Changelog#
All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog and this project adheres to Semantic Versioning.
v1.5.0#
π Features#
pt.md.Cytokineprovides the cytokine response signatures of the Human Cytokine Dictionary for 90 cytokines in 24 PBMC cell types, as gene sets that theEnrichmenttests take astargets(#1186) @Jenniliu12, @ZethsonStatsmodelsfits linear mixed models when the design contains a random intercept such as~condition + (1 | donor), by REML with Satterthwaite t-tests likelmerTest, so single cells can be tested while accounting for the samples they come from; the kernel is shared with Dialogue and the(1 | variable)syntax with Milo (#1147) @Sizerta, @Zethsonassign_mixture_modelalso writes the binary cell-by-guide assignment toadata.layers[output_layer], since the per-cell string in.obscaps and joins the assigned guides (#1173) @Zethson
β‘ Performance#
Augur scores each cross-validation fold in one pass over the forest instead of six scorer passes, skips scikit-learnβs per-tree parameter validation and no longer ships the fitted trees back from the workers, with identical metrics and importances on seeded runs (#1178) @Zethson
Dialogue runs the PMD iterations in a numba kernel (20x per fit) and solves the iterative NNLS on the normal equations, replacing the
sparseccadependency; whole runs take 3 to 13 s on 5k to 27k cells, about 2x faster than before with up to half the peak memory, where R needs 2 to 16 minutes (#1177) @Zethsonassign_mixture_modelfits each guide on its distinct count values, 20x faster on Papalexi (2.9 s to 0.14 s) and with a third of the peak memory at 100k cells x 500 guides (#1172) @Zethson
β οΈ Changed results#
Dialogue reproduces the R reference end to end: REML mixed models with Satterthwaite t-tests like
lmerTest, Rβs PMD penalty, permutation test and cell-type pruning, Pearson pre-selection of signature genes, one-sided gene p-values, andtest_phenotype_associationnow followsDIALOGUE.pheno(first level orTrueas the indicator, pooledAllrow, no Fisher combination) (#1177) @ZethsonDialogue orders the shared samples like Rβs ICU collation, since that order decides the signs of the PMD initialization and with them the programs found (#1177) @Zethson
Augur is reproducible by default: unseeded estimators and folds are seeded by the subsample index like R Augur, and
predict_differential_prioritizationdraws its permuted means without replacement from a seeded generator, as R does (#1178) @ZethsonDialogueβs
min_cells_per_sampledefaults to Rβs 15 and only gates the ANOVA feature filter, which now raises below five retained features instead of keeping all of them;n_permutationscounts the observed fit, so values below 11 cannot pass the defaultempirical_alpha(#1177) @Zethsonassign_mixture_modelfits the Poisson-Gaussian mixture by multi-start EM instead of 500 Adam steps from one start, which left thresholds unconverged, swapped the components on high-MOI guides and dropped small guides; thresholds now match crispat run to convergence on 105 of 106 Papalexi guides,learning_rateis deprecated andn_itercaps the EM iterations (#1172) @Zethson
π Bug Fixes#
Augur encodes the condition labels after filtering to two conditions, so binary precision, recall and F1 are no longer NaN when the kept labels were not the first two; those metrics now take the first condition as the positive class like R Augur,
min_cellsdefaults tosubsample_sizeand applies per condition like R, data sets with fewer than 1000 features skip feature selection like R, per-cellaugur_scoreis assigned by exact cell type match, and regressors use unstratified folds (#1178) @ZethsonDialogue treated signed log10 p-value z-scores as normal z-scores in phase 3, which inflated strong genes, lost the up/down direction and dropped underflowing genes from the Fisher combination (#1177) @Zethson
