Guido Barducci

Abstract Predicting how residue variations affect protein stability is crucial for rational protein design and for assessing the impact of disease-related mutations. Recent advances in protein language models have revolutionized computational protein analysis, enabling more accurate predictions of mutational effects. However, balancing predictive accuracy with the fundamental laws of thermodynamics remains a challenge for sequence-based models.

1 INTRODUCTION Predicting protein stability changes upon single-point mutations is a longstanding challenge in computational biology (Benevenuta et al., 2022; Pucci et al., 2022; Sanavia et al., 2020), with significant implications in drug design, enzyme engineering, and understanding disease mechanisms (Thomas et al., 1995). Protein stability is typically quantified by measuring the Gibbs free energy change (ΔG) between the folded and unfolded states Δ G = G F − G U . $$ \Delta G={G}_F-{G}_U.