BC-Predict Database: A Curated Resource of Experimentally Validated Markers in Multidrug Resistance in Breast Cancer

Abstract

BackgroundIn this study, we aim to develop a yearly updatable database that could predict chemotherapeutic drug resistance and overall survival probability in breast cancer patients. Existing drug sensitivity databases depend on correlation-based predictions. In our study, candidates involved in drug resistance are chosen based on cell line validation (overexpression or downregulation or inhibition of candidates) studies, curated manually. Method28,773 mRNA expression signatures from 914 breast cancer patients were extracted from cProsite. 106 of these patients had clinical information and log2 fold change information required for this study. We categorized these patients into deceased and surviving groups from TCGA. To prepare a database that can predict drug resistance and overall survival, we included mRNAs that were over-expressed in at least 80% of the breast cancer patients and mRNAs over-expressed in deceased and surviving groups. In addition, we also reported breast cancer-associated drug resistance candidates which have been reported in cell-line based studies. The database matrix preparation involved an approximate of 15000 manual searches of cell validated studies. (750 candidates x 20 drugs). The database was validated using a publicly available breast cancer patient proteomics data. ResultsOur analysis identified a list of top priority candidates associated with multidrug resistance, categorized based on their resistance to >15 drugs, 5-15 drugs, and 2-4 drugs. Analysis of patient profiles in the database revealed that the number of proteins contributing to drug resistance was high in the poor prognosis category compared to the good prognosis category. ConclusionsOur study highlights the probable gaps in breast cancer drug resistance research, as only a small subset of overexpressed mRNA candidates found in patients are studied in vitro or in vivo experiments focusing on drug resistance. We also identified candidates involved in multidrug resistance, whose role in drug resistance has not been studied in more than 15 drugs. After further validations, this will benefit the clinicians and upcoming CRISPR gene therapeutics.