mlmip / piano

PIANO

Graph masked self-distillation learning for prediction of mutation impact on protein-protein interactions.

This source code is tested with Python3.8 on Ubuntu20.04.

Step 1: Clone the GitHub repository

git clone https://github.com/MLMIP/PIANO.git
cd PIANO

Step 2: Build required dependencies

Before installing the relevant dependency packages, please make sure Anaconda3 has been installed. If not, please click here to install it.

source install.sh

Executing the above command will automatically install the Anaconda virtual environment. Upon completion, a virtual environment named "piano" will be created. In addition, the model weights can be obtained through Zenodo-PIANO. After downloading, place it in piano/Data/model_params.

Step 3: Download required software

The download links for various software packages are provided below. After downloading, you can install it directly according to the official tutorial.

PSI-BLAST

PSI-BLAST Database:Uniref90

HHblits

HHblits Database:Uniref30

NACCESS

MSMS

Please install these software in the piano/Software. In addition, mkdssp needs to be granted certain permissions, which can be operated by executing the following command：

chmod a+x mkdssp

Step 4: Running PIANO

Activate the installed piano virtual environment and ensure that the current working directory is PIANO.

conda activate piano

If you want to obtain the prediction results of a single sample, directly execute the following command on the command line to obtain the prediction results, and the results are saved in the PredictedResults.txt:

python run.py 0 [pdb name] [mut_chain] [wildtype] [mutant] [resid] [partnerA_partnerB]

where the digit 0 signifies that the program performs predictions for individual samples only. [pdb name] is the name of the complex to be predicted, such as 1a4y. [mut_chain] is the name of the mutated chain. [wildtype], [mutant], and [resid] are wild-type amino acid, mutant amino acid, and the mutation position, respectively. [partnerA_partnerB] describes the two interaction partners in the protein complex, such as A_B.

A specific example is:

python run.py 0 1a4y A E A 401 A_B

If you want to perform batch prediction of multiple samples, please first organize the relevant mutation sample data in pred_data.csv. The input format is consistent with the data format within an individual sample. After the completion of filling in pred_data.csv, execute the following command in the command line to obtain the prediction results. The results are saved in the PredictedResults.txt:

python run.py 1

where the digit 1 indicates that the program conducts predictions for multiple samples.