search

google-deepmind / alphafold

Open source code for AlphaFold 2.
Apache License 2.0
12.75k stars 2.26k forks source link

Unstable iptm+ptm #839

Closed cfz1998 closed 1 year ago

cfz1998 commented 1 year ago

Hi! @tomhennigan

https://github.com/HommaF/AlphaFold2_chapter/issues/1

FelixHomma commented 1 year ago

Hi! @tomhennigan

https://github.com/HommaF/AlphaFold2_chapter/issues/1

Just responded in the linked repo where you raised the question originally.

I think the issue was that when modeling a protease-inhibitor complex, you hadn't removed the prodomain of the protease. I assume you will be able to see the prodomain occupying the binding groove in the output PDB files. Let me know if the complex between the mature protease (Pip1) and the inhibitor (EpiC2B) gave higher scores. For more details and the sequences, see the linked repo.

cfz1998 commented 1 year ago

Thank you very much! @FelixHomma After removing the prodomain of the protease, I got stable iptm+ptm values.

"iptm+ptm": {
    "model_1_multimer_v3_pred_0": 0.9130940645044535,
    "model_2_multimer_v3_pred_0": 0.9080438523780853,
    "model_3_multimer_v3_pred_0": 0.9171108252434494,
    "model_4_multimer_v3_pred_0": 0.9183709633181361,
    "model_5_multimer_v3_pred_0": 0.9209148930632896