Greg Caporaso

"Lab Person" could be better named

6

I'm not really sure what's supposed to go in this field. Should this be something like "Study contact" or "Data depositor"?

rename _input_splitter to something more appropriate

1

since all of the input handlers don't split the input (e.g., alpha diversity)

centralize the functionality for building a blast database before starting a parallel run

1

this is currently implemented as _precommand_initiation in ParallelPickOtusBlast, ParallelAssignTaxonomyBlast, and probably ParallelBlast. Multiple inheritance might be in order here, including a new, very lightweight class (or maybe ParallelPickOtusBlast and ParallelAssignTaxonomyBlast...

generalize ParallelWrapper._get_poller_command so developers don't need to override to pass process_run_results_f, etc

1

currently user has to override this function (as in ParallelPickOtus), but we could come up with a clever way to get around that

uclust pre-sort by abundance should collapse sequences which are prefixes of other sequences

4

this would ensure (1) that abundances are not effected by differing quality across reads of the same sequence, and (2) that prefixes cluster to the same OTU as the sequence...

workflow scripts should compress large but infrequently used directories

2

Directories like rarified OTU tables, jackknifed distance matrices, etc (i.e., large directories that get created for record keeping but not frequently used) should be gzipped as part of the workflow...

Missing dependencies send an error vs. a warning

1

When we move to python 2.7 we should take advantage of conditional testing to put this in place.