| Predicting drug response of a patient based on various genetic information is a funda-mental problem in current research of precision medicine. Nowadays, regression model or classification model with machine learning method was established to predict anticancer drug sensitivity based on various genetic information. However, there is a strong correlation between groups of genes, so research on the correlation between the genetic information systems, especially the relationship between the gene expressions co-expressions and the anticancer drug sensitivity, are of great significance.In this paper, we developed a systems biology framework to identify gene expressions, co-expressions, and co-expression modules differentially changed with drug sensitivity, and then applied this framework to the Cancer Cell Line Encyclopedia gene expression and drug response data. More than 4,000 genes are inferred to be drug response associated (DRA) genes for at least one drug. While the number of DRA genes for each drug varies dramatically from almost 0 to 1,226. Functional enrichment analysis shows that the DRA genes are significantly enriched in genes associated with cell cycle and plasma membrane. There are significantly shared DRA genes between male and female for most drugs, while very little DRA genes tend to be shared between the two genders for a few drugs targeting sex-specific cancers (e.g., PD-0332991 for breast cancer and ovarian cancer). Our analyses also show substantial difference for DRA genes between young and old samples, suggesting the necessity of considering the age effects for personalized medicine in cancers. Lastly, differential module and key driver analyses confirm cell cycle related modules as top dif-ferential ones for drug sensitivity. The analyses also reveal the role of TSPO, TP53, and many other immune or cell cycle related genes as important key drivers for DRA network modules. These key drivers provide new drug targets to improve the sensitivity of cancer therapy. |
PDF Full Text Request