Optimisation-based modelling for explainable lead discovery in malaria

Yutong Li, Jonathan Cardoso-Silva, John M. Kelly, Michael J. Delves, Nicholas Furnham, Lazaros G. Papageorgiou, Sophia Tsoka

Abstract: Highlights•Interpretability in machine learning for Quantitative Structure-Activity Relationship modelling in drug discovery is important.•An optimisation-based method for piecewise linear regression, modSAR, applied on data from the Open Source Malaria project is reported.•Owing to the mathematical nature of modSAR, inference of regression coefficients by the algorithm is used to derive feature importance and drive compound screening.•Our methodology is validated through virtual screening and wet-lab experiments to allow identification of potential novel antimalarial lead compounds.