Protein Interaction Prediction

This project aims to predict intra-protein interactions using amino acid sequences. The model uses an LSTM neural network implemented in PyTorch.

Setup

1. Clone the repository:

   git clone https://github.com/Bibhuprasadbehera/PPI-predictor-.git
   cd protein_interaction_prediction

2. Install dependencies:

   pip install -r requirements.txt

3. Check if CUDA is available:

   import torch
   print(torch.cuda.is_available())

4. Prepare your data in the data/ directory.

5. Adjust the configuration file config.yaml as needed.

Protein-Protein Interaction (PPI) Prediction Workflow

1. Prepare Data

   • Ensure all required CSV files are in the data/ directory
   • Verify config.yaml has correct paths and parameters

2. Train Model

   python src/train.py

   • Monitor training progress
   • Note the epoch with best validation performance

3. Evaluate Model

   python src/evaluate.py

   • Record MSE and R2 scores
   • Compare with baseline or previous versions

4. Make Predictions

   python src/predict.py --model checkpoints/model_epoch_20.pth --sequence NKVQMHRSEMRPKFFSEHIISILNPHCVV --config config.yaml

   or use

   python src/predict.py --model checkpoints/model_epoch_20.pth --sequence NKVQMHRSEMRPKFFSEHIISILNPHCVV --secondary_structure HHHHHCCCCCCEEEEEECCCCCC --config config.yaml

   • Use for individual sequences or batch processing

5. Analyze Results

   • Compare predictions with known interactions
   • Assess model generalization on new data

6. Iterate and Improve (if needed)

   • Adjust hyperparameters in config.yaml
   • Modify model architecture in src/model.py
   • Collect additional training data

7. Deploy Model

   • Integrate into larger bioinformatics pipeline
   • Create user interface for easy access

Testing

Run unit tests:

python -m unittest discover tests

For more detailed information on each step, refer to the documentation in the docs/ directory.

if facing environment related issue please use this

/mnt/myssd/anaconda3/envs/ml/bin/python "/home/bibhuprasad/Documents/PPI prediction model/PPI-predictor-/src/train.py"

/mnt/myssd/anaconda3/envs/ml/bin/python "/home/bibhuprasad/Documents/PPI prediction model/PPI-predictor-/src/evaluate.py"

/mnt/myssd/anaconda3/envs/ml/bin/python /home/bibhuprasad/Documents/PPI\ prediction\ model/PPI-predictor-/src/predict.py --model checkpoints/model_epoch_20.pth --sequence NKVQMHRSEMRPKFFSEHIISILNPHCVV --config config.yaml

first run the rsa_ss_merger

then do the modify 1 by adding a test interaction score