PyMSQ: a Python package for fast Mendelian sampling (co)variance and haplotype-based similarity in genomic selection

Area of Science:

• Quantitative genetics
• Animal breeding
• Bioinformatics

Background:

• Genomic selection (GS) accelerates genetic gain but risks reduced haplotype diversity and increased inbreeding.
• Mendelian sampling variance (MSV) and covariance (MSC) can mitigate genetic variation loss by utilizing within-family segregation.
• Haplotype-based similarity measures offer direct control over haplotype diversity.

Purpose of the Study:

• Introduce PyMSQ, an open-source Python package for estimating MSV and MSC.
• Implement a novel haplotype-based similarity matrix for parental Mendelian sampling terms.
• Enhance genomic selection strategies by balancing genetic gain with diversity preservation.

Main Methods:

• Developed a matrix-based approach for computing MSV and MSC across various contexts (single-trait, multi-trait, zygotic).
• Created a haplotype-based similarity matrix quantifying shared heterozygous segments.
• Integrated optimized scientific libraries for computational efficiency.

Main Results:

• PyMSQ computes MSV and MSC significantly faster (up to 332-fold) than existing tools like gamevar, maintaining numerical accuracy.
• Demonstrated the utility of MSV, MSC, and the novel similarity measure on a Holstein-Friesian dataset.
• The new similarity measure complements standard genomic relationship matrices by focusing on heterozygous segments.

Conclusions:

• PyMSQ facilitates the practical application of MSV, MSC, and haplotype similarity metrics in breeding programs.
• Enables breeders to implement haplotype diversity constraints alongside optimal contribution selection.
• Supports the development of more sustainable genomic selection strategies by preserving key haplotypic segments.