snATAC-seq and 10x Multiome analysis: from fragments to regulatory networks
snATAC-seq and 10x Multiome analysis: fragment QC, TSS enrichment, iterative LSI, per-cell-type peaks, chromVAR, footprints and SCENIC+ networks.
Key takeaways
- Judge snATAC quality by TSS enrichment, fragment size periodicity and FRiP, not read counts alone.
- Call peaks per cell type on pseudobulk replicates; whole-dataset peak calling misses rare-cell elements.
- Motif enrichment says which TF family could bind. Combine chromVAR with expression and footprints to nominate drivers.
- Multiome pairing turns enhancer-gene correlation into direct, within-cell evidence.
Chromatin accessibility shows which regulatory elements are open in each cell, and by implication which transcription factors are active. Single-nucleus ATAC-seq (snATAC-seq) and 10x Multiome (RNA + ATAC in the same nucleus) are powerful and unforgiving. The data is sparser than scRNA-seq, nearly binary per region, and dominated by technical signal if quality is not controlled.
Step 1: From reads to fragments
Align with Cell Ranger ATAC/ARC or chromap, then work from the fragments file: one line per Tn5-cut fragment, with cell barcode. Remove mitochondrial and duplicate fragments, and make sure the genome build and annotation match any RNA data you will integrate.
Step 2: Fragment-level QC
- TSS enrichment: accessible chromatin concentrates at promoters. Low TSS enrichment means mostly background.
- Fragment size distribution: good libraries show nucleosome-free fragments under 150 bp plus mono- and di-nucleosome peaks. Flat distributions indicate poor nuclei.
- Unique fragments per cell: sets the depth available for peak-level analysis.
- FRiP and blacklist ratio: fraction of fragments in peaks versus artifact regions.
- Doublets: ArchR's doublet scores, AMULET (which uses the fact that a diploid nucleus has at most two fragments per locus), or scDblFinder in ATAC mode.
Step 3: Dimensionality reduction and clustering
ATAC data is usually transformed with TF-IDF and reduced with latent semantic indexing (LSI). ArchR's iterative LSI first clusters on the most accessible tiles, then refines on cluster-variable features, which separates cell types better than a single pass. Integrate samples with Harmony on LSI components. The first LSI component often correlates with sequencing depth and is usually excluded.
Step 4: Annotate cell types
- Gene activity scores (accessibility in and around gene bodies) give a rough expression proxy.
- Label transfer from scRNA-seq (matched or atlas) is far more reliable. In Multiome, the RNA modality annotates cells directly.
- A biologist reviews labels with known cell-type-specific elements.
Step 5: Peak calling done right
Call peaks on pseudobulk groups per cell type (MACS2/3), ideally with pseudo-replicates, then merge into one reproducible peak set with fixed-width peaks (the iterative overlap approach). Calling peaks on all cells together lets abundant cell types dominate and misses elements specific to rare populations.
What our experts avoid: declaring a transcription factor a "driver" from motif enrichment alone. Many TFs share a motif family (dozens of bZIP or ETS factors). We combine chromVAR deviations with TF expression, footprinting and target-gene links before naming candidates.
Step 6: Motifs, footprints and TF activity
- chromVAR computes per-cell motif accessibility deviations, a robust way to compare TF activity between cell types and conditions.
- Footprinting (TOBIAS, ArchR) looks for protected TF-bound sites within accessible regions. Correct for Tn5 sequence bias first.
- Differential accessibility between conditions uses pseudobulk per cell type, as for RNA.
Step 7: Link regulatory elements to genes
- Co-accessibility (Cicero-style) connects distal elements to promoters.
- Peak-to-gene linkage correlates accessibility with expression, across matched cells in Multiome or across integrated groups in unpaired data.
- GWAS enrichment tests whether disease-associated variants fall in cell-type-specific peaks. That is a powerful way to nominate the cell types where disease genetics acts.
Step 8: Regulatory networks with Multiome
With paired RNA and ATAC, SCENIC+ builds enhancer-driven regulons (TF, then enhancer, then target gene), and WNN or MultiVI produce joint embeddings that use both modalities. This is where Multiome pays off: claims rest on within-cell measurements rather than cross-dataset inference.
How OmicsQuest can help
Our snATAC-seq and Multiome services deliver QC reports, reproducible peak sets, cell-type bigWig tracks, ranked TF drivers and regulatory networks. For bulk ATAC-seq, ChIP-seq and CUT&RUN, see bulk epigenomics.
Frequently asked questions
What is a good TSS enrichment score for snATAC-seq?
As a rough guide, cells with TSS enrichment above about 4 to 6 (depending on the calculation method and tissue) are usually considered good quality, and high-quality datasets often have medians well above that. We set thresholds per sample from the distribution, rather than using one fixed number.
ArchR or Signac?
Both are excellent. ArchR is fast and memory-efficient for large datasets, with built-in iterative LSI, peak calling and trajectory tools. Signac integrates tightly with Seurat and its multimodal workflows. We use whichever fits the project and the client's existing ecosystem.
Work with OmicsQuest. Related services:
References
- Granja JM, Corces MR, Pierce SE, et al. ArchR is a scalable software package for integrative single-cell chromatin accessibility analysis. Nat Genet. 2021;53:403-411.
- Stuart T, Srivastava A, Madad S, Lareau CA, Satija R. Single-cell chromatin state analysis with Signac. Nat Methods. 2021;18:1333-1341.
- Zhang Y, Liu T, Meyer CA, et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 2008;9:R137.
- Aibar S, Gonzalez-Blas CB, Moerman T, et al. SCENIC: single-cell regulatory network inference and clustering. Nat Methods. 2017;14:1083-1086.
- Hao Y, Stuart T, Kowalski MH, et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat Biotechnol. 2024;42:293-304.
- Heumos L, Schaar AC, Lance C, et al. Best practices for single-cell analysis across modalities. Nat Rev Genet. 2023;24:550-572.