Mixscapeβs
plot_barplotno longer fails when the number of target genes is not a multiple of five, and Mixscape takes the control, perturbed and non-perturbed class labels fromcontrol,perturbation_typeand the newnon_perturbed_labelinstead of hardcodingNT,KOandNP;plot_heatmapandplot_perturbscoreuse themixscape_classandpert_keycolumns they are given,plot_perturbscorekeeps a custompalettewithbefore_mixscape=True, andplot_barplottakes apalette(#1191) @kkkklck, @ZethsonMixscape takes the minimum number of cells left in a perturbed class from the new
min_cellsinstead ofmin_de_genes, as Seurat does, which changes results only for a non-defaultmin_de_genes(#1189) @kkkklck, @ZethsonMiloβs
plot_nhood_counts_by_condraised with pandas 3.1, which kept the sample columnβs name on the per-sample index (#1181) @ZethsonGuide assignment:
assign_by_thresholdandassign_to_max_guideaccept CSC and sparse-array input, explicitly stored zeros no longer enter the mixture fit,plot_heatmapsaves a reusable cell order andonly_return_results=Trueno longer writes toadata.var(#1172) @Zethson
π§° Maintenance#
Require Python 3.13 (#1172) @Zethson
Build the docs with the shared scverse theme and configuration from
scverse-doc, with the logo blue as its accent for links and highlights (#1179) @ZethsonThe docs take their accents from the scverse-doc registry and colour the sidebar captions in the logo red (#1182) @Zethson
Group the docs sidebar into Getting started, Guides, Reference and Project like the other scverse core packages, with a new quickstart that replaces the background page and a new glossary (#1180) @Zethson
Skip the network tests when a download fails with a
URLError(#1184) @ZethsonKeep R from loading the system libpython in the tutorial notebooks on Ubuntu 26.04 (#1185) @Zethson
Skip PyQt6βs stubs in mypy, whose invalid syntax made mypy abort (#1190) @Zethson
Require LLM usage disclosure in the pull request template (#1165) @Zethson
v1.4.0#
π Features#
Add signature reversal scoring to enrichment (#1082) @daveringelberg
Add native Hill dose-response fitting (#1097) @daveringelberg
Expose datasets as
pt.dsin addition topt.dt(#1103) @ZethsonMake Milo differential abundance testing match R Milo, including the pydeseq2 solver (#1109, #1110) @Zethson
Add
n_jobstofit_dose_responseandbalance_classestoMLPClassifierSpace.compute(#1142, #1153) @ZethsonSeed the control gene sampling of
Enrichment.scoreviarandom_state(#1128) @ZethsonAccept
hdi_probin scCODA summaries with arviz 1.x (#1159) @Zethson
β‘ Performance#
Hot loops across all tools were vectorized or moved into numba kernels, with results verified against the previous implementation or the reference library on real data.
Distances:
DistanceTest35x,ks_test27x,pairwisewitht_test/sym_kldiv9-13x, MMD without n x n kernel matrices, numba KDE formean_var_distribution, exact NB2 fits fornb_ll(#1118, #1134, #1135) @ZethsonDifferential expression: Statsmodels contrasts 11-20x and batched OLS fits, sparse Wilcoxon and t-test kernels, vectorized
PermutationTeststatistic (#1124, #1136) @ZethsonMilo: batched neighbourhood refinement and pydeseq2 likelihood-ratio fits (3-7x), numba GLMM fits (~34x) (#1125, #1137) @Zethson
Dialogue: numba random-intercept fits (
test_celltype_pairs119 s to 0.9 s at 30k cells) (#1124, #1133) @ZethsonMixscape and Mixscale: sparse signature assembly, numba EM, one DE call per reference (
mixscale11x,mixscape6x) (#1140, #1143) @ZethsonPerturbation spaces, enrichment and metadata:
