BayeStab Method

BayeStab evaluates the effects of single mutations on protein stability by calculating the quantitative changes in unfolding Gibbs free energy. The predictions are based on the protein structure. So, the users need upload the wild protein and mutation protein pdb format files.

The Rosetta can be used to generate the mutation protein files.

1. Go to https://els2.comotion.uw.edu/product/rosetta to get an academic license for Rosetta.

2. Download Rosetta 3.13 (source + binaries for Linux) from this site: https://www.rosettacommons.org/software/license-and-download.

3. Extract the tarball to a local directory from which Rosetta binaries can be called by specifying their full path.

4. Run the following fommand to refine the give protein structure xxxxx.pdb:

relax.static.linuxgccrelease -in:file:s XXXXX.pdb -relax:constrain_relax_to_start_coords -out:suffix _relaxed -out:no_nstruct_label -relax:ramp_constraints false

5. Run the following command to create a structural model for each of the variants in the given list:

rosetta_relax.py --rosetta-bin relax.static.linuxgccrelease -l VARIANT_LIST --base-dir /path/to/where/all/XXXXX_relaxed.pdb/is/store

BayeStab model was trained on experimental data of the unfolded Gibbs free energy change (ΔΔG) of 2648 mutations from 131 proteins.For the mutation (ΔΔGwt→mut), the 3D structure of the wild-type protein was obtained from the Protein Data Bank (PDB) ⁽¹⁾ .

Predicted result. The Gibbs free energy change of a single point mutation of the protein was predicted using the BayeSatb model, and the prediction result and uncertainty were obtained. We can also retrieve the model's noise as well as the dataset's noise using this way. The picture below represents the output of the BayeStab model.

More details can be found in our paper.