MHC-I epitope presentation prediction based on transfer learning
Software install:
Download the zip file and uncompress it, you are good to go. However, make sure the following dependencies are installed before you use Pluto:
- pandas
- numpy
- tensorflow
Here are some illustration about how to used the scripts contained in this repository:
pretrain.py is for pretraining a model to recognize common features of the epitopes, which will be used to train the pluto model.
Usage:
python pretrain.py [training dataset] [development dataset]
Example:
python pretrain.py data/pretrain/train_peptide.csv data/pretrain/dev_peptide.csv
pluto.py is for training the final allele specific epitope presentation prediction model.
Usage:
python pluto.py [HLA] [training dataset] [developent dataset]
Example:
python pluto.py A0101 data/A0101/train_peptide.csv data/A0101/dev_peptide.csv
predict.py is for predicting epitope presentation for specific allele by using the already trained model.
Usage:
python predict.py [checkpoint dir] [input file] [output file]
Example:
if you want to predict a list of peptide will be presented by HLA-A0101, you can use:
python predict.py model/A0101/pretrain/ data/A0101/dev_peptide.csv output.csv