LeafViz

Jack Humphrey, David A. Knowles, Yang I. Li 2017-2021

Modified Scott I. Adamson 2026

A lightweight, standalone version of the Leafcutter Visualisation Shiny app.

This was created due to the complexities of installing Leafcutter on local machines.

Installation:

## in R:
install.packages("remotes")
remotes::install_github("jackhump/leafviz")

Docker image now available, created by @NaotoKubota

See https://hub.docker.com/r/naotokubota/leafviz

Running the shiny app

Setting the browser (for Windows only)

First set the browser option (using double backward slash or a single forward slash is due to windows) such as for the main 3 browers: (thanks to bsouthey for this tip)

options(browser="C:\Program Files\Mozilla Firefox\firefox.exe")
options(browser="C:\Program Files (x86)\Microsoft\Edge\Application\msedge.exe")
options(browser="C:\Program Files\Google\Chrome\Application\chrome.exe")

Running app on example dataset:

library(leafviz)
leafviz()

Running leafviz on your own dataset:

Update as of version 0.1.1

You can use new functions to prepare both the database from a gtf and the Rdata object from your leafcutter output.

To find example data, navigate to the example data directory and load up R.

cd example_data/leafcutter_ds
R

Then build the annotation and Rdata object. Note that the output prefix must be in the format "._hg38" for UCSC lookup to be functional.

library(leafviz)

# Prepare annotation for leafviz
gtf2leafcutter(
    gtf_file      = "gencode.v43.basic.annotation_sample.gtf.gz",
    output_prefix = "gencode.v43.basic_hg38_leafviz"
)

# Generate RData object for leafviz
prepare_results(
  counts_file               = "Geuvadis_M_vs_F_perind_numers.counts_sample.gz",
  cluster_significance_file = "cluster_significance.txt",
  effect_sizes_file         = "effect_sizes.txt",
  annotation_code           = "gencode.v43.basic_hg38_leafviz",
  groups_file               = "groups.txt",
  output                    = "results.RData"
)

Then you can load it into leafviz and it will launch:

library(leafviz) # if you haven't loaded it previously
leafviz("<path/to/your_leafcutter_results.RData>")

Note that this should be compatible with data generated from the R version, and the leafcutter-ds version generated for regular or leafcutter2 workflows.

Generating simple reports

Leafviz now allows you to export information to reports. Currently under development.

make_report(infile = "<path/to/your_leafcutter_results.RData>", outfile = "./report.html")

The following output is produced:

<outfolder>/<code>_report.html - a HTML file viewable in your browser, containing the same information as in the summary tab of the LeafViz browser session.

Exporting tables

You can now export the clusters and introns tables to TSV with one command:

export_tables(infile = "<path/to/your_leafcutter_results.RData>", outfolder = ".")

The following outputs are produced:

<outfolder>/<code>_sig_clusters.tsv - the cluster-level information on each significant cluster <outfolder>/<code>_sig_introns.tsv - the intron-level information for each significant cluster

<code> is already set in the RData object

Classify clusters as cassette exons and determine directionality and novelty

classify_clusters(infile = "<path/to/your_leafcutter_results.RData>", outfolder = ".")

Leafviz also contains a function to classify leafcutter junction clusters as cassette exons, if they contain 3 introns with thecorrect topology (2 child (inclusion) introns; 1 parent (skipping) intron)

It then works out the directionality based on the leafcutter model effect sizes for the 3 introns. A skipped cassette exon would have a positive effect size for the parent intron and negative effect sizes for the two child introns, whereas an included exon would have the reverse.

It then uses the annotation status of each intron to determine whether the exon is novel (cryptic exon; the two inclusion introns are novel) or the skipping event is novel (skiptic exon; the skipping intron only is novel); or completely unannotated.

Exon and intron coordinates are then inferred from the junction coordinates.

The following files are produced:

<outfolder>/<code>_classifications.tsv - the classification for each cluster <outfolder>/<code>_summary.tsv - the summary, tallying by classification

code is already set in the RData file.

Options:

full_results - logical TRUE or FALSE - whether to also output the clusters that couldn't be classified in the output files.