A minor release rather than a patch: eight modules are new since 2.2.4.
ov.flow
| FCS reading, display transforms, the gating strategy tree, and Gating-ML that validates |
ov.mol
| GPU molecular dynamics — simulate, analyse, MM-GBSA |
ov.pl
| table-first statistical plots with an AnnData adapter; clinical statistics and publication figure assembly |
ov.alignment
| homology search (DIAMOND/BLAST+) and Sanger .ab1 verification
|
ov.space
| SPATA2-style coordinate utilities; the arrangement layer (nb / niche / geom); a Stereo-seq reader
|
ov.synbio
| the D-segment gaps closed — reconstruction, omics→GEM, manufacturability, biosecurity |
It also carries user-visible corrections that should not wait for another cycle: roughly forty synbio defects that produced plausible wrong numbers, Gating-ML output that was not valid Gating-ML, CopyKAT failing silently when its genome tables are missing, a [synbio] extra that could not be installed, and the rpy2 mclust bridge replaced by the pure-Python pymclustR backend.
Where a claim can be checked, the notes give the number: the ov.space statistics against squidpy 1.6.5 on an identical graph (interaction_matrix, nhood_enrichment counts and centrality exact; co_occurrence and all three Ripley modes at Pearson r = 1.0000; var_by_distance to 1.11e-16), and pymclustR against R mclust 6.1.2 (log-likelihood and BIC to six decimals, every cell in the same cluster).
Full notes, including the 2.2.4 entry that was never written: Release_notes.md