Enrichment.score88x,label_transfer19x,CellLine.correlate6x, chunked MLP training, parallel PubChem lookups (#1139, #1142) @ZethsonscCODA:
make_arvizno longer simulates a discarded N x N counts tensor (minutes to under a second) (#1138) @ZethsonGuide assignment 120x on sparse input and faster CINEMA-OT without N x N intermediates (#1140) @Zethson
β οΈ Changed results#
Mixscape holds the control variance fixed during EM as Seurat does; about 0.4% of KO/NP calls change on Papalexi 2021 (#1144) @Zethson
Euclidean e-distance and mean pairwise distance average over all n^2 cell pairs, so
Distance.__call__,pairwiseandDistanceTestagree (#1151) @ZethsonDistance.bootstrapresamples the second group with its own size (#1152) @ZethsonscGen
batch_removalapplies its latent correction, which was previously discarded (#1119) @ZethsonDialogue p-values move slightly because the mixed-model fit converges where statsmodelsβ BFGS stopped early (#1133) @Zethson
nb_llfits the overdispersed genes that statsmodels silently dropped (#1135) @ZethsonWilcoxon, t-test, e-distance and MMD statistics on float32 input are accumulated in float64 (#1134, #1136) @Zethson
CellLine.correlatecomputes every pair instead of mirroring the asymmetric matrix (#1121) @Zethsonevaluate_clusteringcomputes ASW on the cells instead of on rows of the distance matrix (#1120) @ZethsonAdjusted p-values of the simple DE tests are kept when some variables have NaN p-values (#1122) @Zethson
π Bug Fixes#
Keep scCODA and tascCODA working under numpyro 0.22 (#1104) @Zethson
Use regression-appropriate scorers for Augur regressor estimators (#1105) @Sizerta
Dialogue: use
additional_covariates, handle degenerate genes and unused categorical levels (#1132, #1156) @ZethsonMixscape and Mixscale:
scale=Falseon sparse layers,epsilonfor every split, control labels with spaces, duplicateobs_names(#1129, #1130, #1131, #1145, #1149, #1155) @ZethsonDistances: sparse MMD in
pairwise, bounded memory, symmetric-matrix assumption, one-cell t-test groups (#1150, #1154) @ZethsonPerturbation spaces: sparse
.X, pseudobulks keep obs columns that are NaN for some groups (#1146, #1158) @ZethsonCellLine.annotate_from_prismmatches drug names case-insensitively (#1127) @ZethsonScgen.plot_reg_var_ploton sparseX, scCODA tree plots aftersave=, deterministicfrom_scanpycolumn order (#1160, #1161, #1162) @ZethsonMake the PubChem compound annotation test robust to outages (#1111) @Zethson
π§° Maintenance#
v1.3.0#
π Features#
Make JAX optional and reimplement scGen without scvi-tools (#1084) @Zethson
π Bug Fixes#
Raise a clear error when a
da_nhoodsdesign has no residual degrees of freedom (#1077) @ZethsonDrop empty annotation rows from
plot_da_beeswarm()(#1078) @ZethsonKeep permutation p-values aligned with their group labels in
DistanceTest(#1079) @ZethsonEnsure right fit object is in R scope (#1073) @sagar87, @Zethson
π§° Maintenance#
Read the Lightning intersphinx inventory from Read the Docs (#1083) @Zethson
v1.2.0#
π Features#
Rework perturbation spaces around a perturbation-level API (#1032) @Zethson
Milo: add mixed models to differential abundance testing (#1069) @Zethson
Milo: add neighbourhood grouping (#801) @MaximilianNuber, @Zethson
Add
plot_ma()for DGE results (#1066) @ZethsonAdd
plot_credibleoption toplot_effects_umap()(#1063) @ZethsonAdd text and HTML representations to DE model classes (#1070) @Zethson
Rename
DistanceTestalphatopadj_threshold(#1062) @ZethsonVectorize and parallelize the DE methods that support it (#1051) @Zethson
Speed up Augur
predict()(#1043) @ZethsonShow output filename in download progress (#1056) @corkang
π Bug Fixes#
