Closed cflerin closed 4 years ago
For the pyscenic ctx
step, the results are already deterministic, but when csv
is used as an export format (as is the case here), the context
column is sorted differently based on which module type was calculated first. For instance:
frozenset({'activating', 'top50', 'genome-ranking'})
vs:
frozenset({'genome-ranking', 'top50', 'activating'})
If the regulons are exported in gmt
format, the file hashes will match exactly.
For pyscenic aucell
, https://github.com/aertslab/pySCENIC/pull/143 adds the ability to use a fixed seed in both AUCell and the regulon threshold calculation step. This produces loom files that are identical except for last_modified
attribute.
pySCENIC processes that need to have a seed parameter added:
These require edits to the pySCENIC codebase before being implemented here. Related to https://github.com/vib-singlecell-nf/vsn-pipelines/issues/43