GoliathAlign

GoliathAlign is a standalone toolkit for whole-genome bisulfite alignment. It runs on Linux x86-64 and Linux ARM64. Compute devices are CPU, NVIDIA CUDA, and AMD HIP, selected through Mojo DeviceContext.

The linear command, MojoFq2bamMeth, maps a paired FASTQ against a C→T reference the way bwa-meth and NVIDIA Clara Parabricks fq2bam_meth do, and writes a coordinate-sorted BAM plus a QC JSON report. Align and MojoGiraffe map the same kind of reads onto C→T and G→A genome graphs and write GAF. MethylCall turns that GAF into per-cytosine methylation counts.

Install a checkout with pixi, or unpack the Ubuntu 22.04 host runtime from release 0.1.0-jammy. The command in both cases is methylGrapher.

Topic	Page
Pixi install and external tools	Install
Hosts, devices, and aligner engines	Platforms and engines
Linear BAM aligner	Linear alignment
Graph GAF aligner and methylation calls	Graph alignment
Every command, with inputs and outputs	Command line
v0.1.0-jammy tarball	Release 0.1.0-jammy

Package notes that sit next to the code:

Path	Subject
fq2bam-meth/docs/LINEAR_FQ2BAM_SPEC.md	Linear pipeline, devices, environment variables
fq2bam-meth/docs/LINEAR_ENGINES.md	fm, parity, and speed
fq2bam-meth/docs/LINEAR_PARITY.md	Clara fq2bam_meth concordance harness
fq2bam-meth/docs/BENCHMARK_FQ2BAM_METH.md	Linear wall-clock notes
giraffe/docs/GIRAFFE_SPEC.md	GBZ / GFA GAF contract
giraffe/docs/BENCHMARK_GIRAFFE.md	Giraffe timing notes
giraffe/docs/ROCM_GIRAFFE_GATES.md	AMD device target
giraffe/docs/PHASE0_GH200_ALIGN.md	Why Parabricks giraffe is BAM-only
methylgrapher/docs/BENCHMARK_MCALL.md	MethylCall subset timing
numeric/README.md	Centroid-stream kernels

This site is published at https://goliath-research.github.io/GoliathAlign/. From a checkout, mkdocs build writes site/. Pull requests run that build in the CI docs job and do not deploy. Pushes to main publish with GitHub Pages.