hebaebrahim
commented
5 years ago I read your paper "Deep learning-based drug-target interaction prediction". It make me interested to use deep belief nets to predict drug-target interaction. And I use tensorflow to implement deep belief nets, following your idea and code. But the nets' performance is worse. The strategy I build training set is:
- use approved drug-target interaction as positive sample, almost have 4500 samples.
- random select drug and target who not in approved data set as negative sample, number of negative sample is same as positive.
And I use same strategy to build test set, using experimental data. After train the nets, I use test set to evaluate nets, and I see almost drug-target records are predict as positive, only a few records predict as negative (maybe 200 of 3000). I read the code in file 'code/load_data.py', and the function 'load_labeled_data' is need a parameter for a pkl format file. But I can't find this file or convert data code in Github. So how to build a training set for this nets? Thanks.
please if you have the training data set for this paper
upload here for uses in my research
thanks
89-oniichan
ruru5697
I read your paper "Deep learning-based drug-target interaction prediction". It make me interested to use deep belief nets to predict drug-target interaction. And I use tensorflow to implement deep belief nets, following your idea and code. But the nets' performance is worse. The strategy I build training set is:
And I use same strategy to build test set, using experimental data. After train the nets, I use test set to evaluate nets, and I see almost drug-target records are predict as positive, only a few records predict as negative (maybe 200 of 3000). I read the code in file 'code/load_data.py', and the function 'load_labeled_data' is need a parameter for a pkl format file. But I can't find this file or convert data code in Github. So how to build a training set for this nets? Thanks.