search

hychim / molpc2

MoLPC2
1 stars 0 forks source link

Issue with Predicting Large Protein Structures Using molpc2 #1

Open zhangjuan233 opened 1 week ago

zhangjuan233 commented 1 week ago

Hello, I am encountering an issue when using molpc2 to predict the structure of large proteins with known stoichiometry information. For example, when I try to predict the structure of 1A8R (or 5BSE), which has 10 identical chains, the following command: ./pipeline2.sh -f /data1/zsm/molpc2-main/data/1A8R.fasta -o /data1/zsm/molpc2-main/data/res/sto -m 3 -d /data2/database -i A10 results in a structure with only eight chains. I have noticed that during assembly using MCTS, it displays an "early stop" message. Could you please advise on how to resolve this issue? Thank you. 1A8R

hychim commented 1 week ago

Thank you for your question. The early stop means that during the MCTS assembly, it can't find any spot to add a new chain, which is why it stops at 8 chains. One thing you can try is to increase the number of simulation steps in MCTS to enhance the search depth. You may need to clone the repository again for that, as I recently found a minor bug related to the simulation steps. Another thing you could try is to change the overlap threshold in MCTS (see this line).

However, if you know the stoichiometry of your structure, I would recommend using other tools (e.g., AlphaFold-Multimer or CombFold) for predictions, as these tools generally show better performance in predicting protein assembly.