Skip to content

Epigenomics

The epigenomics path processes spatial ATAC-seq and CUT&Tag data generated via DBiT-seq. It takes raw FASTQ and image data through alignment, QC, parameter optimization, and secondary analysis.

How the Workflows fit together

flowchart LR
    FASTQ["FASTQ"]:::startend
    IMG["Input Images"]:::startend
    WT["WT Workflow outputs"]:::startend

    BQC[BarcodeQC]:::process
    PRE[ATX Epigenomic<br/>Preprocessing]:::process
    OPT[optimize ArchR or<br/>optimize snap]:::process
    IMP[ATX impute]:::process
    CLN[clean]:::process
    FF[fragment fixer]:::process
    SEC[create ArchRProject<br/>or ATX snap]:::process
    PLOTS[ATX Epigenomics Plots]:::process
    CMP[compare clusters]:::process
    GLUE[ATX Glue]:::process
    COPRO[ATX CoPro Plots]:::process
    DONE[Analysis Complete]:::startend

    NC{needs<br/>correction?}:::decision
    FA{further<br/>analysis?}:::decision

    IA1["<b>interactive analysis</b><br/>module scoring · annotation<br/>neighborhood analysis<br/>motif logos · track browser"]:::interactive
    IA2["<b>interactive analysis</b><br/>module scoring · annotation<br/>omics comparison · track browser"]:::interactive

    FASTQ -.-> BQC
    FASTQ --> PRE
    BQC -. cleaning / x-talk input .-> PRE
    PRE --> OPT
    IMG --> OPT
    OPT --> NC
    NC -- missing lanes --> IMP --> OPT
    NC -- high lanes --> CLN --> OPT
    NC -- out-of-bounds --> FF --> OPT
    NC -- No --> SEC
    SEC --> PLOTS
    SEC --> FA
    PLOTS -.-> IA1
    PLOTS -. launch from notebook .-> CMP
    FA -- Differential analysis --> CMP
    FA -- No --> DONE
    FA -- CoPro --> GLUE
    WT --> GLUE
    GLUE --> COPRO -.-> IA2

    classDef process stroke:#818cf8,fill:#eef2ff
    classDef decision stroke:#fb923c,fill:#fff7ed
    classDef interactive stroke:#4ade80,fill:#f0fdf4,stroke-dasharray: 5 5
    classDef startend stroke:#a78bfa,fill:#f5f3ff

Walking the flow:

  1. Preprocess. Raw FASTQ goes through ATX epigenomic preprocessing to produce a fragments file. Optionally, run BarcodeQC first — it generates the cleaning and cross-talk correction tables that preprocessing then applies inline.
  2. Optimize. The fragments file plus the spatial images feed optimize archr or optimize_snap, which sweep parameters and produce the first spatial maps of the experiment.
  3. Correct if needed. The spatial maps reveal whether a run needs remediation — ATX impute for missing lanes, clean for over-represented ("high") lanes, or fragment fixer for out-of-bounds coordinates. After correction, re-optimize; once no correction is needed, proceed.
  4. Secondary analysis. create ArchRProject or ATX_snap builds the analysis-ready objects.
  5. Visualize & branch. Results are explored in ATX Epigenomics Plots (interactive module scoring, annotation, neighborhood analysis, motif logos, track browser). From here you can run compare clusters for differential analysis (also launchable from a Plots notebook), finish, or continue to Co-Profiling — ATX Glue combines these outputs with whole-transcriptome results, visualized in ATX CoPro Plots.

When do the correction steps run?

The correction steps above (impute / clean / fragment fixer) only appear as a post-optimization loop when BarcodeQC is not run — in that case, optimization is the first time you get spatial maps, which are what's needed to decide whether correction is required.

  • If BarcodeQC is run, cleaning (and cross-talk correction) is applied during preprocessing, so no post-optimization loop is needed for it.
  • Fragment fixer runs during preprocessing for new runs; only older runs, or fragment files from other pipelines, need it applied separately.

Processing path

Stage Workflow Purpose
Preprocessing ATX epigenomic preprocessing Filter, align, and QC raw reads into a fragments file.
Optimization optimize archr, optimize_snap Sweep dimensionality-reduction / clustering parameters.
Secondary Analysis create ArchRProject, ATX_snap Produce analysis-ready objects (ArchRProject, AnnData).
Plots Epigenomics Plots Interactive visualization of results.

Helper Workflows

Epigenomics-specific, single-step utilities:

  • cram2frags — convert an indexed CRAM alignment into a fragments file (e.g. Ultima runs).
  • atx_convert — convert between Seurat and H5AD.
  • ATX impute — fill in missing lanes.
  • clean — cleaning / cross-talk correction.
  • fragment fixer — screen alignment files for out-of-bounds coordinates.

Advanced Analysis