SAAMBE-3D: Predicting Effect of Mutations on Protein–Protein Interactions.

About

SAAMBE-3D is a tool written in python 3, it is tested to work with python3.7. SAAMBE-3D implements a structure based mathod for predicting the binding free energy change for protein-protein due to a point-mutation. It requires the 3D structure in PDB of the protein-protein complex as input and information about the mutation to make the prediction.

SAMMBE 3D can be used to :

• i) predict the binding free energy change caused by single mutation, and
• ii) predict whether a particular mutation is disruptive or non-disruptive.

Dependepcies

SAAMBE-3D requires following packages and mentioned versions to be installed in the python (Python 3.11.5) to work

• numpy (1.26.0)
• prody (2.4.0)
• xgboost (1.7.6)

It has been tested to work on a machine running Ubunutu 22.04.3. The python is mainted using conda 23.7.2. We created a python environment for SAAMBE-3D using:

conda create -n py311_saambe3d python=3.11

Then if any conda environment is already active deactivate it using:

conda deactivate

Activate the conda python environemnt created above using:

conda activate py311_saambe3d

All the above versions of packages are installed using following commands.

conda install -c conda-forge numpy=1.26.0
conda install -c conda-forge prody=2.4.0
conda install -c conda-forge xgboost=1.7.6

downgrade the pyparsing to 3.1.1 if higher version installed, as it leads to error in selection parsing using prody. One can use below command for it.

conda install -c conda-forge pyparsing==3.1.1

Alternatively, create python environment from environment.yaml.

Alternatively one can create the python enviroment py311_saambe3d from the environment.yaml file using command:

conda env create --file=environment.yml

Prediction Models

The method provides two different models for predictions:

• regression.model (for predicting: binding free energy change)
• classification.model (for prediction: if the mutation is disruptive or non-disruptive)

How to run

To have a look at and familiarize yourself with all the options supported use:

python saambe-3d.py --help

The SAAMBE-3D can be run for a single point mutation by executing following command from inside of the SAAMBE-3D directory:

python saambe-3d.py -i PDBfile -c Chain -r Resid -w wild -m mutation -d model

To make predictionsof a list of point-mutations execute:

saambe-3d.py -i PDBfile -f mutation_list -d model

Example

Single point-mutation:

For example if user want to predict binding free energy for protein-protein complex with PDB ID 1A22 due to mutation from Cysteine (C) to Alanine (A) at res ID 182 in chain A, they should run the script by typing

python saambe-3d.py -i 1A22.pdb -c A -r 182 -w C -m A -d 1

After running it check the file output.out or the file name if you provided any with sammbe-3d.py -o option for the output.

List of point-mutations.

If user want to get multiple predictions for many single mutations at the same complex, user can provide a file say mutations_list.txt, which should be formatted as follows for the above example. The columns in the mutation list file are: ChainID resID wildtype_residue mutant_residue

A 182 C A   

and the command will be:

python saambe-3d.py -i 1A22.pdb -f mutations_list.txt -d 1

After running it check the file output.out or the file name if you provided any with sammbe-3d.py -o option for the output.

# Similarly. for predicting disruptive/non-disruptive mutation, user can type the same command, just need to change the model from 1 to 0

python saambe-3d.py -i 1A22.pdb -c A -r 182 -w C -m A -d 0

After running it check the file output.out or the file name if you provided any with sammbe-3d.py -o option for the output.

OR

python saambe-3d.py -i 1A22.pdb -f mutations_list.txt -d 0

After running it check the file output.out or the file name if you provided any with sammbe-3d.py -o option for the output. The 'mutations_list.txt' will be exactly same.

Citation

Pahari, S., Li, G., Murthy, A. K., Liang, S., Fragoza, R., Yu, H., & Alexov, E. (2020). SAAMBE-3D: predicting effect of mutations on protein–protein interactions. International journal of molecular sciences, 21(7), 2563.