Hi,
I am using FUpred and DeepMSA to generate a deep MSA and predict the domain boundaries. Many of the target sequences seemed to be genus/species-specific and do not have enough homologs. In one case, DeepMSA outputs a MSA of 13 similar sequences generated after hmmsearch with Nf >= 0.2, and FUpred predicted two domains 1-76 and 77-120, which I wonder would be reliable. I am running the pipeline for hundreds of sequences, and it's tricky to look at every single MSA myself. I wondered based on your experience, how much depth of the MSA you would recommend to have for reliable prediction of domain boundaries. Or in a similar sense, would there be any Nf cutoffs to filter out unreliable domain boundary predictions?