I have installed the MLST plugin with a scheme using webservices, and some of my stop trimming patterns are very degenerate. A colleague that installed it without webservices, has stop patterns that are perfectly fine. What is causing this difference?
The procedure to generate trimming patterns can give different results depending on whether webservices are used or not.
When using webservices, the patterns are generated on pubMLST. The procedure used for this is creating a consensus of the last 20 positions of the locus size for all alleles. In case of loci with variable allele length, this can result in very degenerate trimming positions. If the length of the locus is for instance 465 bp, the positions from 445 to 465 are considered for the stop trimming. Any allele that is not full length, will not contribute 20 bases to this consensus, generally leading to all Ns at the positions that are not present in all alleles.
The procedure used by BioNumerics when no webservices are used, is taking the consensus of the last 20 bases for all alleles. This leads to less degenerate stop patterns in most cases, so a first solution to degenerate stop patterns should be to update the trimming patterns after turning off the webservices and updating the alleles. The webservices can be turned on again after this update, but the automatic update of the trimming patterns should be turned off.