miraculix

miraculix implements optimized C/CUDA routines for mathematical operations on compressed genotype data, accelerating matrix-vector multiplications of genotype covariates for large-scale genomic analyses.

Key Features:

• Efficient Computation: Provides optimized routines for genotype matrix-vector multiplications on compressed genotype data used in statistical analyses and repeated operations in single-step models.
• Integration and Interoperability: Offers programmatic interfaces for integration into C utilities and interoperability with R, Julia, and Fortran codebases.
• Optimized for Modern Hardware: Leverages Nvidia GPUs via CUDA and optimizes for a range of CPU architectures to reduce computation time on large population datasets.

Scientific Applications:

• Genome-Wide Association Studies (GWAS): Accelerates computations required in GWAS to enable analysis of larger genotype datasets.
• Genomic Breeding Value Estimation: Enables rapid genotype matrix multiplications needed for estimating breeding values in large populations.
• Population Summary Statistics: Supports computation of summary statistics across extensive genomic datasets for population-level analyses.

Methodology:

Implemented as a C/CUDA library that performs mathematical operations on compressed genotype data with optimized routines for genotype matrix-vector multiplications used in statistical and single-step models.