Comparing genomes: linear synteny and dotplots

JBrowseR() renders a single linear genome view. For comparative genomics — several genomes stacked, the blocks each pair shares drawn between the rows — JBrowseRApp() drives the full app from a declarative views list. Each entry is list(type = , init = ) — the same vocabulary JBrowse Web serializes into its ?session=spec-… URLs, and the init fields are the view’s own state-model options. There are no R builders for it, so any view type JBrowse has — including one a runtime plugin registers — opens with nothing added to the package.

This vignette stacks four E. coli strains (K12, Sakai, CFT073, NCTC86) tied together by one all-vs-all minimap2 alignment — the same hosted data as the JBrowse all-vs-all synteny tutorial.

The website hosts the same views as live widgets, and everything here also runs on Colab’s R runtime:

library(JBrowseR)

Describe the four assemblies

An assembly is the flat list(name = , uri = ) shorthand JBrowse expands itself — it picks the adapter from the extension and derives the .fai/.gzi index locations, so the URL is all you write. (a list(name=, uri=) is the same thing with the name defaulted from the file.)

base <- "https://jbrowse.org/demos/ecoli_pangenome"
strains <- c("K12", "Sakai", "CFT073", "NCTC86")

assemblies <- lapply(strains, function(s) {
  list(name = s, uri = paste0(base, "/", s, ".fa.gz"))
})

One all-vs-all track between them

A single AllVsAllPAFAdapter track serves every pair from one PAF file. It spans all four assemblies. The track config is a plain list, the same JSON a JBrowse config file would hold.

ecoli_ava <- list(
  type = "SyntenyTrack",
  trackId = "ecoli_ava",
  name = "E. coli all-vs-all (minimap2 PAF)",
  assemblyNames = as.list(strains),
  adapter = list(
    type = "AllVsAllPAFAdapter",
    assemblyNames = as.list(strains),
    pafLocation = list(uri = paste0(base, "/all_vs_all.paf.gz"))
  )
)

Stack them in a synteny view

init$views lists the panels top-to-bottom, one list(assembly = ) each. The four rows have three gaps between adjacent pairs, and each gap draws the same all-vs-all track, so tracks is that trackId once per gap. drawCurves = FALSE draws straight ribbons; minAlignmentLength hides short, noisy blocks.

app <- JBrowseRApp(
  assemblies = assemblies,
  tracks = list(ecoli_ava),
  views = list(
    list(
      type = "LinearSyntenyView",
      init = list(
        views = lapply(strains, \(s) list(assembly = s)),
        tracks = list(list("ecoli_ava"), list("ecoli_ava"), list("ecoli_ava")),
        drawCurves = FALSE,
        minAlignmentLength = 10000
      )
    )
  )
)
app

Four E. coli strains stacked, with the blocks each adjacent pair shares drawn between the rows

In the console this opens a browser tab; in R Markdown it renders inline; in Shiny, pair JBrowseROutput() with renderJBrowseR().

The same PAF as a dotplot

Any one pair opens whole-genome as a dotplot — the long diagonal is the shared backbone, off-diagonal segments are rearrangements.

JBrowseRApp(
  assemblies = assemblies[1:2],
  tracks = list(ecoli_ava),
  views = list(list(
    type = "DotplotView",
    init = list(
      views = list(list(assembly = "K12"), list(assembly = "Sakai")),
      tracks = list("ecoli_ava")
    )
  ))
)

K12 against Sakai as a dotplot: a long shared diagonal with off-diagonal rearrangements

Building the PAF from your own genomes

The hosted all_vs_all.paf.gz comes from concatenating the strains (each contig PanSN-named sample#1#contig) and running minimap2 -c -x asm20 --eqx all pairs. The tutorial walks through generating it and loading per-strain gene tracks alongside.