Detection of heterozygosity-rich regions (HRRs) in the genome of domestic ruminants: which parameters for which breed

The detection of heterozygosity-rich regions (HRR) in livestock genomes is a recent concept: HRR promise to be a useful tool to estimate heterozygosity and to identify regions of the genome that are under balancing selection or are the result of introgres- sion and admixture events. Very little is known on how sensitive results are to the detection parameters. We used data on three domestic ruminant species-cattle (Holstein), sheep (Lacaune) and goat (Saanen)- to explore the effect on HRR detection of: (i) minimum number of SNP in the HRR (range 10-20); (ii) mini- mum length of the HRR (range 150-350 kb); (iii) maximum number of homozygous SNP in the HRR (range 0-5); (iv) max- imum number of missing SNP in the HRR (range 0-5). These parameters were tweaked around the base scenario with mini- mum 15 SNP, minimum 250 kb long, maximum 3 homozygous SNP and 2 missing SNP. Before the analysis, data were filtered for MAF >5%, SNP missing-rate <5%, individual missing-rate <10%; only SNP on autosomes (cow: 1-29; sheep: 1-26; goat: 1-29) were used. The filtered data consisted of 43,737 SNP for 104 sheep, 40,025 SNP for 63 cows and 44,624 for 163 goats. HRR were detected using the consecutive methods with the R package detectRUNS. The following evaluation metrics were used: the number of HRR detected (total and per sample), the average length of HHR and the average number of SNP inside HRR. The minimum number of SNP and the maximum number of homo- zygous SNP in a HRR showed the largest impact on the results. With minimum 10 SNP, over 600 HRR per animal were detected in all species, which dropped to below 100 and below 10 HRR per animal when the minimum number of SNP was increased to 15 and 20. The number of HRR showed an inverse relationship with their length, doubling from 588 kb (sheep) - 670 kb (cow) to 1057 kb (goat) - 1261 kb (sheep) with a 20-SNP threshold. When no homozygous SNP were allowed, only 1-2 HRR per animal were detected on average, which jumped to 295 (sheep) - 384 (goat) when as many as 5 homozygous SNP were allowed. This was in a direct relationship with the size of HRR, which increased from 830 kb to 996 kb in cows and from 751 kb to 836 kb in goats. Interestingly, in sheep this relationship was reversed, with the longest HRR found when no homozygous SNP were allowed (986 kb vs 882 kb). These results represent a fundamental basis to better understand HRR and to define guidelines for the detec- tion parameters in domestic ruminants.