search

dptech-corp / Uni-Dock

Uni-Dock: a GPU-accelerated molecular docking program
174 stars 37 forks source link

Seed doesn't seem to be working properly. #133

Open matija-marijan opened 1 month ago

matija-marijan commented 1 month ago

Describe the bug

Uni-Dock's results don't seem to be reproducible. When calling unidock for the same seed and for the same ligand, it often returns different docking scores.

Any ideas as to why this occurs, and how it should be addressed?

Expected behavior

Uni-Dock should produce identical docking scores for the same molecule.

To Reproduce

Call Uni-Dock several times for the same ligand, and the same seed.

I am calling Uni-Dock with: unidock --receptor {receptor} --ligand_index {l_index} --dir {exp_name} --center_x {center[0]} --center_y {center[1]} --center_z {center[2]} --size_x {size[0]} --size_y {size[1]} --size_z {size[2]} --scoring vina --search_mode balance --num_modes 1 --seed {seed} > {log_file}

Environment

WSL2 2.2.4.0 Ubuntu 22.04 Windows 11 Pro CUDA 12.3 NVIDIA RTX 3060

ysyecust commented 1 month ago

Hi @matija-marijan ，thank you for your reply, we have not found this in our experiments, can you try the next example of single digit ligands and observe if you still have the same problem？

matija-marijan commented 1 month ago

Hello @ysyecust, I'm not entirely sure what example you are referring to? Thanks.

matija-marijan commented 1 month ago

Hello @ysyecust, I followed the instructions in the README.md, and ran the screening test examples for the same config.json file and the same seed, and got completely different results.

I attached the two CSV results: result_1_def_vina_nt128_ns20_seed5_V100.csv result_2_def_vina_nt128_ns20_seed5_V100.csv

Do you have any information regarding this problem, and how to solve it? Thanks.