search

huankoh / PSICHIC

PSICHIC (pronounced Psychic) - PhySIcoCHemICal graph neural network for learning protein-ligand interaction fingerprints from sequence data
Apache License 2.0
68 stars 9 forks source link

screening mode #3

Open CLG68 opened 2 months ago

CLG68 commented 2 months ago

Hi,

The screening script uses a CSV file containing in each row the sequence of the protein. However, it is not very efficient when you want to screen let say one protein against 1 million smiles. Moreover, if you want to screen the same library against many different proteins, a huge file is going to be generated for each target as the protein sequence will be different. One target vs a library of compounds is probably the setup that people mostly uses for screening. Is there a way to separate the protein from the library, ex by adding the protein sequence in a fasta file or in an argument instead of merging the protein seq to the library?

Is there a way to preprocess the protein info instead of assuming the protein sequence is going to always be different at each matching smiles? I'm not sure if this is possible to preprocess the protein info however if it is, it would probably make your script much faster.

Best, Christian

huankoh commented 1 month ago

Hi Christian,

Thanks for the question and the suggestion. As you pointed out, it is true that the current code is not efficiently handling the setup and data loading process.

Yes, it is possible to preprocess the protein informationI will update the code over the next two weeks to have the option of separating protein from the library. In the meantime, I will keep this issue open and update you once this is option is update to the code.

Cheers, Huan