Discovering Drug Resistance Signatures From Gene Expression Profiles Of Cell Line Models

Currently a problem prevalent in cancer treatment is the emergence of drug resistance. This is true not only for the traditional chemotherapy, but also for the targeted drug therapy. The method mainly used by researchers is establishing drug-resistant cell lines model in vitro to study drug resistance mechanisms and screening gene expression biomarker for drug resistance. However,there were some flaws in the analysis of cell lines, which include lack of standardized operating procedures for the establishment of the resistant cell line model and strict statistical analysis is not performed, as a result, discovering gene expression biomarker for drug resistance in different studies is not reproducible. To solve this problem, in this paper we integrating multiple datasets of resistant cell lines that were established by different treatment, developing the new method to standardize the analysis process for cell line, finally we obtain reproducible gene expression biomarker for drug resistance.The main contributions are as follows: firstly, contrary to using artificial fold change threshold to find differentially expressed genes in previous work, we proposed a statistical model to assess the reproducibility of differentially expressed genes in technique replicates of resistant cell lines, then filtering consistent pattern of differentially expressed genes among the independent datasets that had different ways of cell culture and treatment as gene biomarkers for drug resistance. In addition, we extracted the reproducible functions associated with drug resistance from significantly enrichment functions within different datasets. Secondly, the algorithm we proposed was applied for sceening gene expression biomarkers for drug resistance in the three commonly clinical cancer targeted drug resistant cell lines models. The first one is tamoxifen-resistant breast cancer cell lines, we using the proposed algorithm identified 91 genes as drug-resistant gene biomarker, and found tamoxifen resistance primarily associated with lysosomes, RNA transport and other functions. The second one is oxaliplatin-resistant colon cancer cell lines, we using our algorithm identified 20 genes as gene biomarker for drug resistance, and found that oxaliplatin resistance was associated with mRNA modification, base conversion or substitution editing. The third case is erlotinib-resistant lung cancer cell lines, with our improved algorithm, we found 18 differentially expressed genes as gene biomarker for drug resistance. and through reproducible functional analysis, we found that erlotinib resistance was related to cell proliferation.