Single-Cell ENCODE Data

When to Use

• User wants to find or analyze single-cell data (scRNA-seq, scATAC-seq, snRNA-seq) from ENCODE
• User asks about "single-cell", "scRNA-seq", "scATAC-seq", "cell type annotation", or "single-nucleus"
• User needs to integrate ENCODE single-cell data with bulk epigenomic profiles
• User wants to identify cell-type-specific regulatory elements from single-cell chromatin accessibility
• Example queries: "find scRNA-seq data in ENCODE for brain", "what snATAC-seq is available?", "integrate single-cell with bulk ChIP-seq"

Help the user find and work with ENCODE single-cell genomics data, understand quality limitations relative to bulk assays, and integrate single-cell with bulk ENCODE profiles for cell-type-resolved regulatory analysis.

Literature Foundation

Available Single-Cell Assays in ENCODE

Step 1: Search for Single-Cell Data in ENCODE

Search for scRNA-seq and scATAC-seq experiments in the tissue of interest:

# Single-cell RNA-seq
encode_search_experiments(
    assay_title="scRNA-seq",
    organ="pancreas",           # user's tissue of interest
    biosample_type="tissue",
    limit=50
)

# Single-cell ATAC-seq
encode_search_experiments(
    assay_title="snATAC-seq",
    organ="pancreas",
    biosample_type="tissue",
    limit=50
)

If no results, try broader search terms:

encode_search_experiments(search_term="single cell RNA", organ="pancreas", limit=50)
encode_search_experiments(search_term="single cell ATAC", organ="pancreas", limit=50)

Check facets first to understand what organs have single-cell data:

encode_get_facets(assay_title="scRNA-seq")
encode_get_facets(assay_title="snATAC-seq")

Present a summary to the user showing:

• Number of scRNA-seq and scATAC-seq experiments found
• Organs/tissues represented
• Platforms used (10X Chromium, Smart-seq2, Drop-seq)
• Labs contributing data
• Number of unique donors/biosamples

Step 2: Understand ENCODE Single-Cell Data Structure

scRNA-seq Files

Use encode_list_files to see what is available per experiment:

encode_list_files(
    experiment_accession="ENCSR...",
    assembly="GRCh38",
    preferred_default=True
)

Typical file hierarchy:

• FASTQ (output_type="reads"): Raw sequencing reads with cell barcodes and UMIs
• Gene quantifications (output_type="gene quantifications", format TSV): Count matrices (genes x cells) after ENCODE uniform pipeline processing
• Filtered counts (output_type="filtered feature barcode matrix"): Post-QC cell-filtered matrices ready for analysis
• h5ad: AnnData format when available (convenient for Scanpy workflows)

scATAC-seq Files

encode_list_files(
    experiment_accession="ENCSR...",
    file_format="bed",
    assembly="GRCh38"
)

Typical file hierarchy:

• FASTQ (output_type="reads"): Raw reads with cell barcodes
• Fragments (TSV/BED): Fragment files with cell-barcode assignments — the primary input for ArchR/Signac
• Peaks (BED narrowPeak): Aggregate peak calls across all cells (pseudo-bulk)
• Cell assignments: Barcode-to-cluster or barcode-to-cell-type mapping files

Key difference: scATAC-seq data is extremely sparse at the single-cell level. Most analyses operate on the fragment file, not on per-cell peak calls.

ENCODE Blacklist filtering (required for scATAC-seq): Before any downstream analysis of scATAC-seq peaks or fragments, remove reads/peaks overlapping ENCODE Blacklist regions (Amemiya et al. 2019, Scientific Reports, 1,372 citations). These regions produce artifactual signal in chromatin accessibility assays and inflate per-cell quality metrics (TSS enrichment, FRiP). Both ArchR and Signac apply blacklist filtering by default when provided, but verify it is active. Download blacklists from Boyle-Lab/Blacklist:

• Human GRCh38: hg38-blacklist.v2.bed.gz
• Mouse mm10: mm10-blacklist.v2.bed.gz

Step 3: Assess Single-Cell Quality (ENCODE-Specific Considerations)

Check experiment-level quality:

encode_get_experiment(accession="ENCSR...")

Quality Metrics for scRNA-seq

Quality Metrics for scATAC-seq

ENCODE Audit Flags

Apply the same audit hierarchy as bulk data:

• ERROR: Avoid unless no alternative
• NOT_COMPLIANT: Usable with caveats
• WARNING: Generally safe; document
• INTERNAL_ACTION: DCC processing notes; usually not a concern

Track passing experiments:

encode_track_experime