Signac TF Footprint Visualization

Purpose

Compute and plot TF footprint profiles directly on a Seurat object with a Signac ChromatinAssay. Output is publication-ready ggplot2 figures showing observed vs expected Tn5 cutting around motif sites, grouped by any obs column (typically cell type or condition).

Use when:

• Producing footprint figures for a manuscript or report
• Confirming binding patterns for a small set of selected TFs
• Inspecting footprints grouped by Seurat cluster / cell type / condition

Upstream: Seurat object with ChromatinAssay, fragments file indexed, motifs added via AddMotifs() Downstream: ggplot2 / patchwork composition for figure panels

Setup Motif Information

library(Signac)
library(Seurat)
library(motifmatchr)
library(JASPAR2020)
library(TFBSTools)
library(BSgenome.Mmusculus.UCSC.mm10)

obj <- readRDS("integrated_object.rds")

# Get motif PWMs (10090 = Mus musculus; 9606 = Homo sapiens)
pwm <- getMatrixSet(
  x = JASPAR2020,
  opts = list(species = 10090, all_versions = FALSE)
)

# One-time, slow: add motif positions to object
obj <- AddMotifs(obj, genome = BSgenome.Mmusculus.UCSC.mm10, pfm = pwm)

Compute and Plot Footprints

# Compute footprint data for specific TFs
obj <- Footprint(
  object     = obj,
  motif.name = c("IRF8", "BATF::JUN", "SPI1", "ZEB1"),
  genome     = BSgenome.Mmusculus.UCSC.mm10,
  in.peaks   = TRUE,      # CRITICAL for speed — restrict to peaks
  upstream   = 250,
  downstream = 250
)

# Plot, grouped by celltype
p <- PlotFootprint(
  obj,
  features = c("IRF8", "BATF::JUN", "SPI1"),
  group.by = "celltype"
)
p + patchwork::plot_layout(ncol = 1)

ggsave("footprint_profiles.pdf", width = 8, height = 12)

Interpreting Signac Footprint Plots

Top panel (Observed):

• Inverted "V" pattern = active TF binding (flanking enrichment + central depletion)
• Flat profile = no evidence of TF occupancy
• Deeper central valley = stronger / more frequent binding

Bottom panel (Expected / Background):

• Shows expected Tn5 cutting from sequence bias alone
• Compare observed vs expected to distinguish true footprints from Tn5 preference artifacts

Comparing conditions:

• Higher flanking signal + deeper valley = more active binding in that condition
• Similar profiles across conditions = constitutive binding (not condition-specific)

Output Interpretation

• Flanking enrichment: Height of the "shoulders" around the motif center
• Central depletion: Depth of the "valley" at position 0 (binding site)
• Footprint depth: Difference between shoulder height and valley depth

Common Pitfalls

Issue	Cause	Fix
Footprint() extremely slow	Many TFs × large fragment file	Always set in.peaks = TRUE; limit motif.name to selected TFs
Profiles look identical across groups	Group sizes very imbalanced	Subsample large groups or check group.by mapping
Empty plot for a TF	Motif not added to object	Re-run AddMotifs(); verify TF is in pwm set
Species mismatch	Wrong JASPAR species code	10090 = mouse, 9606 = human
Genome style mismatch	chr1 vs 1 between fragments and BSgenome	Ensure seqlevelsStyle() matches across object and genome

Performance Tips

• Use in.peaks = TRUE (massive speedup vs whole-genome scan)
• Pre-restrict motif.name to the TFs you care about
• Cache the Footprint()-augmented object; recomputing per figure is wasteful
• For quantitative differential analysis at scale, switch to tobias-footprint-bindetect

Resources

• Signac footprint vignette: https://stuartlab.org/signac/articles/footprint
• Signac GitHub: https://github.com/stuart-lab/signac
• JASPAR 2024: https://jaspar.elixir.no/

---

When not to use

• Do not use for quantitative differential binding — Signac Footprint() is visualization-first; use tobias-footprint-bindetect.
• Do not use on very large fragment files for many TFs — slow; restrict with in.peaks = TRUE or use TOBIAS.

---

See also

• tobias-footprint-bindetect
• hint-atac-differential-footprint
• signac-chromatin-analysis