Profile-profile alignment does not preserve columns of input profiles #62
Description
I’m using profile-profile alignment to add sequences to an existing reference MSA, one query at a time. I expected reference alignment columns to be preserved, but observed that the aligned reference sequences differ depending on the query sequence being added. This makes it impossible to stitch multiple independent “ref + one query” profile alignments into a single consistent MSA.
FAMSA: 2.4.1
OS: Linux (Ubuntu)
Expected behavior:
I expected that align_profiles(ref, query) would preserve the internal column structure of the reference profile (i.e., the relative alignment of reference residues should remain fixed), while allowing insertion columns to be introduced as needed to accommodate the query.
This expectation is based on the traditional definition of profile-profile alignment which states that input MSA columns are kept intact, with only gap columns inserted.
Observed behavior
The relative alignment of the reference sequences differs between runs. Specifically, the aligned reference strings change in a way that is not explainable solely by inserting additional all-gap columns into the reference block: gap patterns within the reference alignment shift depending on which query is added.
./famsa Proto.aln.fa query1.fa out1.fa
./famsa Proto.aln.fa query2.fa out2.fa