Error in vina score

[tom832]

Describe the bug

I used latest released unidock on my 4090 gpu, and it successfully worked, but has an abnormal result in vina score about -100 in certain docking results (most results are normal).

MODEL 1
REMARK VINA RESULT:  -108.179      0.000      0.000

Expected behavior

using autodock vina, I can get normal results.

To Reproduce

Grid center: X -26.346 Y -25.7025 Z 31.2255
Grid size : X 30 Y 30 Z 30
Grid space : 0.375

receptor pdbqt file: receptor.txt

ligand pdbqt file: ligand.txt

my unidock output pdbqt file: ligand_unidock_vina_out.txt

using autodock get a normal output file: ligand_autodock_vina_out.txt

Environment

• CUDA 12.3
• unidock 1.1.0

Additional Context

No response

[ysyecust]

Hi @tom832 Thank you for your feedback. Can you try such an instruction unidock --receptor ./receptor.pdbqt --ligand_index ./index.txt --center_x -26.346 --center_y -25.7025 --center_z 31.2255 --size_x 30 --size_y 30 --size_z 30 --dir ./result/ --search_mode detail --max_step 20 --num_modes 12 --scoring vina --refine_step 5 --seed 5 --verbosity 1 the index.txt is the absolute path of ligand. My output like this :
mode | affinity | dist from best mode
| (kcal/mol) | rmsd l.b.| rmsd u.b.
-----+------------+----------+----------
1 -1.549 0 0
2 -1.373 4.499 7.039
3 0.2898 12.91 15.65
4 3.319 5.264 7.722
5 6.576 6.822 10.21
6 8.77 10.05 14.17
7 29.93 6.42 9.252
8 41.2 8.23 10.92
9 48 10.39 12.86
10 208.4 15.39 17.4
11 219.3 18.63 21.18
12 3.048e+06 13.9 17.47

[tom832]

Hi @tom832 Thank you for your feedback. Can you try such an instruction unidock --receptor ./receptor.pdbqt --ligand_index ./index.txt --center_x -26.346 --center_y -25.7025 --center_z 31.2255 --size_x 30 --size_y 30 --size_z 30 --dir ./result/ --search_mode detail --max_step 20 --num_modes 12 --scoring vina --refine_step 5 --seed 5 --verbosity 1 the index.txt is the absolute path of ligand. My output like this : mode | affinity | dist from best mode | (kcal/mol) | rmsd l.b.| rmsd u.b. -----+------------+----------+---------- 1 -1.549 0 0 2 -1.373 4.499 7.039 3 0.2898 12.91 15.65 4 3.319 5.264 7.722 5 6.576 6.822 10.21 6 8.77 10.05 14.17 7 29.93 6.42 9.252 8 41.2 8.23 10.92 9 48 10.39 12.86 10 208.4 15.39 17.4 11 219.3 18.63 21.18 12 3.048e+06 13.9 17.47

THX for your reply!
I reproduce your result, which is also strange.

1. Why there is difference between mine and yours due to different input?
   my cmd is like this:

unidock --receptor {receptor_path} \
    --ligand_index unidock_ligands_tmp.list \
    --search_mode {search_mode} \
    --scoring vina \
    --center_x {box['center_x']} \
    --center_y {box['center_y']} \
    --center_z {box['center_z']} \
    --size_x {box['size_x']} \
    --size_y {box['size_y']} \
    --size_z {box['size_z']} \
    --num_modes {save_n_poses} \
    --dir {output_dir} \
    --cpu {n_cpu}

where search_mode I use is detail, num_modes is 10, n_cpu is 0.

2. I use autodock vina with the same receptor and ligand, the vina score is about -6 and get 10 poses around -6/-5.

BTW, the unidock I used is unidock-1.1.0-cuda120-linux-x86_64 in the released page.

[andressanchezr4]

hey @tom832 I think I just ran on the same problem: for some reason when using unidock with macrocycles (special atom types CG0, etc) the variability of the score between different runs is pretty high. In my case I got values for the same protein-ligand that went from -0.07 to -400.

If you prepare this ligand without the --rigid_macrocycles flag, and repeat the docking you will most likely get realistic and reproducible results.