patrickbryant1 / Umol

Protein-ligand structure prediction
164 stars 16 forks source link

How to output more predicted poses #20

Open Zehui7Zhou opened 3 months ago

Zehui7Zhou commented 3 months ago

Hi Patrick, thank you for sharing the Colab version of Umol. It works well and is friendly to Python freshmen. I’m just wondering if Umol can output more predicted poses (e.g., the top 10 poses), which would be more thorough for the next steps of MD filtering and SAR study.

patrickbryant1 commented 3 months ago

Hi, Why would you like more poses? Is it not better to have one correct than many wrong ones?

You can rerun Umol many times and edit the config to activate dropout which should generate alternative predictions - this is something you have to do on your own, however.

Hope this helps!

Patrick

Zehui7Zhou commented 3 months ago

Thanks for your suggestions. I want to explore more ligand binding poses for two reasons. First, to determine whether Umol can identify metastable binding poses. Second, to test if Umol can predict the correct poses when a protein can bind to two molecules of the same ligand.