ResaPred is a two-state (non-strict and strict) flexibility prediction tool with high accuracy. ResaPred combines a variety of features related to flexibility, such as secondary structure, torsion angle, solvent accessibility, etc. ResaPred is a novel deep network based on a modified 1D residual module and self-attention mechanism, which effectively extracts deep key features related to flexibility. The modified 1D residual module consists of three convolution layers, with batchnorm and relu layers added after each layer to prevent gradient explosion or vanishing. The self- attention module facilitates capturing long-range intra or inter-slice dependencies which are often overlooked by convolution layers.
Before installing and running ResaPred, ensure your environment meets the following requirements:
- Python 3.7.0 or higher
- PyTorch 1.8.0 or higher
- Additionally, you need to install the biopython package:
pip install biopythonTo predict the flexibility of a protein, simply run the run_model.sh script:
bash run_model.shIf the running process fails, after verifying the environment setup, you can manually execute the corresponding Python scripts as described in run_model.sh.
We have provided data for two example proteins. You can directly run the tool to predict their flexibility. If you wish to predict the flexibility of other proteins, please prepare the following files:
- FASTA file of the protein sequence
- PDB file of the protein structure
- PSSM file generated by PSI-BLAST
- DSSP file generated by DSSP
- Please place these files in the corresponding folders (main, pdb, pssm, dssp), respectively, then follow the previous instructions to run the tool and predict protein flexibility.