This is the authors' abstract. We don't add key points for this paper.
Hybrid machine-learning and molecular mechanics formulations have been growing over the years. The attractive cost-effectiveness for big systems (such as protein complexes) has made these methods gain attention in the computational chemistry/biology field. In these theories, one of the most discussed themes is embedding, which can add or remove physics from the system. Here, we present POLEN (POLarizable Embedding for Neural-network potentials), an OpenMM plugin for ML/MM simulations with a mutually polarizable AMOEBA embedding, including a variant whose ML-trained head (MuPE(ML)) predicts multipoles, polarizabilities, and charge penetration at each simulation step. We tested it on three systems: a push-pull chromophore which a strong solvatochromic shift (p-nitroaniline in water); a protease-inhibitor complex with a buried salt bridge (trypsin-benzamidinium); and a kinase-inhibitor complex with a halogen bond (CK2α–TBB−).
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Cunha et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: