Drug approval prediction based on the discrepancy between gene perturbation effects in cells and humans

Abstract Safety failures in clinical trials increase the cost of drug development. Appropriate drug target selection with minimal toxicity is critical for successful drug development. However, the discrepancy in drug target perturbation effects between cells and humans results in safety failures of drugs with poor translatability between preclinical and clinical research. To address this issue, we developed a machine learning framework to predict the probability of drug approval in clinical trials based on the discrepancy between effects on cells/humans. We hypothesized that genes with tolerance and intolerance to perturbation effects on cells and humans, respectively, were associated with drugs having safety problems, such as drugs that failed clinical trials and were withdrawn from markets. For the first time, we demonstrated that drug approval can be predicted based on the discrepancy between gene perturbation effects in cells and humans, which explains the safety failure of drugs during preclinical to clinical translation.