multiomeR Manual

Start here

What is multiomeR?

multiomeR is a targets-based workflow for single-nucleus 10x Genomics Multiome ATAC + Gene Expression data. It starts from cellranger-arc count outputs and is meant to be adapted to your own data, compute setup, and biological questions.

NoteUnder peer review

The manuscript describing multiomeR is under peer review (link to come). The running guide shows what to inspect at each checkpoint, but acceptance criteria still depend on the tissue, study design, and intended use. Report problems or questions through GitHub issues.

Workflow at a glance

The primary module processes each GEM well, combines selected GEM wells into an aggregation, clusters and labels cell types in its gene-expression (GEX) and ATAC data, and integrates both modalities with weighted nearest neighbors (WNN). Three optional modules extend a completed aggregation with differential analyses, genetic enrichment for human traits, or peak–gene correlation.

How this manual is organized

Read the first four parts in order, as far as your analysis needs them:

• Try the public demo: install multiomeR, then run and inspect a small example analysis.
• Analyze your own data: check your inputs, then configure, run, and review the primary module one checkpoint at a time.
• Add an optional analysis: run differential analyses, genetic enrichment, or peak–gene correlation on a completed aggregation.
• Operation and scaling: run locally or on a scheduler, and troubleshoot failed or outdated targets.

Look up the rest when you need it:

• Reference: the parameter browser, the configuration tables, the output files and the output gallery, with an example of each plot.
• In depth (advanced): what each analysis does, from the primary module to the optional modules, how the reimplemented algorithms were validated, and how to change the pipeline itself. The running guide and module pages link to the matching methods section.

Your first analysis

Start with the public demo, even if you plan to analyze your own data. It runs one aggregation of two human GEM wells with supplied configuration and shows how to read its cell metadata and multimodal Seurat object.

You need basic R skills, a Linux terminal, and a machine with sufficient memory and disk space. You do not need to know how to write a targets pipeline. Run blocks labeled Bash in the terminal and blocks labeled R in the R session opened during installation, both from the repository folder unless stated otherwise. A block labeled with a file name, such as cfg_aggregations.yaml, shows content for that file.

Terms used in this manual

• A GEM well is one 10x Chromium chip channel and its cellranger-arc count output directory. It may contain nuclei from several donors, the individuals identified by donor_id.
• An aggregation is a joint analysis of one or more GEM wells.
• A checkpoint is one of the eight stages of the primary module, each ending with plots to review before you continue. Its name, such as 8_multimodal_QC, appears in target names and plot folders.
• A target is a named result, such as a metadata table, matrix directory, or plot. You request the targets you want; targets builds the earlier results they depend on, in order, and reuses those that are up to date. Most target names end with their GEM well or aggregation, as in multimodal_Seurat_object.8_multimodal_QC.immune_human_2x.
• The store is the folder where targets keeps results and the records needed for reruns. Paths in this manual write it as <store>.

For a small worked introduction to targets, see the targets walkthrough.

Citing multiomeR

If you use multiomeR, cite the release you used: its citation file gives the authors, and the release’s Zenodo record gives its DOI. multiomeR is built on targets (Landau, 2021) and BPCells (Parks and Greenleaf, 2025), so please cite them too. Also cite the methods behind the results you report: the In depth pages, starting with the primary module, cite each method where it is used.

Continue to Install and prepare the demo.