Search for a command to run...
Hydrogen-deuterium exchange mass spectrometry is widely used to detect ligand-linked conformational responses, but it remains unclear how far simple observables from molecular dynamics can reproduce these patterns under fixed, auditable rules.A particular problem is that many MD-to-HDX studies rely either on post hoc directional proxies or on minimal forward models without testing whether those models actually recover the experimental signal on a defined benchmark.Here, I treat the problem as a constrained benchmark rather than a general prediction task.Using matched apo-holo simulations of a KRAS G12D ligand system and a 6OIM extension, I first tested a preregistered directional metric based on windowed changes in solvent exposure and conformational mobility, then compared it against a minimal forward model based on backbone amide hydrogen-bond occupancy, and finally introduced a frozen two-gate geometric state definition as a method-development upgrade.The directional benchmark showed conditional success in G12D: switch II had the most negative motif-average signal, switch I was also negative but weaker on average, and the P-loop was not negative overall.However, the strongest individual negative windows were not confined to switch II, and the same directional pattern did not reproduce robustly in the 6OIM extension, where the dominant negative windows remained off-target.The minimal forward model failed in both systems, with motif-level apo-holo differences close to zero and no meaningful recovery of the expected switch-centered contrast.The two-gate upgrade restored strong apo-holo state discrimination in G12D, but the same frozen gates failed to transfer to 6OIM and moved in the wrong direction.Together, these results show that simple MD-derived HDX proxies can have benchmark value in a favorable system, but neither a naive one-descriptor forward model nor a frozen two-gate upgrade yet provides a portable Cross system mapping from MD ensembles to HDX directionality Graphical Abstract caption.Benchmark workflow for evaluating simple MD-derived HDX proxies against a published KRAS directional anchor.A composite directional benchmark shows conditional success in the KRAS G12D system, whereas a minimal local forward model based on backbone amide hydrogen-bond occupancy fails to recover the same switchcentered contrast.A frozen two-gate geometric state definition improves apo-holo separation in G12D but does not transfer cleanly to the 6OIM extension, indicating benchmark value in a favorable system but limited cross-system portability.