Disambiguate the source of target-dependent electrostatic-embedding performance

Determine whether the neutral performance of electrostatic embedding for CDK2, Thrombin, p38, and JNK1 arises from increased sampling cost and statistical variance caused by a more rugged potential-energy landscape, or from static charges not being the dominant error source for those targets.

Background

Electrostatic embedding substantially improves the TYK2 results but performs comparably to GAFF2 and mechanical-embedding AceFF-1.0 for CDK2, Thrombin, p38, and JNK1. The paper proposes two competing explanations: dynamic charge coupling may make the energy landscape more rugged and increase sampling demands, or static ligand charges may simply not be the principal source of error in these more flexible or complex systems. Because target rigidity, size, and net charge vary together in the benchmark, the study cannot identify which explanation is responsible.

References

We cannot yet distinguish this from the simpler explanation that static charges were never the dominant error source for these four, more flexible targets, since target rigidity, size, and net charge all vary together across our five systems.

— Evaluating Electrostatic Embedding MLIP/MM for Relative Binding Free Energy Calculations  (2608.13355 - Farr et al., 13 Aug 2026) in Section 2, subsection “Relative Binding Free Energy (RBFE) Benchmarks”