MLSolvA: solvation free energy prediction from pairwise atomistic interactions by machine learning

Abstract Recent advances in machine learning technologies and their applications have led to the development of diverse structure–property relationship models for crucial chemical properties. The solvation free energy is one them. Here, we introduce a novel ML-based model, which calculates from pairwise atomistic interactions. novelty proposed model consists simple architecture: two encoding functions extract atomic feature vectors given structure, while inner product between results 6239 experimental measurements achieve outstanding performance transferability enlarging training data owing its solvent-non-specific nature. An analysis interaction map shows that our has significant potential producing group contributions on energy, indicates provides not only predictions target properties but also more detailed physicochemical insights.