hint-atac-differential-footprint

HINT-ATAC (rgt-hint) HMM-based TF footprint detection and multi-condition differential footprinting. Use when comparing TF activity across 3+ conditions at once and you want overview heatmaps from one

SKILL.md

GitHub

HINT-ATAC Multi-Condition Differential Footprinting

Purpose

HINT-ATAC (part of the RGT suite) uses an HMM to learn footprint states from ATAC-seq Tn5 cuts and supports multi-condition differential footprinting in a single command — its main advantage over TOBIAS BINDetect, which is pairwise only.

Use when:

  • Comparing TF binding across 3+ cell types or conditions simultaneously
  • Producing overview heatmaps of TF activity dynamics
  • Bias-corrected bigwig tracks for genome browsers (rgt-hint tracks)

Upstream: Peak calling (MACS2/MACS3), pseudobulk BAM generation Downstream: Multi-condition activity heatmaps, scatter plots of TF dynamics

Installation

python3.9 -m venv /opt/venvs/rgt
source /opt/venvs/rgt/bin/activate
pip install rgt

# Download genome data (REQUIRED for bias correction)
python -c "from rgt.helper import download_genome; download_genome('mm10')"

rgt-hint footprinting --help

Workflow

BAM + Peaks → footprinting → motifanalysis matching → differential → heatmaps

Step 1: Footprint Calling

# Per celltype pseudobulk BAM
rgt-hint footprinting \
  --atac-seq \
  --paired-end \
  --organism=mm10 \
  --output-location=./footprints/ \
  --output-prefix=${CELLTYPE} \
  ${CELLTYPE}.bam ${CELLTYPE}_peaks.narrowPeak

Critical: --organism=mm10 requires RGT genome data downloaded first (see Installation).

Step 2: Motif Matching

rgt-motifanalysis matching \
  --organism=mm10 \
  --output-location=./motif_matching/ \
  --input-files ./footprints/${CELLTYPE}.bed

Step 3: Differential Footprinting (Multi-Condition)

# Compare multiple cell types in one run
rgt-hint differential \
  --organism=mm10 \
  --bc \
  --nc 64 \
  --mpbs-files=./motif_matching/cDC1_mpbs.bed,./motif_matching/cDC2_mpbs.bed,./motif_matching/Mac_mpbs.bed \
  --reads-files=cDC1.bam,cDC2.bam,Mac.bam \
  --conditions=cDC1,cDC2,Mac \
  --output-location=./diff_footprints/ \
  --output-prefix=celltype_comparison

Outputs:

  • *_statistics.txt: Tag count and protection score per condition per TF
  • Lineplots/: ATAC-seq profiles per TF per condition
  • Scatter plot of TF activity dynamics across conditions

Step 4: Bias-Corrected Tracks (Genome Browser)

rgt-hint tracks \
  --bc \
  --bigWig \
  --organism=mm10 \
  ${CELLTYPE}.bam ${CELLTYPE}_peaks.narrowPeak \
  --output-prefix=${CELLTYPE}_BC

Output Interpretation

  • Protection score: Higher = stronger footprint (more TF occupancy)
  • Tag count score: Reads around motif sites (library-size dependent — normalize before comparing libraries)
  • Statistics file: Per-TF activity scores across all conditions; suitable for heatmap input
  • Activity scatter: TFs with high condition-specific activity appear as outliers

Common Pitfalls

Issue Cause Fix
--organism=mm10 errors Genome data not downloaded python -c "from rgt.helper import download_genome; download_genome('mm10')"
Library-size confounds activity Tag count not normalized Compare protection scores, or downsample BAMs to matched depth
Slow differential step Single-threaded by default Set --nc 64 to parallelize
Motif name mismatch with TOBIAS/ChromVAR Different PWM sets Use a shared motif database across tools

Performance Tips

  • --nc controls parallelism in rgt-hint differential — use it
  • For >5 conditions, expect heavy memory use; split into thematic groups if needed
  • Pre-compute motif matching once per peak set, reuse across condition combinations

Resources


When not to use

  • Do not use for single-cell TF activity — HINT-ATAC requires pseudobulk BAMs.
  • Do not use for quantitative pairwise contrasts — use tobias-footprint-bindetect.

See also

  • tobias-footprint-bindetect
  • signac-footprint-visualization
  • chromvar-motif-accessibility