Open taprs opened 1 month ago
thanks for this suggestion. This is a feature that was part of an earlier GENESPACE release (to allow parallelization within blast runs, instead of only among). I removed it at v1 to reduce the chances for compute architecture conflict. You can definitely run orthofinder separately, and if you set -op, then alter the number of threads in those blast commands and run them separately. This would save substantial time when running a couple very large genomes, but will have little savings once the number of blast jobs significantly exceeds the number of cores.
I noticed that on my HPC the
diamond
step oforthofinder
takes forever for the octoploid species with ~100k gene models being BLASTed against itself.My workaround was (1) manually running the exact
diamond
command in the orthofinderWorkingDirectory
providing more--threads
to it and then (2) killing the originaldiamond
processes. Then, everything gets through very quickly andorthofinder
does not notice the interception :)But I think there might be a nicer way to provide more threads per
diamond
run. I haven't found one neither inGENESPACE
nor inorthofinder
— should I go down toorthofinder
issues then?