Agricultural and Food Science (Sep 2020)

Snp_blup_rel: software for calculating individual animal SNP-BLUP model reliabilities

  • Hafedh Ben Zaabza,
  • Esa A. Mäntysaari,
  • Ismo Strandén

DOI
https://doi.org/10.23986/afsci.95617
Journal volume & issue
Vol. 29, no. 4

Abstract

Read online

The snp_blup_rel program computes model reliabilities for genomic breeding values. The program assumes a single trait SNP-BLUP model where the breeding value can include a residual polygenic (RPG) effect. The reliability calculation requires elements of the inverse of the mixed model equations (MME). The calculation has three steps: 1) MME calculation, 2) MME coefficient matrix inversion, and 3) reliability computation. When needed, the inverted matrix can be saved after step 2. Step 3 can be used separately to new genotypes which do not contribute information to Step 2. When an RPG effect is included, an approximate method based on Monte Carlo sampling is applied. This reduces the MME matrix size and allows including many genotyped individuals. The program is written in Fortran 90/95, and uses LAPACK subroutines which enable multi-threaded parallel computing. The program is efficient in terms of computing time and memory requirements, and can be used to analyze even large genomic data. Thus, the program can be used in calculating model reliabilities for large national genomic evaluations.