rrblup-rs

A Rust implementation of rrBLUP for mixed model solving and genomic prediction. Based on R/rrBLUP version 4.6.3.

This is a library crate.

For GWAS workflows, see gwaspoly-rs (which uses this crate) or the Binx CLI (binx gwas) for a complete command-line interface.

Features

mixed_solve() - REML-based mixed model solver
a_mat() - Additive relationship matrix computation
kin_blup() - Genomic BLUP with kinship matrix
Faithful port of R/rrBLUP algorithms

Usage

use rrblup_rs::{mixed_solve_reml, MixedSolveOptions};
use nalgebra::DMatrix;

let y = vec![1.0, 2.0, 3.0, 4.0, 5.0];
let x = DMatrix::from_row_slice(5, 1, &[1.0, 1.0, 1.0, 1.0, 1.0]); // intercept

let result = mixed_solve_reml(&y, Some(&x), None, MixedSolveOptions::default())?;
println!("Vu: {}, Ve: {}", result.vu, result.ve);

Benchmarks

mixed.solve with marker matrix Z

Varying markers (n=500 genotypes):

Markers (m)	R (ms)	Rust (ms)	Speedup
1,000	1,014	532	1.9x
5,000	2,447	706	3.5x
10,000	4,222	1,000	4.2x
20,000	7,841	1,604	4.9x

Varying genotypes (m=10,000 markers):

Genotypes (n)	R (ms)	Rust (ms)	Speedup
100	159	31	5.1x
200	638	117	5.5x
500	4,237	936	4.5x
1,000	19,937	5,960	3.3x

kin.blup

n_geno	R (ms)	Rust (ms)	Speedup
100	93	54	1.7x
200	704	474	1.5x

Summary: Rust is 3-5x faster for marker-based workflows, with advantages growing as problem size increases.

Citation

If you use rrblup-rs, please cite the original R/rrBLUP paper:

Endelman, J.B. (2011). Ridge regression and other kernels for genomic selection with R package rrBLUP. The Plant Genome 4:250-255.

This crate is a Rust reimplementation of mixed.solve and related functionality. For a command-line interface, see Binx.

Related Crates

gwaspoly-rs - GWASpoly implementation (uses this crate)
binx-cli - User-facing CLI for GWAS workflows

License

GPL-3.0-or-later

rrblup-rs 0.1.0