The source code for our paper: OnionMHC: A deep learning model for peptide — HLA-A*02:01 binding predictions using both structure and sequence feature sets
- Tensorflow 2.0.0
- Pandas 0.25.3
- scikit-learn 0.21.3
- scipy 1.3.2
- Numpy 1.17.4
Structure based features can be generated by running generate_features_cam_o.py
. Each line in the input file <input.dat>
contains the path to the pdb structures.
python generate_features_cam_o.py -inp <input.dat> -out <output.csv>
Parallelizing generation of structure based features using mpirun
mpirun -np 16 python generate_features_cam_o.py -inp <input.dat> -out <output.csv>
- Clone this repository
git clone https://github.com/shikhar249/OnionMHC
- Run onionmhc.py
python onionmhc.py -struc <structure-based features file> -seq <sequence file> -mod path/to/models/fold{0..4}_model{0..2}_bls_lstm.h5 -out <output file>
The example of prediction results will be shown as:
peptide_sequences | OnionMHC_score | Binding_affinity(nM) |
---|---|---|
FLIAYQPLL | 0.901299 | 2.909335 |
NLLTTPKFT | 0.483376 | 267.669274 |
GTHSWEYWG | 0.062278 | 25487.509380 |
... | ... | ... |
Please cite our paper
@article{doi:10.1142/S2424913020500095,
author = {Saxena, Shikhar and Animesh, Sambhavi and Fullwood, Melissa J. and Mu, Yuguang},
title = {OnionMHC: A deep learning model for peptide — HLA-A*02:01 binding predictions using both structure and sequence feature sets},
journal = {Journal of Micromechanics and Molecular Physics},
volume = {05},
number = {03},
pages = {2050009},
year = {2020},
doi = {10.1142/S2424913020500095}