Design of a Protein with Improved Thermal Stability by an Evolution‐Based Generative Model

Efficient design of functional proteins with higher thermal stability remains challenging especially for highly diverse sequence variants. Considering the evolutionary pressure on protein folds, sequence design optimizing evolutionary fitness could help designing folds with higher stability. Using a generative evolution fitness model trained to capture variation patterns in natural sequences, we designed artificial sequences of a proteinaceous inhibitor of pectin methylesterase enzymes. These inhibitors have considerable industrial interest to avoid phase separation in fruit juice manufacturing or reduce methanol in distillates, averting chromatographic passages triggering unwanted aroma loss. Six out of seven designs with up to 30 % divergence to other inhibitor sequences are functional and two have improved thermal stability. This method can improve protein stability expanding functional protein sequence space, with traits valuable for industrial applications and scientific research. Protein design bears great interest in both research and industry. Evolutionary‐based methods have been promising in designing new sequences based on the concept that the evolutionary fitness of a sequence is closely related to folding stability. Here we present a computational pipeline that, thanks to a combination of sequence‐ and structure‐based methods, can design new sequences translating in functional proteins with enhanced thermal stability.