`cluster`¶

K-means clustering on the single-key sample-level embedding obsm['X_DR_sample']. Writes the CSV mapping of samples to clusters and a 2D scatter of the first two embedding dimensions colored by cluster. The cluster assignments can be fed directly into proportion_test and raisinfit via the sample_to_clade argument.

Single-key embedding

The current pipeline produces one sample embedding, X_DR_sample. The legacy use_expression / use_proportion flags and the old X_DR_expression / X_DR_proportion keys are gone — both flags are accepted for backward compatibility but ignored, and the returned tuple has the same mapping in both slots.

Source: sampledisco/sample_clustering/cluster.py:11

Signature¶

def cluster(
    pseudobulk_adata: ad.AnnData,
    output_dir: str,
    number_of_clusters: int = 5,
    use_expression: bool = True,
    use_proportion: bool = True,
    random_state: int = 0,
) -> Tuple[Optional[Dict[str, int]], Optional[Dict[str, int]]]

Parameters¶

Name	Type	Default	Description
`pseudobulk_adata`	AnnData	—	Sample-level AnnData with `X_DR_sample` in `.obsm`.
`output_dir`	str	—	Writes under `{output_dir}/sample_cluster/`.
`number_of_clusters`	int	`5`	K for K-means.
`use_expression`	bool	`True`	Legacy argument — accepted but ignored.
`use_proportion`	bool	`True`	Legacy argument — accepted but ignored.
`random_state`	int	`0`	Seed for reproducibility.

Returns¶

Tuple[Dict, Dict] — (expr_results, prop_results), kept for backward compatibility. Both slots contain the same {sample_id: cluster_label} mapping computed on X_DR_sample.

Output files¶

Under {output_dir}/sample_cluster/:

kmeans_clusters_sample.csv
kmeans_sample_embedding.png

The passed-in AnnData also gets a cluster_sample_kmeans column in .obs.

Usage¶

from sampledisco.sample_clustering.cluster import cluster

sample_clusters, _ = cluster(
    pseudobulk_adata=pseudo_adata,
    output_dir="sampledisco_demo_output/rna",
    number_of_clusters=4,
)

cluster¶