search

BHSAI / BRILIA

B cell receptor repertoire annotation and lineage construction tool
7 stars 3 forks source link

Bug: Lineage tree clustering % off by factor 2 #12

Closed dleebhsai closed 7 years ago

dleebhsai commented 7 years ago

The lineage tree cut off distance uses a hamming distance %, but applied to BRILIA's internal hamming distance + shm-adjusted distance metric, which contains fractional 0.5 values. To reduce memory usage, I use integer matrices such as uint16, and I double the internal metric by 2 to ensure integer values. This was not being carried over to the clustering %, hence we're off by 2. Will be fixed AFTER v3.0.7, but workaround is to use a DevPerc (cluster cutoff deviation %) that is twice as larger than you want.

dleebhsai commented 7 years ago

Fixed in v3.0.8 so that the DevPerc is the correct hamming % distance.