Predicting mutation-driven conformational shifts from one MD run
- Reweights one reference MD ensemble, state by state, to predict how a mutation redistributes conformational populations, avoiding a fresh converged simulation for each variant.
- Built-in diagnostics flag predictions the sampled data cannot support.
- Benchmarked on model peptides, a miniprotein, T4 lysozyme and the EGFR kinase domain; applicable only where the reference landscape is well sampled.
💡 Points Based on the abstract
If the reweighting holds, this is the cheap triage step I'd want before spending alchemical FEP budget on a mutant panel: rank variants by predicted population shift, then run full simulations only on the ones that matter. The catch is that it's only as good as the reference run's coverage, so the built-in diagnostics matter more than any benchmark number. I'd test it first on a kinase DFG-in/out question where we already have data.