scib.metrics.trajectory_conservation

scib.metrics.trajectory_conservation(adata_pre, adata_post, label_key, pseudotime_key='dpt_pseudotime', batch_key=None)

Trajectory conservation score

Trajectory conservation is measured by spearman’s rank correlation coefficient \(s\), between the pseudotime values before and after integration. The final score was scaled to a value between 0 and 1 using the equation

\[trajectory \, conservation = \frac {s + 1} {2}\]
Parameters:

  • adata_pre – unintegrated adata

  • adata_post – integrated adata

  • label_key – column in adata_pre.obs of the groups used to precompute the trajectory

  • pseudotime_key – column in adata_pre.obs in which the pseudotime is saved in. Column can contain empty entries, the dataset will be subset to the cells with scores.

  • batch_key – set to batch key if you want to compute the trajectory metric by batch. By default the batch information will be ignored (batch_key=None)

This function requires pseudotime values in .obs of the unintegrated object (adata_pre) computed per batch and can be applied to all integration output types. The input trajectories should be curated manually as the quality of the metric depends on the quality of the metric depends on the quality of the annotation. The integrated object (adata_post) needs to have a kNN graph based on the integration output. See User Guide for more information on preprocessing.

Examples

# feature output
scib.pp.reduce_data(
    adata, n_top_genes=2000, batch_key="batch", pca=True, neighbors=True
)
scib.me.trajectory_conservation(adata_unintegrated, adata, label_key="cell_type")

# embedding output
sc.pp.neighbors(adata, use_rep="X_emb")
scib.me.trajectory_conservation(adata_unintegrated, adata, label_key="celltype")

# knn output
scib.me.trajectory_conservation(adata_unintegrated, adata, label_key="celltype")