broadinstitute / catch / 425

94%

master: 94%

push

travis-ci-com

Add heuristic to decide which clustering method to use

The comment added by this commit to probe_designer details a problem
with the recently-introduced default method for clustering input
sequences ('simple'). The problem occurs on long genomes (e.g.,
bacterial) when also breaking the genomes into fragments and clustering
those.

This commit introduces a crude heuristic to select whether to use the
newer 'simple' method (which finds connected components) or the older,
most resource-intensive 'hierarchical' method (which uses agglomerative
hierarchical clustering). 'hierarchical' avoids the problem.

The argument --cluster-and-design-separately-method now has a new
choice, 'choose', to use the heuristic to decide the method, and
'choose' is the default setting for that argument.

1649 of 1908 branches covered (86.43%)

14 of 14 new or added lines in 1 file covered. (100.0%)

5725 of 6084 relevant lines covered (94.1%)

1.88 hits per line