Too many None for scoring

Good afternoon,

Thank you for the amazing Haddock3. I used it for scoring already docked models. But, there is an issue that I faced when I ran it on a batch of pdb files using pdb_ensemble.py. I used “mdscoring” for scoring but still I got None instead of scores for too many complexes. Below is a part of the results that I got:

mdscoring_1283.pdb output_ensemble_2_982_haddock.pdb None -52.42295
mdscoring_1268.pdb output_ensemble_2_969_haddock.pdb None -51.466750000000005
mdscoring_777.pdb output_ensemble_2_526_haddock.pdb None -4.9204022599999995
mdscoring_1202.pdb output_ensemble_2_909_haddock.pdb None 11.91816
mdscoring_643.pdb output_ensemble_2_405_haddock.pdb None 885.8238799999999
mdscoring_626.pdb output_ensemble_2_40_haddock.pdb None 2699836.4326219996
mdscoring_95.pdb output_ensemble_2_1083_haddock.pdb None None
mdscoring_96.pdb output_ensemble_2_1084_haddock.pdb None None
mdscoring_97.pdb output_ensemble_2_1085_haddock.pdb None None
mdscoring_98.pdb output_ensemble_2_1086_haddock.pdb None None
mdscoring_99.pdb output_ensemble_2_1087_haddock.pdb None None
mdscoring_100.pdb output_ensemble_2_1088_haddock.pdb None None
mdscoring_101.pdb output_ensemble_2_1089_haddock.pdb None None

how can I solve it?
Another concern that I have is that scores have a large range. for example, in above for one complex the score is -51.4667 and for the other complex the score is 2699836.4326.
So I am suspicious of the validity of those score.

Thank you so much for your help and taking time in advance.

Probably a lot of clashes in your models…

You can add to your workflow a caprieval step which will output a file with all components of the score

PS: does this also happen when using emscoring?

Fortunately emscoring works.
Now I have score for all docking models.
Thank you.

But, because I want to compare the performance of scoring tools including HADDOCK, is it fair comparison if I use emscoring instead of dmscoring?

We typically use emscoring for our CAPRI scoring participation.

The score remains the same, the optimisation protocol is different between em and mdscoring.

But as you must have bad clashes the MD refinement part must be crashing for some complexes.

And of course you can compare both :slight_smile: