tradeSeq-python 1.13.12

Python port of the R tradeSeq package (tracks statOmics/tradeSeq 1.13.12 @ 02a9050) — trajectory-based differential expression analysis (NB-GAM)

tradeSeq-python

Python tradeSeq package — trajectory-based differential expression analysis with negative-binomial generalized additive models (NB-GAM).

Installation

pip install tradeSeq-python                # from PyPI

Quickstart

import tradeseq as ts

adata = ts.load_paul15()                              # AnnData (2660 × 240)

ts.fit_gam(adata, n_knots=6)                          # NB-GAM per gene

assoc = ts.association_test(adata)                    # expression vs pseudotime
start = ts.start_vs_end_test(adata)                   # progenitor markers
diff  = ts.diff_end_test(adata)                       # between-lineage endpoints
patt  = ts.pattern_test(adata)                        # between-lineage patterns
early = ts.early_de_test(adata, knots=(1, 2))         # early drivers

ts.plot_smoothers(adata, gene=start['waldStat'].idxmax())
ts.plot_gene_count(adata, gene=start['waldStat'].idxmax())

Choose the number of knots with evaluate_k / evaluate_k2 and plot_evaluatek_results. evaluate_k2 is a speed-up version of the original evaluate_k.

Tutorials

Runnable notebooks that reproduce the R tradeSeq vignettes live under tutorials/:

Notebook	Coverage
tradeSeq.ipynb	Full Wald battery on the Paul-2015 myeloid trajectory — mirrors vignettes/tradeSeq.Rmd
fitGAM.ipynb	Model-fitting options — covariates, parallelism, list-mode output, convergence diagnostics