Question about eval_all.pkl

[ZJoYuan]

Hello! Thanks for your great work!
I only find the eval_all.pkl file for target pair 8-226, but I couldn't locate the eval_all.pkl files for other target pairs. Could you please provide information on how to obtain the eval_all.pkl files for all the pairs in the test set?

[flowerwjl]

Thank you for your valuable contribution @zhouxiangxin1998. I'm encountering the same issue originally reported by @ZJoYuan.

Hello @ZJoYuan, and thank you for raising this - would you be able to share where you located (or how to generate) the eval_all.pkl file for target pair 8-226?

Also, I'd appreciate if anyone could advise whether it's possible to obtain the complete set of eval_all.pkl files for all test set pairs.

[ZJoYuan]

Hello @flowerwjl ! Sorry, I also don’t know how to obtain eval_all.pkl yet.

[flowerwjl]

​​@ZJoYuan​ After carefully reviewing the code and paper, I think that I've figured out this issue! After further examination, I understand that the eval_all.pkl used in the alignment step actually represents the protein-ligand binding prior mentioned in the paper. Here's my current understanding:

1. The authors first employed TargetDiff to generate 10 molecular samples for each of the 438 unique binding pockets.
2. These sampled molecules were then evaluated against both target pockets in the dual-target dataset, resulting in the creation of the two eval_all.pkl files visible in your screenshot, and produced the corresponding docking conformers for each molecule-pocket pair. (the 'docked_mol')

3. From these evaluations, the molecule demonstrating the optimal combined Vina score across both pockets was selected as the probe/anchor molecule. Then calculated the transform matrix according to the two docked_mol.
4. This anchor molecule subsequently serves as the reference for aligning the two binding pockets and preparing the system for the dual-target sampling algorithm.

@zhouxiangxin1998​ Dear author, I'd be grateful if you could verify whether this understanding is correct. Please feel free to correct any misinterpretations I may have made. Thank you for your time and for this valuable work!