BPS2026 – Mechanistic insights into glucose transport for PfHT1 from molecular dynamics and graph attention models

Transporter proteins regulate molecular flux across membranes through complex cycles of large-scale conformational changes, yet these transitions remain difficult to resolve experimentally. To address this challenge, we performed over 800 μs of adaptive molecular dynamics simulations to capture the complete conformational cycle of the Plasmodium falciparum hexose transporter PfHT1 in both apo and holo states. The resulting trajectories revealed transitions among outward-facing, occluded, and inward-facing states with atomistic detail, allowing us to map the free-energy landscape for conformational cycle and of glucose translocation. Construction of Markov state models (MSMs) quantified transition probabilities and mean first passage times across states, providing a kinetic framework for directional transport. Through our simulation we also explored the role of TM7b kinking in conformational change as well as substrate transport process. Comparison of apo and holo simulations highlighted conformations preferentially stabilized by glucose, shedding light on the energetic basis of substrate recognition and specificity. To interpret the complex residue-residue interaction patterns that drive these conformational shifts, we applied a graph attention network (GAT) trained on interaction graphs derived from the simulations. The GAT framework identified influential “hinge-like” nodes and key interaction motifs that remain preserved throughout the transport cycle, linking local substrate coordination with global conformational switching. This machine learning approach provided a data-driven complement to conventional contact analysis, enabling mechanistic interpretation of how structural elements orchestrate transport. Together, these results establish an atomistic transport mechanism for PfHT1 by integrating adaptive sampling, MSM kinetics, free-energy landscapes, and graph-based learning. Beyond PfHT1, our work demonstrates a broadly transferable computational framework for studying the conformational dynamics, energetics, and substrate specificity of membrane transporters and other complex molecular machines.