single-cell-vector-search

To find similar cell states across scRNA-seq studies, you can use embedding similarity searches against large atlases. This skill applies vector-search to identify matching phenotypes and construct cross-study cohorts using SCimilarity or Census.

Use SCimilarity for centroid and nearest-neighbor searches on human 10x data with local models. Use Census for cross-species queries and API-driven analyses, allowing robust decision making based on your dataset's target organism and scale.

scRNA-seq embedding similarity searches require a preprocessed AnnData object. You must ensure gene alignment, apply lognorm normalization, verify QC metrics, and account for memory constraints before running the vector-search workflow.

Yes, you can perform cross-species scRNA-seq comparisons using the Census API. It enables cross-species queries to identify biologically similar cell states and interpret phenotypes across different organisms within large-scale atlases.

KNN-based cross-study scRNA-seq analysis is limited by strict gene alignment requirements, lognorm normalization constraints, and memory limitations. Ensure your input data meets QC metrics to avoid failures when searching across large atlases.

You can derive gene signatures from single-cell embeddings by performing centroid-based searches. This identifies similar cell states across scRNA-seq atlases and extracts representative gene expression profiles for phenotype discovery.