Physical–chemical approach to identify local structural determinants of molecular mechanisms: Case study of antimalarial drug pyronaridine and crystal-growth inhibition

Understanding how specific molecular substructures control chemical behavior is central to rational molecular design and the development of new materials. However, most current predictive models offer limited mechanistic resolution at the fragmental level. We present a conceptually novel use of fragment-based structure-activity reasoning, based on systematically perturbing a parent molecule, to quantify fragment-level contributions to both a specific mechanistic action and a broader functional outcome. As a case study, we investigated local structural contributions to pyronaridine (PY), a clinically used antimalarial drug with a mechanistically distinctive mode of inhibition of hematin crystal growth via step-bunching. Chemically plausible PY molecular analogs have been computationally generated by selectively removing or substituting functional groups hypothesized to influence either step-bunching mechanisms or whole-parasite blood-stage activity. For each analog, we predicted the probability of four different crystal-growth inhibition mechanisms using a centroid-based similarity model based on a small dataset of experimentally verified crystal-growth inhibitors. The blood-stage antimalarial activity has also been estimated using the MAIP platform. A systematic comparison of molecular analogs revealed that step-bunching mechanisms depend primarily on two protonated pyrrolidines, with chlorobenzene as a strong secondary contributor. In contrast, antimalarial activity is more distributed, relying on coordinated interactions between aromatic-heteroatom scaffolds and an amine linker. The obtained results demonstrate that our approach can disentangle position-specific and cooperative fragmental effects, offering mechanistically interpretable guidance for the design of mechanism-optimized inhibitors. The framework might be broadly applicable across chemical and materials domains where linking local structure to specific mechanisms is essential.