Comparative analysis of molecular representations in prediction of drug combination effects

Abstract

Application of machine and deep learning (ML/DL) methods in drug discovery and cancer research has gained a considerable amount of attention in the past years. As the field grows, it becomes crucial to systematically evaluate the performance of novel DL solutions in relation to established techniques. To this end we compare rule-based and data-driven molecular representations in prediction of drug combination sensitivity and drug synergy scores using standardized results of 14 high throughput screening studies, comprising 64,200 unique combinations of 4,153 molecules tested in 112 cancer cell lines. We evaluate the clustering performance of molecular fingerprints and quantify their similarity by adapting Centred Kernel Alignment metric. Our work demonstrates that in order to identify an optimal representation type it is necessary to supplement quantitative benchmark results with qualitative considerations, such as model interpretability and robustness, which may vary between and throughout preclinical drug development projects.