Use ThermoMPNN

The input protein structure to analyze and predict the stability changes caused by specific mutations within the protein.

The chain(s) of the input structure to analyze and predict the stability changes caused by specific mutations within the structure. You can leave this field blank to analyze the entire structure.

ThermoMPNN offers three prediction modes tailored to different mutation scenarios. The **Single Mode** predicts stability changes for individual mutations. The **Additive Mode** estimates the effects of double mutations by summing the contributions of each mutation independently, without accounting for interactions between them. For a more detailed analysis, the **Epistatic Mode** captures the complex interactions between mutations, delivering a comprehensive prediction of stability changes.

The Threshold option determines which mutations are saved based on their predicted stability changes. By default, ThermoMPNN saves only stabilizing mutations (with ΔΔG ≤ -0.5 kcal/mol). To include all mutations, including destabilizing ones, set the threshold to a high value (e.g., 100).

The Distance option sets the threshold for identifying "nearby" residues used in additive or epistatic predictions. This helps filter residues likely to exhibit epistatic interactions. Smaller values apply stricter filtering, while the default is 15 Å, measured based on Cα–Cα distance.

The Include Cysteine option allows predictions for cysteine mutations; however, it is not recommended due to potential assay artifacts from disulfide bond formation, which may result in overly favorable predictions.

The Penalize option ensures that breaking disulfide bonds is explicitly penalized. While ThermoMPNN usually detects and penalizes disulfide breakage automatically, this option applies an additional penalty to ensure such changes are always disfavored.