Explore, Diagnose and Filter Genomic Data • radr

radr explores, diagnoses, visualizes, and filters individual genomic data. It works primarily with GDS files and objects created by genometranslator.

The two packages have deliberately different responsibilities:

genometranslator reads, standardizes, and writes genomic formats;
radr investigates data quality and applies explicit filters.

explore_genomes() offers a guided first exploration. It is not a universal filtering recipe: established analyses should use selected detect_*() and filter_*() functions in an order justified for the dataset.

Installation

Starting from a basic R installation, install the CRAN installer and required Bioconductor foundation first:

install.packages(c("BiocManager", "remotes"))

BiocManager::install(c(
  "gdsfmt",
  "Rsamtools",
  "SeqArray"
))

remotes::install_github("thierrygosselin/tgbase")
remotes::install_github("thierrygosselin/genometranslator")
remotes::install_github("thierrygosselin/radr")

The Remotes field in radr’s DESCRIPTION records its GitHub dependencies, but the explicit sequence above makes a clean installation easier to diagnose.

Check the installation without changing it:

radr::radr_dependencies()

The returned table distinguishes required components from optional components and states which workflow uses each optional dependency.

Optional R packages

Install only what is needed for the planned analysis:

# LD, linkage pruning, and IBS calculations on GDS
BiocManager::install("SNPRelate")

# Tidy-data distances, sex markers, and fast IBM PNG rendering
install.packages(c("amap", "quantreg", "ragg"))

Function documentation identifies its additional dependencies. An optional package is not required merely to install or load radr.

Optional command-line tools with Conda

Some VCF-level filters use bcftools, while run_bayescan() uses the BayeScan executable. These are programs, not R packages. A shared Conda environment can provide both:

conda create --name genomics --channel conda-forge --channel bioconda bcftools bayescan=2.1
conda activate genomics
bcftools --version
bayescan --help

For an existing environment:

conda activate genomics
conda install --channel conda-forge --channel bioconda bcftools bayescan=2.1

Start R or RStudio from the activated environment, then verify visibility:

Sys.which(c("bcftools", "bayescan"))
radr::radr_dependencies()
radr::check_bayescan()

A minimal workflow

Import and standardize a genomic file with genometranslator, then diagnose and filter the resulting GDS with radr:

genome <- genometranslator::read_genome(
  data = "individuals.vcf.gz",
  strata = "strata.tsv"
)

# Preserve the original sample and marker order for the first missingness view
ibm <- radr::detect_ibm(
  data = genome,
  filename = "initial_missingness.png"
)

# Guided exploration for a new dataset
screened <- radr::explore_genomes(data = genome)

Filtering order should follow what is known about the project rather than a fixed recipe. Filtering individuals first changes marker statistics, while filtering markers first changes individual statistics.

The getting-started vignette develops two contrasting examples: a marker-first workflow for a noisy callset and a sample-first workflow for known sequencing failures. It also explains how to return to guided exploration after correcting a known problem and how to compare alternative filtering orders reproducibly.

Citation

citation("radr")
packageVersion("radr")

Until a dedicated publication or DOI is available, cite the version and, for a development build, record the Git commit and access date:

Gosselin, T. (2026). radr: Explore, diagnose and filter genomic data. R package version 0.0.0.9000. https://github.com/thierrygosselin/radr. Accessed 2026-08-26.

Website and support

Documentation and articles are available at https://thierrygosselin.github.io/radr/. Report problems or request features through the GitHub issue tracker.