Chimera Slayer Instability

Hi everyone,

We just finished a small MOTHUR tutorial with ~50 people using identical datasets on different platforms and probably different versions of MOTHUR as well. Almost everything went off without a hitch except for the chimera checking excercise. As an example, we used the chimera slayer with default settings on an identical dataset:

MOTHUR> chimera.slayer(fasta=…, template=…)

Interestingly, there were a lot of differing results in terms of how many, and which, sequences were flagged. So maybe some of these differences can be chalked up to stochasticity in the bootstrapping, but the degree to which results differed was a little startling. Perhaps this is one of those cases where default settings don’t hack it.

Any thoughts?

So there is some stochastiticity involved because there’s a bootstrapping procedure involved. However, one of the patches we just released changes the default from minsnp=10 to minsnp=100. The latter gives the better performance for our implementation. I suspect/hope that’s the source of the differences.