biostudies-literatureWozniak SDivision of Molecular and Cellular BiosciencesNational Institute of General Medical SciencesNIGMS NIH HHS4830-4845https://www.ebi.ac.uk/biostudies/studies/S-EPMC12080100biostudies-literatureUnknown21(9)Machine learning has emerged as a promising approach for predicting molecular properties of proteins, as it addresses limitations of experimental and traditional computational methods. Here, we introduce GSnet, a graph neural network (GNN) trained to predict physicochemical and geometric properties including solvation-free energies, diffusion constants, and hydrodynamic radii, based on three-dimensional protein structures. By leveraging transfer learning, pretrained GSnet embeddings were adapted to predict solvent-accessible surface area (SASA) and residue-specific p<i>K</i>_<i>a</i> values, achieving high accuracy and generalizability. Notably, GSnet outperformed existing protein embeddings for SASA prediction and a locally charge-aware variant, aLCnet, approached the accuracy oJournal of chemical theory and computationAccurate Predictions of Molecular Properties of Proteins via Graph Neural Networks and Transfer Learning.PMC12080100R35 GM126948MCB 2210228Janson GFeig MWozniak SfalseAccurate Predictions of Molecular Properties of Proteins via Graph Neural Networks and Transfer Learning.Machine learning has emerged as a promising approach for predicting molecular properties of proteins, as it addresses limitations of experimental and traditional computational methods. Here, we introduce GSnet, a graph neural network (GNN) trained to predict physicochemical and geometric properties including solvation-free energies, diffusion constants, and hydrodynamic radii, based on three-dimensional protein structures. By leveraging transfer learning, pretrained GSnet embeddings were adapted to predict solvent-accessible surface area (SASA) and residue-specific p<i>K</i>_<i>a</i> values, achieving high accuracy and generalizability. Notably, GSnet outperformed existing protein embeddings for SASA prediction and a locally charge-aware variant, aLCnet, approached the accuracy o2025-01-01T00:00:00Z2025 May2026-06-01T14:59:17.914Z2026-04-08T13:31:28.008ZS-EPMC120801004027030410.1021/acs.jctc.4c01682