PhosVarDeep: a deep-learning model for phospho-variant prediction using sequence information
Liu Xia from Health Informatics Lab, School of Information Science and Technology, University of Science and Technology of China
keras==2.0.0
numpy>=1.8.0
backend==tensorflow
testdata.csv
The input file is an csv file, which includes proteinName, postion, reference sequences，mutation and labels.
ori_var_input.py helps to get variant sequences, and get testdata_ori.csv, testdata_var.csv, which separately includes reference sequences and corresponding variant sequences.
If you want to use the model to predict your test data, you must prepare the test data (testdata_ori.csv, testdata_var.csv) as an csv file, the first column is protein Name, the second col: position, the third col: sequences
The you can run the predict.py
You can change the corresponding parameters in main function prdict.py to choose to use the model to predict for S/T or Y sites
If you want to train your own network, your input file are four csv files (positive reference sequences and corresponding variant sequences; negative reference sequences and corresponding variant sequences), while separately contains 4 columns: label, protein Name, position, sequence
You can change the corresponding parameters in main function train.py to choose to use the model to predict for S/T or Y sites
Please feel free to contact us if you need any help: lxlovetf@mail.ustc.edu.cn