`CCA_Call`¶

Supervised pseudotime through Canonical Correlation Analysis. CCA projects the single sample embedding (adata.uns['X_DR_sample']) onto the one-dimensional axis most correlated with a phenotype column in adata.obs (e.g. severity, age, disease stage). The leading CCA axis is reported as a sample-level pseudotime and visualized as a 2D scatter using the PCs that contribute most to the axis. Useful when you already have a phenotype gradient and want to inspect how samples order along it.

Source: sample_trajectory/CCA.py:372

Signature¶

def CCA_Call(
    adata: AnnData,
    output_dir: str = None,
    trajectory_col: str = "sev.level",
    n_components: int = 2,
    auto_select_best_2pc: bool = True,
    verbose: bool = False,
    show_sample_labels: bool = False,
)

Parameters¶

Name	Type	Default	Description
`adata`	AnnData	—	AnnData with the sample embedding in `.uns['X_DR_sample']` (falls back to `.obsm['X_DR_sample']`).
`output_dir`	str, optional	`None`	If provided, writes to `{output_dir}/CCA/`.
`trajectory_col`	str	`"sev.level"`	`.obs` column used as the CCA target.
`n_components`	int	`2`	PCA components drawn from DR before running CCA. Use `10` for RNA, typically `2` for ATAC.
`auto_select_best_2pc`	bool	`True`	Pick the best 2-PC combination for the 2D scatter.
`verbose`	bool	`False`	Print diagnostics.
`show_sample_labels`	bool	`False`	Overlay sample IDs on the plot (clutters when many samples).

Returns¶

Tuple[float, float, Dict, Dict] — a legacy 4-tuple (score, score, pseudotime, pseudotime). With the single-key embedding, both score slots and both pseudotime slots hold the same X_DR_sample result; the 4-tuple shape is preserved for back-compat with the wrapper.

score — CCA correlation coefficient along the leading axis.
pseudotime — {sample_id: float} dict mapping each sample to its position along the CCA axis (used later by run_trajectory_gam_differential_gene_analysis).

Output files¶

Under {output_dir}/CCA/:

pca_{n_components}d_cca_sample.pdf — 2D scatter with the CCA direction overlay.
pca_{n_components}d_cca_sample_contributions.pdf — PC-contribution diagnostic plot.
pseudotime_sample.csv — per-sample pseudotime (sample, pseudotime columns).

Usage¶

from sampledisco.sample_trajectory.CCA import CCA_Call

score, _, pseudotime, _ = CCA_Call(
    adata=adata,
    output_dir="sampledisco_demo_output/rna",
    trajectory_col="sev.level",
    n_components=10,
    auto_select_best_2pc=True,
    verbose=True,
)

CCA_Call¶