Recent advances in animal breeding and genetics have made it possible to combine records, genotypes, and pedigree into a single input for predicting breeding values, making genomic evaluations easier and more efficient. However, large genomic models remain computationally expensive. As data grows, the question of whether old data is useful for predicting breeding values for young animals arises. We explored how data truncation can reduce computing time in genomic evaluations for the American Angus growth model without compromising prediction accuracy. We tested five phenotype truncation levels and two genomic data truncation levels. Our findings showed that moderate data truncation (removing 42% of the phenotypes and 22% of genotypes) reduced computing time by 66% while maintaining prediction accuracy. Our results suggest that truncating phenotypic and genotypic data creates a scenario where the same predictive power as using the full dataset is achieved in less time, especially for this multi-trait model with a strong data structure and traits with medium to high heritability.

Read the full article in the Journal of Animal Science.