cytorete — Greek cyto- (κύτος, "cell") + Latin rēte ("net"), the
cell's network. Pronounced sy-toh-REE-tee (/ˌsaɪtoʊˈriːtiː/), with
rete as in rete mirabile. The name is the claim: a regulatory network
is a property of a cell type, not of a tissue.
cytorete infers regulons (a TF and its target set, with weights) from
scRNA-seq, on the PIASO stack:
- Promoter cistrome — strand-aware promoter windows from the PIASO-data
TSS BEDs, sequences from the 2bit genome, scanned withpiaso.pp.scan_motifs
(JASPAR / CIS-BP) → a TF × gene motif-support matrix. inferRegulon— motif-supported edges scored by trans
co-specificity (piaso.tl.cospecificity_trans): a TF and a target
qualify when their COSG cell-type specificity profiles agree, per cell
type, with NES pruning.regulonActivity/regulonSpecificity— per-cell regulon scores
(PIASO's scoring engine; AnnData or cytome streaming) and cell-type
specificity of each regulon.- Plots — regulon heatmaps and TF-vs-activity scatters.
Works on AnnData or on .cytome files, in memory or streamed. inferRegulon
also takes a .cytome path directly and closes the file when it is done.
Names for the multiome (RNA+ATAC) chain exist but raise an informative
ImportError in this distribution.
Scale. One Stereo-seq section — 121,767 spatial bins × 28,204 genes —
takes 29 s for the regulon step at roughly a gigabyte of RSS, because the
matrix is streamed and never held whole.
Requires piaso-tools >= 1.2.2, cosg >= 1.1.2, cytome >= 0.2.5.
Layout. cytorete.tl and cytorete.pp hold their modules directly —
there is no grn subpackage, because the whole package is the GRN method
and the extra segment would name nothing. So it is cytorete.tl.inferRegulon,
not cytorete.tl.grn.inferRegulon.
Suite (this distribution): 60 passed, 0 skipped, against the released
piaso-tools / cosg / cytome wheels. Tests skip locally without the
optional motif extra (py2bit) or a JASPAR MEME file; the release
workflow requires the extras and fetches JASPAR, so a release run has the
full suite rather than a quietly reduced one.
Documentation lives with the rest of the stack on
piaso.org: RNA regulon inference and regulons on spatial data,
which runs the whole workflow across a mouse embryo section and checks
the result against the published SCENIC regulons for the same tissue.