search

xypan1232 / IPMiner

Predicting ncRNA-protein interaction using high-level features based on deep learning
15 stars 8 forks source link

Scores #1

Open agolicz opened 7 years ago

agolicz commented 7 years ago

Hello, I have two questions.

  1. What score is considered a putative interaction. Using RPI488 for training I seem to be getting a lot of scores close to 1. I'm using a plant species.
  2. Is it possible to train on NPInter as opposed to RPI488 for new interaction prediction? I've noticed I can do that by changing the source: def predict_new_samples(RNA_file, protein_file): train, trainlabel = get_data_deepmind('NPInter', seperate = True) But I'm not sure how safe that is.

Agnieszka

xypan1232 commented 7 years ago

Hi,

  1. i do not test plant species, but i think you need retrain the dataset that maybe related to your data
  2. Yes, you can do that, but i did not test it.