Skip to content

spatialLIBD

Interactive visualization and analysis of spatially-resolved transcriptomics in R/Bioconductor — LIBD's data + Shiny package for the human DLPFC Visium atlas and the surrounding infrastructure for adapting SpatialExperiment workflows to Visium HD.

Talk at GBCC2025

  • Speakers: Nicholas J. Eagles, Sarah E. Maguire, et al. (Lieber Institute for Brain Development)
  • Day / session: Day 3 (Wed Jun 25, 2025) — Oral Session 3, Grace Auditorium (chair: Lambda Moses)
  • Talk title: "Computational considerations for analysis of Visium HD: A Bioconductor user's perspective"
  • Slides: TBD
  • Video: GBCC2025 playlist
  • Abstract / schedule: GBCC2025 program

What it does

spatialLIBD ships the LIBD DLPFC Visium dataset and an accompanying Shiny app for interactive layer-annotation, gene-expression overlay, and cluster comparison on SpatialExperiment objects. The Day-3 GBCC talk is about the Visium HD transition — Visium HD's 2-µm bins explode the data-volume and memory cost vs. classic 55-µm spots, and the talk walks through the practical Bioconductor-side adaptations (binning strategies, memory-mapped backends, plotting at scale) needed to keep the workflow tractable.

How it works

Core idea. spatialLIBD is a golem-structured Shiny application backed by SpatialExperiment objects: the SE carries counts, spot coordinates, and image tiles, and the Shiny layer exposes interactive overlays (cluster labels, gene expression, layer annotations) on top of the histology image.

Inputs / outputs. Input is a SpatialExperiment (≥ 1.3.3) — typically built from 10x Genomics Visium SpaceRanger output — plus optional reference annotations (manual layer labels, registration coordinates). Output is the running Shiny app for interactive exploration, plus publication-quality static figures via ggplot2 / plotly / cowplot and downstream cluster/annotation tables.

Key innovation. Pairing the LIBD human DLPFC Visium reference atlas with a reusable visualization app: the same run_app() machinery used for the DLPFC dataset is the substrate any new spatial dataset can plug into, so a lab gets the LIBD-style explorer for its own Visium data by supplying its own SpatialExperiment.

Parameters / API surface worth knowing. - run_app() — launches the Shiny explorer against a supplied SpatialExperiment. - vis_gene() / vis_clus() — static spot-overlay plotting for gene expression or cluster labels. - Spatial-registration helpers (per the "Guide to Spatial Registration" vignette) for aligning new datasets to the LIBD reference. - Multi-gene plotting helpers (per the "Guide to Multi-Gene Plots" vignette).

Canonical example. From the intro vignette: pull the LIBD DLPFC SpatialExperiment via fetch_data(), then run_app(sce) to open the interactive explorer — users can overlay any gene's expression on the H&E image, toggle between manually annotated cortical layers, and compare cluster assignments across samples. The Visium HD talk extends this pattern with binning and memory-aware backends for the 2-µm grid.

Where it fits in the corpus

  • AACR 2026: axis = single-cell & spatial omics; Visium HD adoption is one of the operative storylines for spatial-omics dossiers
  • OSTA (eurobioc entry) — both target spatial transcriptomics; OSTA is the book, spatialLIBD is one of the canonical case-study packages
  • SpatialData (eurobioc entry) — Python-side spatial framework; the Visium HD scaling pressure is shared
  • EuroBioC 2025: no direct Visium HD counterpart talk on the schedule
  • Nextflow Summit 2026: not mentioned

Notes

The talk frames "what changes when bins get small" as a Bioconductor-user problem rather than a new method — pragmatic, useful for anyone migrating an existing spatialLIBD pipeline to HD.