Leave-One-Out-Cross-Validation Based Machine Learning Algorithms for Efficient, Scalable, and Robust Prediction of Artemisinin Resistance in Plasmodium falciparum Kelch13 mutants.

The emergence of resistance to frontline artemisinin derivatives in Plasmodium falciparum poses a significant threat to global malaria control efforts. Assessing the ART-sensitive (ART-S) or ART-resistant (ART-R) nature of parasite strains with Kelch13 (K13) mutations is currently a time-consuming process. The ring-survival assay (RSA), a gold standard for ex vivo laboratory adaptation, often fails to accurately predict clinical ART-responsiveness, making it unsuitable for quickly identifying novel K13 variants.

AI-derived models, however, show promise in screening 100 K13 variants within hours, enabling rapid near-real-time prediction of ART-S/R status. To achieve this, researchers have engineered a panel of forty representative K13 variants by computing their relative deviations from the wild-type K13 across ten physicochemical and structural descriptors.

These descriptors capture mutation-induced changes in folding, stability, and solvent exposure, providing a compact feature set that can be used to train and validate an AI classifier for predicting resistance.