Azimuth References

Human - PBMC

This PBMC reference dataset was generated as part of the Hao and Hao et al, bioRvix 2020 paper. It is comprised of data from eight volunteers enrolled in an HIV vaccine trial from which three time point samples were taken at day 0, 3, and 7 following vaccination. All 24 samples were processed with a CITE-seq panel of 228 TotalSeq A antibodies to generate single-cell RNA and ADT data. The data were then integrated using metholody described in the pre-print linked above to generate a weighted nearest neighbor (WNN) representation of the RNA and protein data jointly. This WNN representation is used in the Azimuth app to assign celltypes, embed in the reference UMAP, and impute protein levels for the query dataset.

Annotation Details

Human - Motor Cortex

This reference of human motor cortex was generated from data and annotations available from the Allen Institute. Originally described in Bakken et al, bioRxiv 2020, this data was the integrated across the two individual donors to create the Azimuth reference used here. Annotations are provided at the level of class (e.g. GABAergic), subclass (e.g. L2/3 IT), cluster (e.g. Inh L1-2 VIP WNT4), and cross-species cluster (e.g. Sst_6).

Annotation Details

Mouse - Motor Cortex

The data used in this reference was generated as part of the effort by Yao, Liu, Xie, Fischer, et al, bioRxiv 2020 to generate an integrated transcriptomic and epigenomic atlas of the primary motor cortex in mouse. To generate the reference used in Azimuth, we applied a reference-based integration strategy to harmonize the 10x v3 snRNA-seq data across 12 individual mice. Annotations are provided at the level of class (e.g. GABAergic), subclass (e.g. L2/3 IT), cluster (e.g. L2/3 IT_3), and cross-species cluster (e.g. Sst_6)

Annotation Details

Human - Pancreas

This integrated pancreas reference dataset is an aggregate of multiple pancreas single-cell RNA-seq datasets from the literature. It is comprised of data from six different studies, generated using six different single-cell technologies, and integrated into a single reference using Seurat. Details of the integration procedure can be explored in the Snakemake workflow linked above.

Annotation Details

Human - Fetal Development

This reference consists of 378,000 subsampled human fetal cells from Cao et al., 2020. Details of the reference construction procedure can be explored in the Snakemake workflow linked above.

Annotation Details

Human - Lung v1

This reference consists of 65,662 human lung cells (10x Genomics v2) from Travaglini et al. 2020. Cells were sourced from three donor patients and from all lung tissue compartments (bronchi, bronchiole, alveoli) as well as circulating blood.

Annotation Details

Human - Kidney

This reference consists of 64,693 kidney nuclei (10x Genomics snRNA-seq) from healthy reference states generated by investigators in the Human Biomolecular Atlas Program (HuBMAP) and the Kidney Precision Medicine Project (KPMP). The dataset represents 21 samples across 13 patient donors. Samples were derived from nephrectomy, biopsy, or deceased donor tissue, spanning three anatomical regions: medulla, cortex and papilla. The annotations and the markers are provisional (especially levels 2 and 3) and will likely get curated as more data are added to the atlas. More details regarding cell-type specific positive and negative markers can be found in the supplementary tables 4, 5 and 8 here.

Annotation Details

Human - Bone Marrow

This reference consists of 297,627 bone marrow cells from 39 donors and 3 different studies: Oetjen et al., 2018, Granja, Klemm, McGinnis, et al. 2019, and the Human Cell Atlas Immune Cell Census. All samples, which were generated using 10x Genomics v2, were integrated together and used to define a reference UMAP visualization and list of cell type annotations.

Annotation Details

Human - Lung v2 (HLCA)

The Human Lung Cell Atlas (HLCA, Sikkema et al. 2023) includes 2.4M cells of the human lung and nose. Its core (used for mapping), containing only data of healthy/control lung, was built by integrating data of 107 individuals from 14 datasets, including major published datasets as well as unpublished data. The cells in the integrated atlas were re-annotated based on originally published labels and inputs from 6 experts, resulting in consensus on 61 cell types/states in healthy human lung, along with robust marker genes to identify these. The HLCA is freely available for download, exploration, and use as a reference here.

Annotation Details

Human - Tonsil v1

This reference consists of 263,299 cells (10x Genomics 3' v3) from 10 healthy human donors across a range of ages. The celltype.l2 annotations are derived from the original publication while the celltype.l1 assignment and naming has been modified slightly.

Annotation Details

Human - Adipose

This reference consists of 160,075 white adipose cells. scRNA-seq was performed on subcutaneous white adipose tissue from 9 individuals and sNuc-seq was performed on subcutaneous and visceral white adipose tissue from 13 individuals to generate the reference. The celltype.l2 annotations are derived from the original publication while the celltype.l1 assignment and naming has been modified slightly.

Annotation Details

Human - Heart

This reference consists of 656,509 integrated heart cells from data collected in 3 different studies: Hocker et al., 2021, Koenig et al., 2022, Litviňuková et al., 2020. The atlas was constructed using both scRNA-seq and sNuc-seq samples from a total of 63 human donors over a range of ages.

Annotation Details

Human - Liver

This reference consists of 79,492 integrated liver cells from data collected in 5 different studies: Aizarani et al., 2019, Ramachandran et al., 2019, Zhang et al., 2020, Macparland et al., 2018, Payen et al., 2020. The atlas represents scRNA-seq samples from a total of 29 donors integrated together. Certain datasets were generated using modified tissue dissociation protocols enriching for hepatocytes, cholangiocytes, and HSCs, so that the reference represents a broad set of cell types in the liver.

Annotation Details

Human - Tonsil v2

This reference consists of 377,963 cells (10x Genomics 3' v3) from 17 healthy human donors across a range of ages. The celltype.l2 annotations are derived from the original publication while the celltype.l1 assignment and naming has been modified slightly to represent a more general annotation. Datasets for each of the major lineages (e.g. CD4 T cells) can be downloaded and explored through HCATonsilData, which also includes a detailed glossary with the references used to annotate each cell type and state. The markers we include for the celltype.l2 level below were computationally derived and may differ from the markers that were used to annotate the cells. For a detailed list of markers curated by the annotation team, please refer to the glossary.

Annotation Details

Mouse - Pansci

This reference consists of 488,526 cells, with a uniform subsample of 120 cells per subtype from the PanSci mouse organismal aging atlas. Cells were sourced from 14 mouse organs/tissue, including brain, lung, heart, liver, kidney, muscle, stomach, duodenum, jejunum, ileum, colon, brown adipose tissue, inguinal adipose tissue, gonadal adipose tissue, across three different genetic backgrounds: C57BL/6, B6.129S7-Rag1tm1Mom/J, and B6.Cg-Prkdcscid/SzJ. The library for each organ/tissue was constructed using a refined version of the EasySci protocol. Below, we include markers for only the “Main_cell_type” annotation level, which is the most general annotation level in the reference. There are also two more levels of annotation included in the azimuth app.

Annotation Details