pengzhangzhi
commented
4 months ago Hey @Melissaurious ,
Sorry for the late reply! You can follow this notebook https://github.com/pengzhangzhi/ab_opt/blob/main/AbDock/ab_opt_analysis_4mutations.ipynb. It basically implements the antibody optimization pipeline, which takes a PDB file of three chains, Antigen, heavy and light, and outputs the CDR design specified by the contig. contig specifies the range of the CDR to be designed, set it None to design the whole CDR.
For example, in the following configs, you want to specify the pdb_path = 'wet_exp/Gn-Ab1C5-coot-19.pdb', design_contig = '7-11', and heavy_chain_id = 'B'.
pdb_path = 'wet_exp/Gn-Ab1C5-coot-19.pdb'
outdir = 'wet_exp/dock_single'
seq_design_dir = 'wet_exp/seq_design'
design_contig = '7-11'
screen_dir = 'wet_exp/screen'
heavy_chain_id = 'B'
num_gpus = 2
process_per_gpu = 10
Hello!
I used the training and inference scripts and configurations provided, however, when I analyze my results I see that even that the reference and the mutant structure almost overlap, the sequence of the whole structure changed and therefore, I cannot locate the CDRs automatically following the chothia numbering as the conserved positions are no longer there.
Do you have any suggestions? As I gave as input the config json files, I dont see how it could had been mixed up with the fixbb training mode.