Step 3: fine-tuning domain boundaries based on similarity to other proteins.
Whole targets and suggested domains from Step 2 were run through PSI-BLAST 16 and HHblits 17/HHsearch18 programs to establish sequence-based similarity to proteins of known fold: