Skip to content

plyxp

rlang data masks for SummarizedExperiment — a Bioconductor package that lets users evaluate dplyr-style unquoted expressions against SummarizedExperiment objects across rowData, colData, and assay contexts, without converting to a rectangular data.frame first.

Talk at GBCC2025

  • Speakers: Justin T Landis, Michael I Love (UNC Chapel Hill)
  • Day / session: Day 4 (Thu Jun 26, 2025) — Oral Session 5, Grace Auditorium (chair: Scott Cain)
  • Talk title: "plyxp: reimagining dplyr syntax for SummarizedExperiment objects"
  • Slides (Zenodo DOI): TBD — Zenodo deposits typically 2–4 weeks post-conference
  • Video: GBCC2025 playlist
  • Abstract / schedule: GBCC2025 program

What it does

plyxp uses rlang data masks to make a SummarizedExperiment behave like a tidy data context for dplyr verbs — mutate(), filter(), summarise() — across the three logical "frames" of an SE: row metadata, column metadata, and assay matrices. The user keeps the SE structure (no flatten / pivot round-trip) but writes expressions in dplyr syntax. It's a syntactic bridge, not a replacement for the underlying S4 contract.

How it works

Core idea. plyxp wraps a SummarizedExperiment in a plyxp object that exposes three rlang data masks — one each for the assay matrix, rowData, and colData — and dispatches standard dplyr verbs so expressions evaluate against the appropriate context without flattening the SE to a tibble.

Inputs / outputs. Input is a SummarizedExperiment (the vignette uses airway). Output is a plyxp object that retains the SE structure; se() extracts the underlying SummarizedExperiment back out when the user is done piping.

Key innovation. Context routing inside a single mutate()/filter() call. Rather than an @-prefix grammar, plyxp uses contextual helper functions: rows() directs an expression to rowData, cols() directs it to colData, and bare expressions operate on the assay matrix — so one piped verb can write to all three frames simultaneously.

Parameters / API surface worth knowing. - new_plyxp(se) — constructor wrapping an SE in the plyxp context. - rows(...) / cols(...) — helpers that scope expressions to rowData / colData; bare expressions hit assays. - Supported verbs: mutate, select, summarize, pull, filter, arrange — standard dplyr API. - se() — accessor that returns the underlying SummarizedExperiment.

Canonical example. From the vignette: xp <- new_plyxp(airway); xp |> mutate(log_counts = log1p(counts), cols(treated = dex == "trt"), rows(new_id = paste0("gene-", gene_name))). A single piped mutate() adds a log-transformed assay, a treatment indicator on colData, and a renamed gene ID on rowData — three frames updated in one expression.

Where it fits in the corpus

  • Tidyomics (entry) — same axis (tidy syntax for Bioc data classes); plyxp + tidyomics together signal the maturity shift in how Bioc users interact with their data
  • DeeDeeExperiment (entry) — related "extensions to SummarizedExperiment" pattern; DDE adds result-table slots, plyxp adds expression-language ergonomics — orthogonal but composable
  • AACR 2026: plumbing rather than science; relevant insofar as AACR analysts increasingly hand off SE/SCE objects in tidy-friendly form
  • ISMB 2026: scaffold

Notes

Mike Love's group (UNC) is bridging Bioconductor's S4 world to dplyr / tidyverse syntax — plyxp is the SummarizedExperiment-specific surface, and Tidyomics is the broader umbrella. Together they signal a maturity shift: Bioc data classes are the ground truth, but the user-facing language is the tidyverse. Mature package (Bioconductor 3.23, v1.6.1).