omicverse 2.2.1 — install with pip install -U omicverse (PyPI).
A large feature release: 8 brand-new top-level modules, a rebuilt trajectory/pseudotime stack, out-of-core preprocessing for datasets that don't fit in RAM, and an Apple-Silicon GPU UMAP backend. ~190 commits since v2.2.0.
New modules
ov.genetics— statistical genetics / GWAS pipeline,simulate_gwas_study, cross-trait colocalization (#743, #745)ov.protein— bulk proteomics + real datasets + DEqMS benchmark (#730)ov.mol— molecular structure & drug-binding (#765)ov.epi— epigenomics (wraps epione) (#800)ov.airr— immune-repertoire / AIRR-seq, TCR/CoNGA (#751, #754, #795)ov.single.Metabolism+MetaboliteCCC— single-cell metabolism (#764)ov.single.ev— single extracellular-vesicle proteomics (#759)ov.space.histo— H&E → spatial transcriptomics prediction (#789)
Highlights
- Out-of-core preprocessing (AnnDataOOM) — chunked
filter/HVG/normalize/regress/scrubletfor data that doesn't fit in RAM;ov.pp.ambient(#806, #804, #802, #762) - Trajectory & pseudotime —
PseudotimeFate, MIRA + CellRank downstream, RegVelo GRN, StaVIA, pseudotime-velocity streamplots (#787, #792, #786) - Bulk & DEG — edgeR + limma-voom backends,
timecourse_deg, blockwise PyWGCNA, lipidomics (#734, #746, #738, #741) - Unified
MetaCell— 7 backends + mcRigor validator (#703, #729) - GPU non-parametric UMAP —
ov.utils.gpuex.umap, Apple-Silicon (MLX/metal) + CUDA backends (#812, #814) - In-silico perturbation, simpleaf/salmon/alevin-fry alignment,
ov.pl.funky_heatmap(#781, #750, #776)
Full changelog: Release notes — v2.2.1.