Bioinformatics analysis for drug safety assessment based on the difference in genotype–phenotype relationship between preclinical models and humans

Abstract

Inconsistent drug target perturbation effects between preclinical models (cell and mouse models) and humans have been a major hurdle in drug development, resulting in high costs and adverse effects on patients. To address this challenge, it is crucial to accurately assess drug approval and safety with consideration of the discrepancy in drug target perturbation effects between preclinical models and humans. My graduate research focuses on developing a machine learning model based on the preclinical models/humans discrepancy to predict drug approvals in clinical trials and explain the drug safety failures during translation from preclinical to clinical studies. In my research, I presented a method to predict drug approvals in clinical trials based on the cells/humans discrepancy in gene perturbation effects. In detail, CRISPR-Cas9 knockout and loss-of-function mutation rate-based gene perturbation effects on cells and humans, respectively, were analyzed to evaluate those abilities in predicting drug approvals. I found that the cells/humans discrepancy is a key feature in predicting drug approvals. In addition, I validated that genes, tolerant to perturbation effects on cells but intolerant in humans, were associated with drugs with safety issues such as failed and withdrawn drugs. Motivated by previous studies focused on assessing drug safety through chemical properties, drug approval prediction was improved by integrating chemical information with the cells/humans discrepancy. Discrepancy-based machine learning framework holds the promise of significantly reducing the drug attrition rates. The outcomes of our investigation provide valuable biological insights that can guide the development of new drug candidates with low toxicity and the identification of safe therapeutic drug targets for the benefit of patients.