> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tamarind.bio/llms.txt
> Use this file to discover all available pages before exploring further.

# ThermoMPNN

> Point mutations to improve stability

### Find stabilizing mutations

ThermoMPNN scores point mutations by their ΔΔG°, or change in thermostability from the wildtype. The method shows state-of-the-art performance on established benchmarks, including *Ssym* and *S669*.

### Methods

ThermoMPNN is built on top of ProteinMPNN, a machine learning method that generates novel sequences to fold into a given structure. Specifically, ThermoMPNN is trained on a large dataset (776,000 data points) of ΔΔG° measurements. It uses ProteinMPNN's learned knowledge about sequence-structure relationships by first generating a representation of the protein with ProteinMPNN and using that representation to predict stability.

### Speed

ThermoMPNN is very fast, and can evaluate all possible point mutations of a given input within seconds to minutes, while showing greater accurate than more computationally intensive approaches like MD and MM.

### ThermoMPNN-D

In addition to scoring single mutations, ThermoMPNN-D also supports making two mutations at a time, while considering the other residue's context.

On Tamarind, you can automatically score the top mutations from ThermoMPNN with Alphafold2 to verify their stability.

**Inputs:**

* pdbFile - input wildtype structure to be scored
* chains - chain(s) you want to evaluate mutations on
* multi - select this if you have multiple chains and want them all included in context

**Outputs:**

* ddG\_pred - Predicted ΔΔG° between wildtype and mutant protein, negative means stabilizing mutation
* Heatmap of stability of each residue x amino acid pair

[Try ThermoMPNN](https://app.tamarind.bio/tools/thermompnn)