Improved computation of genomic and pedigree inbreeding and relationships accounting for the X-chromosome

J.P. Nani1, G.B. Jansen2, and P.M. VanRaden1*

1USDA, Agricultural Research Service, Animal Genomics and Improvement Laboratory, Beltsville, MD 20705-2350, USA
2Council on Dairy Cattle Breeding, Bowie, MD 20716, USA

*Corresponding author


Background: Breeders for many decades used pedigrees to limit increases in inbreeding, but genomic measures of relationship and inbreeding can provide more precise control. Previous calculations of pedigree inbreeding (Fped), genomic inbreeding (Fgen), pedigree expected future inbreeding (EFIped), and genomic expected future inbreeding (EFIgen) ignored the influence of the X-chromosome when estimating relationships. However, excluding the X-chromosome can cause an increase in inbreeding by mating two individuals with the same X-chromosome. Because the X-specific region has 3.0% of the 79,060 markers used in U.S. genomic evaluation and those markers are coded as 100% homozygous in males, homozygosity of females appeared to be 3% less than for males. Allele frequency also has an impact on computing Fgen. Programs to compute pedigree and genomic measures were revised to improve speed and memory use as well as to better account for the X-chromosome.

Results: Revised software reduced computational time from 33 hours to 13 minutes for Fped and EFIped with 88 million animals in the pedigree and from 19 hours to 28 minutes for Fgen and EFIgen for 3,280,753 genotyped animals of 5 breeds. With the X-chromosome excluded, correlations were high between Fgen computed using either an allele frequency of 0.5 or a base population frequency for most but not all breeds; mean Fgen was higher for males than for females. After adjusting Fgen for the X-chromosome using an allele frequency of 0.5, correlations across breeds and sex increased, and X-adjusted Fgen was more similar to Fped. Mean correlation with Fped across breeds was 0.67 for Fgen with an allele frequency of 0.5, 0.67 for X-adjusted Fgen, and 0.54 for Fgen with base population allele frequency; corresponding EFI correlations were 0.83, 0.83, and 0.84. Correlation of haplotype-based inbreeding with Fped was 0.64.

Conclusions: Revision of inbreeding software allowed simpler and more accurate comparison of genomic and pedigree relationships and much faster computation. Use of the X-chromosome in estimating genomic inbreeding measures did not affect the correlation between sexes but did improve mean differences. Breeds with smaller populations were more sensitive to the use of different allele frequencies.