search

bixiangpeng / HiSIF-DTA

8 stars 5 forks source link

question about contact map #1

Open HeJunhong1107 opened 5 months ago

HeJunhong1107 commented 5 months ago

While I was reviewing the code, I noticed that the contact_map seems to be a requisite for the input data of model. Does this imply that if I only have the sequence of a protein but not its actual three-dimensional structure, it would not be able to use HiSiF for inference?

bixiangpeng commented 5 months ago

While I was reviewing the code, I noticed that the contact_map seems to be a requisite for the input data of model. Does this imply that if I only have the sequence of a protein but not its actual three-dimensional structure, it would not be able to use HiSiF for inference?

Hello Junhong! Thank you very much for your interest in our article! The contact map is indeed a prerequisite for HiSIF-DTA. If you want to utilize HiSIF-DTA in the absence of actual protein 3D structures or AlphaFold predicted structures, it is advisable to input protein sequences into the ESM-2 (a powerful protein language model ) to predict their contact map. In preliminary experiments, we found that utilizing such predicted contact maps also yields good performance. We hope this information proves helpful to you.

cengyangjie commented 3 months ago

Dear Author,

I am very interested in your model. I would like to use it to test some new data, but I am not sure how to generate the ppi_data.pkl file, even though you provided a description of this data in your paper.