fgumi
fgumi (Fulcrum Genomics Unique Molecular Indexing tools) is a high-performance suite of command-line tools for processing UMI-tagged next-generation sequencing data. Written in Rust, it provides UMI extraction from FASTQ files, read grouping by UMI with configurable assignment strategies, UMI-aware deduplication, simplex and duplex consensus calling, CODEC consensus calling, quality filtering of consensus reads, and overlapping read-pair clipping. fgumi is the intended successor to the Scala-based fgbio toolkit for UMI processing, targeting significantly higher throughput on multi-core systems. It is published on Bioconda and documented at https://fgumi.readthedocs.io.
Warning — Research preview
fgumi is currently a research preview. The Fulcrum Genomics team targets June 2026 for recommending fgumi over fgbio for production use. Verify fitness for your application before deploying in a clinical or production pipeline.
When you’d use it
Use fgumi when your sequencing library includes unique molecular identifiers and you need to group reads by UMI, call simplex or duplex consensus sequences, or remove PCR duplicates in a UMI-aware manner. It handles the standard commercial UMI library preparations (IDT xGen, KAPA, Twist, QIAseq, and others) and the CODEC protocol for duplex sequencing. fgumi is designed to be run after alignment with bwa-mem3 (or bwa-mem2) and before downstream variant calling or methylation analysis.
How it relates to bwa-mem3
fgumi and bwa-mem3 are sibling projects maintained by Fulcrum Genomics and are designed to work together in the same alignment-and-consensus pipeline. bwa-mem3 provides the aligned BAM that fgumi takes as input for grouping and consensus calling. The two projects share build and documentation conventions (mdbook on Read the Docs, Fulcrum theme, conventional commits) and are benchmarked together in the fgumi-benchmarks internal dataset suite. The intended integration path for in-process alignment within fgumi is bwa-mem3-rs, the Rust bindings for bwa-mem3.
Links
- GitHub: https://github.com/fulcrumgenomics/fgumi
- Docs: https://fgumi.readthedocs.io
- License: MIT
See also: bwa-mem3-rs · Aligning short reads (mem) · Best Practices — Multi-sample workflows · bwa-mem3-bench