ATAC-seq

The core enzymatic reaction uses a hyperactive Tn5 transposase loaded with sequencing adapters. This enzyme simultaneously cleaves and tags accessible DNA in a process called tagmentation. Because Tn5 cannot penetrate tightly packed heterochromatin, it preferentially inserts adapters into nucleosome-free regions—typically active promoters, enhancers, and other regulatory elements. This direct in vitro reaction eliminates the need for antibody-based enrichment or crosslinking, distinguishing it from techniques like ChIP-seq.

ATAC-seq identifies multiple classes of regulatory elements in a single experiment:

• Active Promoters: Sharp peaks of accessibility immediately upstream of transcription start sites (TSS).
• Distal Enhancers: Accessible regions located far from gene bodies, often marked by flanking nucleosome arrays.
• Insulators: CTCF-bound boundary elements that partition the genome into topologically associating domains (TADs).
• Nucleosome Positioning: Fragment size distribution analysis reveals the periodicity of phased nucleosomes flanking open regions.

Deep sequencing depth enables digital genomic footprinting—the identification of protected DNA segments where transcription factors (TFs) are bound. When a TF occupies its motif, Tn5 cannot access those specific base pairs, creating a characteristic depletion of insertions within the broader peak of accessibility. This allows simultaneous inference of both chromatin state and TF occupancy from a single assay, providing a functional readout of regulatory activity beyond simple motif prediction.

The protocol has been adapted to the single-cell level, enabling the dissection of cellular heterogeneity within complex tissues. scATAC-seq uses combinatorial cellular indexing or microfluidic partitioning to tag DNA fragments with cell-specific barcodes. This reveals:

• Rare cell populations with distinct regulatory landscapes.
• Regulatory trajectories during differentiation or disease progression.
• Cell-type-specific enhancer usage that bulk assays would average out. The resulting sparse, high-dimensional data requires specialized computational methods for dimensionality reduction and clustering.

Standard analysis involves several key steps:

• Read Alignment: Paired-end reads are mapped to a reference genome using tools like Bowtie2 or BWA.
• Peak Calling: Algorithms such as MACS2 or Genrich identify regions of significant enrichment over background.
• Fragment Size Filtering: Nucleosome-free fragments (<100 bp) are separated from mono-nucleosomal fragments (~180-247 bp) for distinct analyses.
• Differential Accessibility: Tools like DESeq2 or edgeR identify regions with statistically significant changes in accessibility between conditions.
• Motif Enrichment: HOMER or MEME suite identifies overrepresented TF binding motifs within accessible peaks.

ATAC-seq offers distinct practical benefits compared to traditional chromatin accessibility assays:

• Low Input Requirement: Reliable libraries can be generated from as few as 500 to 50,000 cells, making it suitable for rare primary samples.
• Rapid Protocol: The entire library preparation can be completed in under 3 hours, significantly faster than DNase-seq or MNase-seq.
• No Antibody Dependency: Unlike ChIP-seq, it does not require high-quality, validated antibodies for each target protein.
• Dual Readout: Simultaneously provides information on chromatin accessibility, nucleosome positioning, and TF footprinting from a single dataset.

The computational task of identifying distal cis-regulatory elements that activate gene transcription. ATAC-seq data provides the chromatin accessibility signal that is a primary feature for these models.

• Often uses deep learning on histone modification and accessibility data
• Distinguishes active from poised enhancers
• Critical for linking non-coding variants to gene regulation

A computational perturbation method where every nucleotide in an input DNA sequence is systematically mutated to measure the predicted change in a model's output. When applied to models trained on ATAC-seq data, it reveals regulatory motif logic.

• Quantifies the impact of single nucleotide variants on accessibility
• Identifies functional binding sites within open chromatin peaks
• Generates saliency maps for regulatory grammar

A training paradigm where a single neural network is simultaneously trained on multiple related prediction tasks. In the context of ATAC-seq, a model might jointly predict chromatin accessibility, transcription factor binding, and histone modifications.

• Improves generalization through shared representations
• Leverages correlations between epigenomic assays
• Reduces overfitting on sparse individual datasets