was associated with the chemotherapeutic resistance (L-OHP and 5-FU). A total of 6175 genes in the top 35 % of variance were screened from the data set containing 22 CRC patients who were only treated with L-OHP and 5-FU, using the R package WGCNA . In a brief, we chose the power = 6 for weighting the correlation matrix following an approximate scale free topology. Subsequently, gene expression modules with similar patterns were identified based on a gene cluster dendrogram and by using the dynamic tree cut method (minModuleSize = 50, mergeCutHeight = 0.3, deepSplit = 1). The unsigned network type was used to keep the relationships between modules and chemotherapeutic resistance. Generally, the correlation between the phenotype and module eigengenes was considered as the module-trait associations. Therefore, modules with p < 0.05 and person correlation value > 0.4 were considered significantly related to the chemoresistant traits. Additionally, the threshold of genes related to chemotherapeutic resistance was set at person correlation of gene vs. module-membership value > 0.5 and person correlation of gene vs. trait-correlation value > 0.5.
Establishment of the prognostic gene signature
Only CRC patients with a follow-up period longer than 1 months were included for survival analysis. Univariate Cox regression analysis was performed by R package survival (https://CRAN.R-project.org/package=survival) to identify prognostic genes, and genes were considered significant with a cut-off of p < 0.05. Then, Lasso-penalized Cox regression analysis was performed to further select prognostic genes for overall survival in patients with CRC. Then a prognostic gene signature was constructed based on a linear combination of the regression coefficient derived from the Lasso Cox regression model coefficients (β) multiplied with its mRNA expression level.
The risk score=
The optimal cut-off value was investigated by the R package survminer (http://www.sthda.com/english/rpkgs/survminer/) and two-sided log-rank test. Patients were classified into a high-risk and low-risk according to the threshold (medium of risk score). The time-dependent receiver operating characteristic (ROC) curve was drawn to evaluate the predictive value of the prognostic gene signature for overall survival using the R package survivalROC . The Kaplan-Meier survival curve combined with a log-rank test was used to compare the survival difference in the high- and low-risk group using the R package survival. As the prognostic genes could predict the prognosis of CRC, so they were considered as the hub genes among the genes related to the chemotherapeutic resistance. Then the predictive value of the prognostic hub-gene signature was further investigated in the GSE17536 testing cohort.
Independent prognostic role of the gene signature
To investigate whether the prognostic hub-gene signature could be independent of other clinical parameters (including gender, age and stage), univariate and multivariate Cox analyses were performed by R package survival using the Cox regression model method with forwarding stepwise procedure with the cutoff of p < 0.01.
Pathway enrichment analysis
Kyoto Encyclopedia of Genes and Genomes (KEGG) is a data base for systematic analysis of gene function which links genomic information with higher-level systemic function . After obtaining the genes related to the chemotherapeutic resistance by WGCNA, the R package clusterProfiler  was used to perform a KEGG over-representation test, at a cutoff of p < 0.05. Simultaneously, Gene Set Enrichment Analysis (GSEA) were performed  to explore the potential molecular mechanisms, with a cutoff of FDR < 0.05 and p < 0.05. Then, combining the two results to get the key pathway involved in the chemoresistance.
The Human protein-protein interaction (PPI) network used in this analysis was retrieved from the STRING database (medium confidence: 0.2) . The PPI network was constructed with the genes related to the chemotherapeutic resistance and core genes involved in the key pathways were visualized using Cytoscape 3.7.1 .
Shortest pathway analysis
To find possible connections between hub genes and the signaling pathways of interest, we performed a shortest-pathway analysis. The shortest pathway is defined as the minimum number of edges required to travel from one node in the PPI network to another. The Python package NetworkX (http://networkx.github.io) was applied to compute all the shortest paths between hub genes and the key pathway genes.
Identifing of Potential Drug-repurposing
The 3D structure of ANO1 was download from SWISS-MODEL (Q5XXA6, https://swissmodel.expasy.org/) and its biding site was found by Schrodinger maestro 2019-1 . Then, we built a library of 2106 FDA approved drugs obtained from ZINC15 database . Finally, we performed virtual screening and molecular docking by Schrodinger maestro 2019-1 to find the potential drug-repurposing.
Statistical analysis was performed using R 3.5.3 (R Foundation for Statistical Computing, Vienna, Austria). Qualitative variables were analyzed using the Pearson test or Fisher’s exact test; quantitative variables were analyzed using a t-test for paired samples. Multiple groups of normalized data were analyzed using one-way ANOVA. If not specified above, p < 0.05 was considered statistically significant.