search

THUNLP-MT / dyMEAN

This repo contains the codes for our paper "End-to-End Full-Atom Antibody Design"
https://arxiv.org/abs/2302.00203
MIT License
89 stars 8 forks source link

Reproducing dyMEAN model in Fig. 4 #12

Open bud-graziano opened 1 year ago

bud-graziano commented 1 year ago

Hi, great work!

I was trying to reproduce the dyMEAN model for PDB-ID 1ic7 as shown in Figure 4.

However, the models I get do not seem to be positioned as well as shown in the Figure. See the image below with green = ground truth and pink = dyMEAN model 1ic7_test

I was wondering if you would mind sharing the settings to reproduce the model shown in the Figure? I tried to follow the examples from the README but possibly I made a mistake there. I used the api/design.py script with the following parameters:

{'model': {'checkpoint': './checkpoints/cdrh3_design.ckpt'},
 'Interface': {'pdb': './test/1ic7.pdb',
  'receptor': 'Y',
  'ligand': 'H',
  'k': 48,
  'out': './test/epitope_1ic7.json'},
 'Design': {'root_pdb_dir': './test',
  'pdbs': '1ic7.pdb',
  'epitope_defs': './test/epitope_1ic7.json',
  'heavy_chain': 'DVQLQESGPSLVKPSQTLSLTCSVTGDSITSAYWSWIRKFPGNRLEYMGYVSYSGSTYYNPSLKSRISITRDTSKNQYYLDLNSVTTEDTATYYCANWAGDYWGQGTLVTVSAA',
  'light_chain': 'DIVLTQSPATLSVTPGNSVSLSCRASQSIGNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTLSINSVETEDFGMYFCQQSNSWPYTFGGGTKLEIK',
  'out': './test',
  'suffix': 'antibody',
  'remove_chains': 'HL',
  'enable_openmm_relax': True,
  'auto_detect_cdrs': True}}
kxz18 commented 1 year ago

Hi, thanks for your interest in our work. Since there is randomness in the generating process, the results might be different with different random seeds. To reproduce the results in the figure, you can run the testing script which fixes the random seed GPU=0 bash scripts/test/test.sh ./checkpoints/cdrh3_design.ckpt ./all_data/RAbD/test.json ./results and obtain the results for 1ic7.