Relabeling candidate m6As with multiple instance learning.
The misvm package should be installed at first, which can be downloaded from https://github.com/garydoranjr/misvm.
Usage: python {0} positive_dataset negative_dataset method model_file scale_file score_file reserved_file
This script is used to train a bag level miSVM/MISVM model.
method : method of selection
0 -- miSVM
1 -- MISVM
A bag index file 'bag_index'will be created in current path.
A feature encoding file 'train_features.txt' will be created in current path.
A model file 'svc.pkl' will be created in current path.
A normalized file 'scale.pkl' will be created in current path.
A prediction score file 'score.txt' will be created in current path.
A reserved samples file 'reserved.txt' will be created in current path.
Example: python m6A_model_training.py ./traindata/train_pos_example.txt ./traindata/train_neg_example.txt 1 svc.pkl scale.pkl score.txt reserved_samples.txt
Usage: python {0} dataset model_file scale_file predict_res
This script will generate scores to relabel the candidate m6A sites.
Example: python m6A_model_prediction.py ./traindata/test_example.txt ./svc.pkl ./scale.pkl predict_res