Protein mutation recommendation based on thermostability (ThermoMPNN)

ThermoMPNN is a protein thermal-stability mutation scanning (SSM) tool suite based on the ProteinMPNN deep-learning architecture. It contains three specialized models: ThermoMPNN (single-point mutations), ThermoMPNN-D (double mutations, with additive and epistatic modes) and ThermoMPNN-I (insertion/deletion mutations), systematically predicting the effect of every possible mutation on protein stability (ddG values, kcal/mol).

This tool provides three scan modes: single (single-point mutation scan), additive (fast double-mutation approximation by additivity) and epistatic (exact double-mutation prediction with a siamese network). Upload the backbone structure of your target protein, and the system automatically computes the ddG values of all possible mutations and displays them sorted by the selected effect direction.