Trajectory DGE

Given a pseudotime (from CCA or TSCAN), this step fits a Generalized Additive Model per gene and ranks by effect size + FDR. It emits Lamian-style visualizations: a one-page summary, volcano, per-gene curves, sample density, and per-sample curves.

Call

The pseudobulk is built on the fly from the cell-level AnnData (aggregating cells per sample × cell type), so you pass the preprocessed cell-level adata — not a separate sample-level object. pseudotime_source accepts a dict, a DataFrame, or a path to a CSV/TSV emitted by CCA or TSCAN.

from sampledisco.sample_trajectory.trajectory_diff_gene import (
    run_trajectory_gam_differential_gene_analysis,
)

results_df = run_trajectory_gam_differential_gene_analysis(
    adata,                               # cell-level preprocessed AnnData
    pseudotime_source=expr_pseudotime,   # dict, DataFrame, or CSV/TSV path from CCA/TSCAN
    sample_col="sample",
    celltype_col="cell_type",
    pseudotime_col="pseudotime",
    covariate_columns=None,
    fdr_threshold=0.01,
    effect_size_threshold=1.0,
    top_n_genes=100,
    num_splines=5,
    spline_order=3,
    output_dir="sampledisco_demo_output/rna/trajectoryDEG/expression",
    generate_visualizations=True,
    n_clusters=3,
    top_n_genes_for_curves=20,
    anchor_col=None,                     # orient pseudotime so UP/DOWN are stable
)

Output

Writes → sampledisco_demo_output/rna/trajectoryDEG/expression/:

File	Shows
pseudoDEGs.csv	Filtered pseudoDEG table (gene, effect size, p-value, FDR, rank).
all_gene_results.csv	Full per-gene ranking before filtering.
visualizations/dge_results_summary.png	One-page diagnostic.
visualizations/dge_volcano.png	Effect size vs −log10(FDR).
visualizations/dge_gene_curves.png	GAM fits for the top genes.
visualizations/dge_sample_density.png	Sample distribution along pseudotime.
visualizations/dge_sample_level_curves.png	Per-sample raw + fitted curves.

Result

Visualizations from the GAM-based trajectory DGE step.

See the API page for the full parameter list.