MOTIVATION: Accurately identifying and prioritizing protein binding pockets is a foundational element of small-molecule drug discovery. Defining these known pockets currently relies on a laborious manual process of extracting key residue data from selected publications, reconciling inconsistent terminology, and independently computing volumetric representations. This manual curation to ensure biological relevance is time-consuming, error-prone, and represents a major bottleneck for efficient, high-throughput drug discovery. RESULTS: We present a novel approach for the identification and prioritization of protein binding pockets for small molecules by combining geometric pocket detection with large language models (LLMs). Our method leverages Fpocket to generate candidate pockets, which are then validated against published experimental data extracted from research articles using LLM with a series of prompts fine-tuned to identify and extract residue-level information associated with experimentally confirmed binding sites. We developed a curated benchmark dataset of diverse proteins and associated literature to train and evaluate the LLM's performance in paper relevance assessment and pocket extraction. AVAILABILITY AND IMPLEMENTATION: The developed benchmark dataset and methodology are freely available at the GitHub repository (https://github.com/receptor-ai/LLM-benchmark-dataset) and Zenodo (DOI: 10.5281/zenodo.15798647).

Coarse-grained (CG) simulations have become an essential tool to study a large variety of biomolecular processes, exploring temporal and spatial scales inaccessible to traditional models of atomistic resolution. One of the major simplifications of CG models is the representation of the solvent, which is either implicit or modeled explicitly as a van der Waals particle. The effect of polarization, and thus a proper screening of interactions depending on the local environment, is absent. Given the important role of water as a ubiquitous solvent in biological systems, its treatment is crucial to the properties derived from simulation studies. Here, we parameterize a polarizable coarse-grained water model to be used in combination with the CG MARTINI force field. Using a three-bead model to represent four water molecules, we show that the orientational polarizability of real water can be effectively accounted for. This has the consequence that the dielectric screening of bulk water is reproduced. At the same time, we parameterized our new water model such that bulk water density and oil/water partitioning data remain at the same level of accuracy as for the standard MARTINI force field. We apply the new model to two cases for which current CG force fields are inadequate. First, we address the transport of ions across a lipid membrane. The computed potential of mean force shows that the ions now naturally feel the change in dielectric medium when moving from the high dielectric aqueous phase toward the low dielectric membrane interior. In the second application we consider the electroporation process of both an oil slab and a lipid bilayer. The electrostatic field drives the formation of water filled pores in both cases, following a similar mechanism as seen with atomistically detailed models.