Toward Chemical Accuracy in Biomolecular Simulations through Data-Driven Many-Body Potentials: II. Polyalanine in...

Accurate modeling of biomolecular systems under physiological conditions requires potential energy functions (PEFs) that reliably capture many-body interactions both within the biomolecule and between the biomolecule and the surrounding aqueous environment. Here, we extend the intramolecular MB-nrg framework, previously developed for gas-phase molecules, to simulate biomolecules in solution using polyalanine in water as a proof-of-concept system. By integrating machine-learned terms trained on DLPNO-CCSD(T) reference data with physics-based expressions within the many-body expansion of the energy, we develop an MB-nrg PEF that simultaneously accounts for intramolecular interactions within polyalanine chains and intermolecular polyalanine-water interactions that shape the structure and dynamics of polyalanine in solution. The MB-nrg PEF adopts a unified, modular architecture spanning both polyalanine and water. We evaluate the performance of the MB-nrg PEF by comparing interaction energies, conformational free-energy surfaces, and hydration properties against the DLPNO-CCSD(T) reference data and empirical force fields. These analyses demonstrate the transferability of the MB-nrg formalism from the gas phase to aqueous environments for polypeptides, offering a scalable route toward chemically accurate biomolecular simulations.