Colab notebook: Template used in ranking structures and Runs on multimers

[Chengeng-Yang]

Hi there,

Thanks a lot for making this repo available.

I just wanted to ask two questions I have when I tested the colab notebook.

1. Could you tell me is there a template taken when ranking decoy structures (e.g. the native structure)? I was a little bit confused when reading the set up code. Did the NATIVE PATH in the following cell indicate the template to be used?
   

2. When running multimers, even if I have changed model_mode into alphafold-multimer in the settings cell, I still got the following error: "ValueError: Only single model PDBs are supported. Found 0 models.". I also attached a screenshot below. Could you tell me if it's because the chain name only allows a string of length = 1?
   

[sokrypton]

1. We need something to initialize the model (and later to compute the TMscore against). Later the native structure within the model is replaced by the decoys.
2. CHAIN = "A,B"

[Chengeng-Yang]

Many thanks for your reply! I just have a few more concerns.

1. We need something to initialize the model (and later to compute the TMscore against). Later the native structure within the model is replaced by the decoys.

Thanks for the reply. I see, so that cell was for the initialization. I was just curious about which template it takes for prediction and ranking (but maybe it's concerned with MSA taken by alphafold, instead of AF2Rank)

2. CHAIN = "A,B"
   Thanks for the hint, now the model was successfully initialized. But when I run af.predict, I got another error called "ValueError: 'model_1_multimer_v2' is not in list". Could you tell me if there's a way to fix this, like adding an additional parameter in af.predict()?
   

Thanks for reading the message!

[sokrypton]

looks like multimer model got updated to v3. I've updated the notebook. Can you try again.

[Chengeng-Yang]

Yes, that works now! Many thanks for your kind help