Support homeCell Ranger ARCAlgorithms Overview
Algorithms for Cell Type Annotation

Algorithms for Cell Type Annotation

The cloud-based cell annotation model was co-developed by 10x Genomics and the Cellarium AI Lab at the Data Sciences Platform of the Broad Institute. The model is in beta. A preprint describing the method is now available on bioRxiv: Accelerating scRNA-seq Analysis: Automated cell type annotation using representation learning and vector search.

When you enable cell type annotation, your data is securely transmitted to 10x Genomics Cloud Analysis. Since your data is leaving your local environment and entering the 10x Genomics domain, it becomes subject to the terms outlined in the 10x Genomics End User License Agreement (EULA). Please review the EULA carefully to understand how your data will be handled and the associated usage terms. Additionally, please only use this feature if there are no restrictions that preclude your data being sent outside your local environment. The availability of automated cell annotation is subject to restrictions based on U.S. or local laws and regulations. See regional restrictions for the list of impacted regions.

Cell Ranger ARC v2.1 and later enable running automated cell type annotation on the 10x Cloud via a cellranger-arc count analysis. It can be applied to Gene Expression data (human and mouse only; not ATAC data) to generate accurate cell type labels. This method assigns cell types by comparing gene expression profiles to annotated reference datasets, avoiding reliance on marker genes or tissue-specific references.

The Pan-Human Azimuth model was developed by the Satija lab as part of the The Human BioMolecular Atlas Program (HuBMAP).

Cell Ranger ARC v2.2 and later enable running the Pan-Human Azimuth model developed by the Satija lab for Gene Expression data (compatible with human samples only; not ATAC data) in a cellranger-arc count analysis.

The first iteration of this model was trained on scRNA-seq and snRNA-seq data from 23 different tissues and 380 different cell types. Cancer datasets were excluded from the training process. Cell types are organized into a unified cell ontology. For more information, see the Pan-Human Azimuth page from the Satija lab.

Unlike the 10x Genomics models described above, the Azimuth model runs locally with the rest of the Cell Ranger ARC pipeline and does not require 10x Cloud access.