AlphaFold2 Initial release tuned for multimer prediction (github/hegelab/betafold) were used for the prediction. The --full_dbs
option, which runs with all genetic databases used at CASP14, was set. Database versions were the same as in this AF2 release, except UniClust30, for which we used the 2021_06 release, since hhblits
produced error for some sequences when using uniclust30_2018_08
. The --max_template_date
argument was set into the future to utilize all available experimental structures as potential templates. The structure top ranked based on pLDDT score were selected and presented as a PDB formatted file (its name ends with full.pdb). Low complexity regions were visually inspected and removed if concluded non-sense (e.g. traversing membrane region) and presented as the main cif and pdb files.