lh3
commented
1 year ago It is difficult to filter when you look at a single protein. Miniprot is intended for aligning similar proteins (say, identity >50%). The E-value is usually small. Alignment score is roughly proportional to the protein length. We can't easily set a cut off on that score. Mapping quality evaluates whether an alignment is unique. It is not that useful for protein alignment as we expect multi-gene families.
A protein to DNA alignment with translated BLAST can be filtered by e-value so the output only contains matches above a specified e-value. miniprot has an alignment score (sort of similar to Smith-Waterman) and a Mapping score (sort of similar to Bowtie). Can you add a command line parameter to filter output by a minimum quality so the gff can be used directly, for example as a JBrowse track.