Repository navigation
Highlights
- Genes with NaN counts are ignored with a warning instead of raising an error, and get NaN dispersions, LFCs and p-values (#502).
plot_MAandplot_dispersionstake matplotlib axes through a newaxargument (#504).- A fitted
DeseqDataSetcan be written to h5ad (#477, #510). - An empty count matrix raises a clear error (#503).
- LFC shrinkage and the dispersion fitting fallbacks were corrected (#470, #471, #472), as were
poscountssize factors for genes whose positive counts are all 1 (#494). - The synthetic example dataset ships with the wheel, so
load_example_dataworks without the repository (#467). - The documentation has a new tutorial on continuous covariates (#507).
- PyDESeq2 works with pandas 3.1 and anndata 0.13 (#456, #501).
- The test suite compares PyDESeq2 with DESeq2 1.50.2, which it runs through rpy2 (#505).
Changed behaviour
- Status updates and timings are logged through the shared scverse logger and are hidden by default (#506).
Setscverse_misc.logging.config.verbosity = "info"to see them.
summary()andlfc_shrink()still print their results table. uns["trend_coeffs"]anduns["vst_trend_coeffs"]are dicts with the keysa0anda1instead of pandas Series, andobsm["design_matrix"]is a plain DataFrame (#477, #510).- Python 3.12 or later is required, and the minimum supported dependencies follow SPEC 0 (#458).
- scverse-misc is a new dependency (#506).
What's Changed
- docs: fix broken intersphinx links by @Ekin-Kahraman in #442
- [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #439
- [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #446
- Fix AttributeError when design is a DataFrame by @Ekin-Kahraman in #443
- [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #448
- Fix type errors surfaced by anndata 0.13 by @Zethson in #456
- Move test dependencies into a
testdependency group by @Zethson in #455 - Remove CODEOWNERS by @Zethson in #457
- Align minimum supported dependencies with SPEC 0 by @Zethson in #458
- Adopt the scverse cookiecutter template 0.8.0 by @Zethson in #459
- Chore(deps): bump the actions-deps group with 4 updates by @dependabot[bot] in #460
- Split utils into modules grouped by theme by @Zethson in #461
- Move the docs to the template layout and restyle them by @Zethson in #462
- Remove duplicated types from docstrings by @Zethson in #463
- Fix load_example_data for non-editable installs by bundling datasets/ into the wheel by @Zethson in #467
- Fix README logo path after docs restructure by @Zethson in #474
- docs: move to scverse-doc and adopt the shared scverse docs TOC by @Zethson in #487
- ci: add autofix.ci workflow by @flying-sheep in #496
- style: switch to panache markdown formatting by @flying-sheep in #495
- Chore(deps): bump the pre-commit group with 5 updates by @dependabot[bot] in #497
- build: require scverse-doc 0.4 by @Zethson in #488
- Chore(deps): bump the pre-commit group with 2 updates by @dependabot[bot] in #499
- Fix prior curvature in shrinkage Hessian by @Intron7 in #470
- Preserve dispersion regularization in optimizer fallback by @Intron7 in #471
- Honor the requested coefficient in shrinkage grid fallback by @Intron7 in #472
- Chore(deps): bump the actions-deps group across 1 directory with 3 updates by @dependabot[bot] in #486
- fix: adapt to pandas 3.1 deprecations by @Zethson in #501
- Keep genes whose positive counts are all 1 in poscounts size factors by @MMatinGerami in #494
- feat: ignore genes with NaN counts instead of raising by @Zethson in #502
- fix: raise a clear error for an empty count matrix by @Zethson in #503
- style: raise the line length to 120 by @Zethson in #508
- docs: add a tutorial on continuous covariates by @Zethson in #507
- build: add Lukas Heumos to the authors by @Zethson in #509
- feat: accept matplotlib axes in plot_MA and plot_dispersions by @Zethson in #504
- feat: log status updates through the shared scverse logger by @Zethson in #506
- test: compare with DESeq2 through rpy2 in a dedicated environment by @Zethson in #505
- Store dispersion trend coefficients as a dict so the dataset can be written to h5ad by @raashish1601 in #477
- fix: store the design matrix as a DataFrame so the dataset can be written to h5ad by @Zethson in #510
- docs: rewrite the README in the style of the scverse template by @Zethson in #511
New Contributors
- @Ekin-Kahraman made their first contribution in #442
- @dependabot[bot] made their first contribution in #460
- @flying-sheep made their first contribution in #496
- @Intron7 made their first contribution in #470
- @MMatinGerami made their first contribution in #494
- @raashish1601 made their first contribution in #477
Full Changelog: v0.5.4...v0.6.0