search

X-lab-3D / PANDORA

MODELLER-based, anchor restrained, Peptide-MHC Modelling pipeline
Apache License 2.0
17 stars 5 forks source link

benchmarking #163

Closed v-mikhaylov closed 2 years ago

v-mikhaylov commented 2 years ago

Example 3 in the tutorial suggests to use case.model(benchmark=True) to benchmark on an existing pdb structure. I followed that example but couldn't find LRMSD anywhere in the output. It also looks like benchmarking is commented out in Modelling_functions.run_modeller. How to properly use this function?

DarioMarzella commented 2 years ago

Hi @v-mikhaylov , thanks for reporting this issue and I apologize for the inconvenience.

In Modelling_functions.run_modeller, the part commented out (from line 868 to line 883) refers only to the RMSD calculation. You can comment this part out and PANDORA will try to calculate the RMSD by itself, although this function is not robustly tested and might raise errors. Moreover, this function exploits pdb2sql and we noticed some results can differ quite a lot from the ones generated by ProFit when calculating the L-RMSD. We are currently working on a fix on the pdb2sql side as well. In case the L-RMSD calculation works, you will find your results in the .pkl file generated by PANDORA: every object in the list present in the pkl is a PANDORA PMHC.Model object, you will find its L-RMSD with .lrmsd

Instead, the actual place where the benchmark option is passed to prevent PANDORA from using the same structure as template when it's the target as well, is in Modelling_functions.find_template() .

Please let us know if you have further questions.

v-mikhaylov commented 2 years ago

Ah, I see. Thank you!