This is linked to #15 as it leads to production of ambiguous prgs (multiple paths give rise to same sequence)
Clustering code can conclude that there is no meaningful clustering of a set of sequences- e.g. puts each sequence in one cluster.
However the code that calls clustering can re-run prg-building for sets of sequences that only differ in alignment (i.e. gap positioning), not sequence (here)
This leads to spurious 'nested variants' and the following pathological example case:
msp6_ambig.pdf
Of the 4 paths between nodes labeled 55 and 56, two are identical. This cause gramtools to mis-genotype down the line.
I have a fix that I'm implementing and will PR in
This is linked to #15 as it leads to production of ambiguous prgs (multiple paths give rise to same sequence)
Clustering code can conclude that there is no meaningful clustering of a set of sequences- e.g. puts each sequence in one cluster.
However the code that calls clustering can re-run prg-building for sets of sequences that only differ in alignment (i.e. gap positioning), not sequence (here)
This leads to spurious 'nested variants' and the following pathological example case:
msp6_ambig.pdf
Of the 4 paths between nodes labeled 55 and 56, two are identical. This cause gramtools to mis-genotype down the line.
I have a fix that I'm implementing and will PR in