Reference

Purpose

Reference resources help interpret single-cell results.

They can include:

• curated marker gene tables
• cell type annotation references
• public expression references
• ontology or label mapping tables
• locally saved .rds, .rda, or .csv files

References are supporting evidence. They should not replace marker inspection and biological interpretation.

Project Location

Reference files used by this project can be stored under:

assets/data/

For example, local celldex references are stored in:

assets/data/celldex/

Current files:

assets/data/celldex/BlueprintEncodeData.rds
assets/data/celldex/DatabaseImmuneCellExpressionData.rds
assets/data/celldex/HumanPrimaryCellAtlasData.rds
assets/data/celldex/ImmGenData.rds
assets/data/celldex/MonacoImmuneData.rds
assets/data/celldex/MouseRNAseqData.rds
assets/data/celldex/NovershternHematopoieticData.rds
assets/data/celldex/celldex_reference.rda

This keeps reference files separate from analysis notebooks.

celldex References

celldex provides curated reference datasets commonly used with cell annotation workflows such as SingleR.

The local celldex folder contains seven individual reference datasets and one combined .rda file.

File	General use
BlueprintEncodeData.rds	human immune and hematopoietic reference
DatabaseImmuneCellExpressionData.rds	immune cell expression reference
HumanPrimaryCellAtlasData.rds	broad human primary cell reference
ImmGenData.rds	mouse immune cell reference
MonacoImmuneData.rds	human immune cell reference
MouseRNAseqData.rds	mouse RNA-seq reference
NovershternHematopoieticData.rds	human hematopoietic reference
celldex_reference.rda	combined local reference object file

Choose references by species, tissue, platform, and the biological question.

Load References

Load one reference:

hpca_ref <- readRDS(
  file = "assets/data/celldex/HumanPrimaryCellAtlasData.rds"
)

Load the combined reference file:

load("assets/data/celldex/celldex_reference.rda")

After loading, check available labels:

colnames(SummarizedExperiment::colData(hpca_ref))

Common label columns include broad labels and finer labels, depending on the reference.

Reproducibility

For reproducible annotation, keep track of:

• reference source
• reference version or download date
• file path
• label column used
• species and tissue compatibility

Locally saved references make annotation less dependent on remote download behavior.

Note

Reference data should be used together with cluster markers, known biology, sample metadata, and visualization.

The final cell type label should be treated as an interpretation, not as a direct output of the reference file alone.