scGen: vendor JAX backend to support scvi-tools 1.5.0 (#1042) @Zethson,
scGen: make the JAX implementation reproduce the PyTorch one (#1028) @Zethson
Fix import of removed scanpy
_get_obs_rep/_set_obs_rep(#1050) @Zethsonpseudobulk: only transfer constant obs columns (#1061) @Zethson
Correct the version of the Milo deprecations (#1064) @Zethson
π§° Maintenance#
Adopt the src layout, VCS versioning and supply chain pinning (#1057) @Zethson
Type check pertpy and tests with mypy (#1053) @Zethson
Adopt anndataβs mypy configuration (#1060) @Zethson
Deduplicate metadata download boilerplate (#1027) @Zethson,
Align dataset download progress (#1059) @corkang
Add discoverability metadata (#1067) @Zethson
Update template to v0.8.0 (#1045) @scverse-bot, @Zethson
v1.1.1#
π Features#
π Bug Fixes#
tascCODA: fix no credible effects returned (theta collapse) (#1017) @Zethson
tascCODA: fix always returning insignificant results (#1016) @Zethson
PseudobulkSpace: fix all-NaN extra obs with
groups_col(#1006) @Zethsonperturbation_space: skip None-keyed layers in
_combine(#1000) @Zethsonplot_effects_barplot(): remove FutureWarning (#1013) @LuisHeinzlmeiercoda: fix ImportError (#1014) @LuisHeinzlmeier
π§° Maintenance#
v1.1.0#
π Features#
milo: speed up pydeseq2 da_nhoods ~100x via poscounts size factors (#992) @Zethson
Guide RNA assignment: crispat-matching JAX MAP fit (10β30Γ faster) (#987) @Zethson
DIALOGUE: from-scratch port of the R reference (PMD + HLM + iterative NNLS) (#988) @Zethson
PyDESeq2: optional apeGLM LFC shrinkage (#985) @Zethson
Add layout option to coda plot_boxplots (#718) (#932) @emekaokoli19
π Bug Fixes#
scCODA: fix run_hmc TypeError; clarify log2-fold-change sign (#993) @Zethson
scCODA/tascCODA: make make_arviz stateless (#991) @Zethson
scCODA: unify covariate index format across get_effect_df / summary / credible_effects (#989) @Zethson
Reject all-zero contrasts in test_contrasts (#981) @Zethson
Fix PseudobulkSpace.compute losing X with anndata main (#979) @Zethson
Fix plot_effects_umap KeyError on Sccoda output (#780) (#971) @LiudengZhang
Fix marker positions: calculate marker positions from DataFrame, not rendered labels (#970) @kimjune01
Fix plot_multicomparison_fc ValueError with default figsize (#965) @kimjune01
Fix minor bugs (#960) @Zethson
Fix
addandsubtract(#959) @Zethson
π§° Maintenance#
v1.0.6#
π Features#
Support Python 3.14 (#893) @Zethson
π Bug Fixes#
π§° Maintenance#
Mention paper in README (#935) @Zethson
v1.0.5#
π Bug Fixes#
π§° Maintenance#
v1.0.4#
π Features#
π Bug Fixes#
π§° Maintenance#
fix isort formatting (#878) @flying-sheep
v1.0.3#
π Features#
v1.0.2#
π Bug Fixes#
v1.0.1#
π Features#
π Bug Fixes#
Fix milo output writing (#821) @Zethson
v1.0.0#
π Features#
v0.11.5#
π Features#
π Bug Fixes#
v0.11.4#
π§° Maintenance#
Remove decoupler (#783) @Zethson
v0.11.3#
π Features#
v0.11.2#
π Features#
Simplify Metadata errors (#765) @Zethson
v0.11.1#
π Bug Fixes#
Restructure documentation & fix scCODA not requiring tcoda extra (#762) @Zethson
v0.11.0#
π Features#
Improve CI test setup & speed up mixscape (#757) @Zethson
Speed up mixscape (#756) @Zethson
Optimize mixscape (#735) @Zethson
Cleaner mixscape (#746) @eroell
Speedup Mixscapeβs lda function (#754) @eroell
Skip milotests properly & cache metadata (#752) @Zethson
Sparse threshold max assignment (#744) @Zethson
Speedup import times (#743) @stefanpeidli
Add sparse threshold guide assignment & fix docstrings (#737) @Zethson
More robust download (#721) @Zethson
Support Python 3.13 (#720) @Zethson
PRISM drug response metadata annotation (#716) @Lilly-May
Improve CellLine metadata module UX (#717) @JiriBruthans
