A starting alignment was built if other sequences with a known PDB structure were available, by performing a GH19 domain structure-based alignment generated through the mmaker command implemented in ChimeraX. Other seed sequences in the same superfamily were added to this fixed structural alignment by the use of “--add” flag option available in MAFFT. If no other structures were available rather than the reference, a sequence-based alignment with other seeds was created with MAFFT “L-INS-i” strategy, improved by adding information of up to 600 close homologues obtained from a search in Uniprot non-redundant Uniref50 database using a restrictive E-value threshold of 10-20.
The obtained alignments were manually cut with respect to the length of the GH19 domain of the reference structure and used to generate the new superfamily-specific profile HMMs.