Bioinformatics tools that respect your hardware
irit is the Indonesian and Javanese word for frugal: getting the most out of modest resources.
Measured on a laptop
Section titled “Measured on a laptop”leanqc against FastQC 0.12.1 on one file of 1 million reads, on an AMD Ryzen 5 PRO 4650U laptop.
- leanqc
- 2.5 s
- FastQC 0.12.1
- 7.0 s
- leanqc
- 16 MB
- FastQC 0.12.1
- 483 MB
On a GitHub-hosted runner the same comparison covered 40 module verdicts across four files, and none differed from FastQC. How each number was taken is in the benchmarks.
Ten tools, one job each
Section titled “Ten tools, one job each”Sizes are the Linux x86_64 binaries of 0.5.0.
sheetgen- Builds a pipeline’s sample sheet from a folder of FASTQ files and checks it against the pipeline’s own schema.
- 6.4 MB
leanqc- Read quality control with an offline HTML report, giving the verdicts FastQC gives; long-read views too.
- 1.8 MB
isotable- Joins QC, assembly, typing and classification outputs (fastp, leanqc, QUAST, Shovill, Flye, CheckM2, GTDB-Tk, CoverM, mlst, AMRFinderPlus, Kleborate and more) into one graded row per isolate or MAG.
- 3.1 MB
dblock- Downloads in parallel, verifies and pins reference databases such as GTDB-Tk, CheckM2, Kraken 2, Bakta and AMRFinderPlus, with a lock file.
- 5.8 MB
seqplan- Plans a sequencing run: isolates per run and cost per isolate at a target coverage, from cited kit outputs.
- 1.0 MB
irit-mcp- Gives AI agents these tools and wrapped ones (Shovill, Flye, Medaka, MEGAHIT, MetaBAT2, CheckM2, GTDB-Tk, AMRFinderPlus, Kleborate, …) as checked tools, and records every call in a replayable run record.
- 6.9 MB
cooccur- Which genomes rise and fall together across samples, with permutation p-values and false-discovery control.
- 1.0 MB
fieldsheet- Offline phone worksheet for sample metadata in the field; exports a sample sheet for sheetgen and isotable.
- web app
Install
Section titled “Install”One command
Section titled “One command”Linux and macOS, no sudo (into ~/.local/bin):
curl -fsSL https://raw.githubusercontent.com/dzulfikar08/irit/main/scripts/install.sh | shDownload a release
Section titled “Download a release”Get the archive for your system from the latest release:
Linux x86_64 and ARM64, macOS Intel and Apple Silicon, or Windows. Unpack it and put the binaries on
your PATH. The Linux builds are statically linked, so they run on old distributions too.
Build with Cargo
Section titled “Build with Cargo”cargo install \ --git https://github.com/dzulfikar08/irit \ sheetgen leanqc isotable dblock seqplan cooccur irit-mcp gtdbprepOr run cargo build --release in a checkout; the binaries land in target/release/.
Read next
Section titled “Read next”- Introduction: why these tools exist and the rules they follow.
- Benchmarks: every measurement, the machine and the test data.
- Roadmap: what is done, what is next, and what is further out.
- Tested on real data: an AI agent redoes published genome papers (and a second), nanopore isolates in 8 GB, metagenome to MAGs in 8 GB and how much sequencing depth is enough.
- Two microbes, one dish: an interactive teaching toy on how two species compete for one resource.