Switch from ete3 to ete4 (#618) @dengzq1234
Move pynndescent import out of for-loop (#761) @eroell
Cleaner CinemaOT (#745) @eroell
π Bug Fixes#
Change how guide assignment model saves params from uns to var (#758) @stefanpeidli
A ton of doc improvements & fix mudata warnings & Augur interface (#759) @Zethson
Change cell_wise_metric from βsqeuclideanβ to βeuclideanβ for edist (#733) @stefanpeidli
Check only for numeric in statsmodels (#725) @grst
Remove leidenalg from requirements (#729) @Zethson
Remove pyenchant (#728) @Zethson
Fix coda plot axes (#739) @mschilli87
π§° Maintenance#
v0.10.0#
π Features#
Allow Custom Feature Spaces in Dialogue (#712) @grpinto
Add Poisson-Gauss Mixture model for Guide Assignment (#709) @stefanpeidli
Add perturbation signature calculation from replicate control cells (#695) @Lilly-May
Optimize test speed (#699) @Zethson
Optimize MeanVarDistributionDistance (#697) @Zethson
π Bug Fixes#
v0.9.x#
π Features#
Python 3.12 support (#644) @Lilly-May
Add lazy loading for deferred modules (#647 and #648) @Zethson
Support for scipy 1.14 (#646) @andr-kun
Update scperturb datasets (#649) @tessadgreen
Augur scsim warnings (#670) @Zethson
Add uncertainty score in KNN label_transfer in PerturbationSpace (#658) @stefanpeidli
Incorporate use case tutorials (#665) @Lilly-May
Add plots for DE analysis and unify plotting API (#654) @Lilly-May
Switch to formulaic-contrasts (#682) @grst @emdann
Remove anndata pin (#653) @Lilly-May
π Bug Fixes#
π§° Maintenance#
Fix docs rendering for classes using lazy import (#651) @Lilly-May
v0.8.0#
π Features#
Add hagai dataloader (#627) @Zethson
Add more informative error if not every cell type is represented in e⦠(#620) @Zethson
Add typehints_defaults (#619) @Zethson
Add metadata notebook (#609) @Zethson
Add multicondition analysis (differential gene expression interface) (#607) @Zethson
Add zhang 2021 dataloader (#602) @Zethson
Use all (#588) @Zethson
Use lamin logger (#586) @Zethson
Add two distance metrics, three-way comparison and bootstrapping (#608) @wxicu
Metadata GDSC annotation improvements (#625) @Lilly-May
Add file lock during downloading (#590) @wxicu
π Bug Fixes#
v0.7.0#
Due to a lack of release notes for 0.6.0, this release may contain the some of the changes of 0.6.0 as well.
Note that the syntax for plotting for most tools changed. We removed the pt.pl modules in favor of moving all plots with plot_ prefixes to the respective tools. We hope that this will make the documentation easier to navigate.
π Features#
Silence pandas groupby() warning when running Augur predict (#573) @namsaraeva
Added milo beeswarm example plot (#552) @namsaraeva
Lazily metadata loading (#544) @wxicu
Added tasccoda example plots (#527) @namsaraeva
Add examples to plotting functions (#511) @namsaraeva
Replace plotnine implementation of barplot with seaborn/matplotlib (#441) @kcArtemis
Cinemaot dataset loader (#424) @Lilly-May
Add classifiers (#427) @Zethson
Remove remotezip (#401) @wxicu
Add new distances (#304) @mojtababahrami
wasserstein distance return type float (#386) @eroell
Implementation of CINEMA-OT for pertpy - follow up (#379) @MingzeDong
Implementation of CINEMA-OT for pertpy (#377) @MingzeDong
Update cell line meta data class (#539) @wxicu
Add dgidb and pharmgkb for drug annotation (#575) @wxicu
Add sciplex-GxE dataloader (#576) @namsaraeva
Harmonize plots (#574) @namsaraeva
Compute method for MLPClassifierSpace (#565) @Lilly-May
uv for CI (#569) @Zethson
Logistic regression support for the Discriminator Classifier (#560) @Lilly-May
Allow saving of CODA plots (#559) @namsaraeva
Order plot functions last (#555) @Zethson
Sparse guide RNA plot (#554) @Zethson
Rename cmap to palette (#553) @Zethson
Expand load for sccoda docstrings (#543) @Zethson
Pseudobulk improvements (#529) @Lilly-May
Add more detailed tool explanations (#528) @Zethson
Add knn imputation (#517) @Zethson
Add combosciplex (#512) @Zethson
Add DE set comparisons (#489) @Zethson
More informative images in tutorial gallery (#488) @Lilly-May
Add perturbation space tutorial (#487) @Lilly-May
Add enrichment (#482) @Zethson
Save perturbation and clustering labels as categorical (#481) @Lilly-May
Optimize from_scanpy (#473) @Zethson
Add Drug metadata from chembl (#480) @Zethson
Add DE filtering (#477) @Zethson
PertSpace docs improvements (#471) @Lilly-May
Make dialogueβs load protected (#464) @Lilly-May
New plotting API (#456) @Zethson
Added smillie_2019 dataloader and renamed smillie dataset to tasccoda_example (#450) @Lilly-May
Docs examples for CINEMA-OT (#433) @Lilly-May
Removing dependencies (#426) @Zethson
Added example for pt.pl.milo.nhood_graph (#423) @Lilly-May
scope imports of ete3 in tasccoda (#422) @Zethson
Add cinemaot nb (#419) @Zethson
DIALOGUE attribution (#371) @Zethson
DIALOGUE extensions and plots (#368) @tessadgreen
Set run nuts random key to 0 by default. (#363) @Zethson
Remove statsannotations (#362) @Zethson
Refactoring (#354) @Zethson
Add combosciplex (#512) @Zethson
π Bug Fixes#
Fix stephenson 2021 and refactor (#374) @Zethson
fix: double plot issue related to barplot (#460) @kcArtemis
Small fixes to milo.plot_da_beeswarm (#551) @emdann
Check and fix plotting functions (#579) @namsaraeva
Fix jax config (#550) @Zethson
fixed augur runtime (#547) @zzheng18
Fix Distance docstring rendering (#537) @Zethson
Address some Jax random key issues (#531) @Zethson
Fixed Augur bug (#533) @namsaraeva
Fix discriminator classifier nn dimensions (#475) @Lilly-May
bug fixes, implementation tweaks, and additional distances (#397) @yugeji
Fix subsetting of milo.plot_da_beeswarm (#472) @Lilly-May
CentroidSpace AnnData Annotations (#455) @Lilly-May
DiscriminatorClassifier compatibility with sparse matrices (#453) @Lilly-May
Fix code coverage (#432) @flying-sheep
Fixed set_fdr for tascCODA models (#411) @Lilly-May
Small bug fixes and docs improvements (#409) @Lilly-May
Move to new numpy.random.Generator (#396) @Lilly-May
Fix cinema OT test (#392) @Zethson
fix naming of example data in doc examples (#387) @eroell
Fix stephenson 2021 and refactor (#374) @Zethson
scCODA: add low/high acceptance probability warnings (#366) @johannesostner
actually use
neighbors_keyinmilo(#418) @maarten-devriesFixed some bugs (due to inconsistencies during migration) and typos in CINEMA-OT (#413) @MingzeDong
fix Implicit modification warnings in cellline tests (#404) @wxicu
v0.5.0#
π Features#
π Bug Fixes#
v0.4.0#
π Features#
π Bug Fixes#
docstring enhancement for sample identifiers (#237) @johannesostner
improve mixscape documentation & fix a bug (#222) @xinyuejohn
Fix docs ete3 (#206) @Zethson
Update _dialogue.py (#203) @tessadgreen
scCODA bugfix: duplicates in covariate_obs (#198) @Zethson
fix #207 ete3 warning masks other ImportErrors (#208) @Moomboh
v0.3.0#
π Features#
π Bug Fixes#
π§° Maintenance#
v0.2.0#
π Features#
π Bug Fixes#
π§° Maintenance#
v0.1.0#
π Features#
Augurpy refactoring (#148) @Zethson
add kang dataset (#146) @antschum
add inits for plots, fix error messages and output, tutorial version (#144) @antschum
Feature/furo md (#141) @Zethson
add augurpy (#135) @antschum
add better data draft (#71) @Zethson
Documentation structure (#23) @Zethson
add pypi-latest & architecture (#22) @Zethson
π Bug Fixes#
add inits for plots, fix error messages and output, tutorial version (#144) @antschum