Genome-wide runs of homozygosity revealed sources of inconsistencies in the AxiomTM Equine Genotyping arrayAnnik Imogen Gmel1,2 and Markus Neuditschko11Animal GenoPhenomics, Agroscope, Posieux, Switzerland, 2Equine Department, Vetsuisse Faculty, University of Zurich, SwitzerlandCorresponding author: [email protected] of homozygosity (ROH), haplotypes identical-by-descent (IBD), are key tools for deriving genomic inbreeding1. The genomic inbreeding coefficient (FROH) for an animal is derived by dividing the sum of all homozygous segments (SROH) by the total length of the genome. In livestock, previously reported studies have shown high concordance rates between FROH and pedigree derived inbreeding (FPED). In the framework of a global genetic diversity study of modern horse breeds, we recently observed notably low FROH values compared to the corresponding FPED. In this study, a total of 4’520 horses from 21 different breeds were genotyped on the 670K AxiomTMEquine Genotyping array (602,131 autosomal SNPs mapped to EquCab3.0)2, 3. Quality of genotyping was considered acceptable with a dish QC (DQC) ≥ 0.82 and QC call rate (CR) ≥ 97 according to Affymetrix Axiom best practices4. Runs of homozygosity segments (NROH) for each horse were determined with an overlapping window approach in PLINK v1.95 with the following parameters: a minimum SNP density of one SNP per 50 kb, a maximum gap length of 100 kb, a minimum length of homozygous segments of 500 kb (including more than 80 homozygous SNPs), and allowing for one heterozygous SNP per segment6. In total there were 18 horses with NROH equal to 0 and another 62 horses with NROH less than 30. Within the Franches-Montagnes (FM) breed it was particularly notable, that five purebred horses had fewer ROH segments (NROH < 88) compared to F1 outcrosses. Of particular concern were one horse with NROH equal to 0 and two others with NROHless than 60. We re-genotyped these horses, along with two additional control horses that had NROH between 123 and 180, using the same DNA sample and genotype platform. Based on this data, we computed genotype concordance rates between the two SNP batches. We found that FM horses with none or few NROH exhibited low genotype concordances rates (Figure 1a; green dots). The same result was observed for eight re-genotyped Lusitano (LUS) horses, with NROH ranging from 1 to 99 (Figure 1a; brown dots). A comparison of discordant SNPs between the three FM outlier horses suggested that individual genotype errors occurred randomly, as there was only an 8% overlap in erroneous markers among the horses (Figure 1b). Notably, we observed over 50,000 discordant SNPs for the FM horse showing none NROH. Based on the updated genotype information, all horses (FM and LUS) exhibited significantly higher NROH, ranging from 148 to 175. Our findings demonstrate that NROH is a reliable indicator for assessing the genotype quality of individual horses. Consequently, we recommend excluding horses exhibiting extremely low NROH (e.g., NROH < 30) from downstream analyses. However, further research is necessary to enhance the reproducibility of AxiomTM Equine Genotyping array.Figure 1: Analysis of runs of homozygosity segments (NROH), genotype concordance rate, and discordant SNPs in re-genotyped horses. (a) Correlation between NROHand genotype concordance between two independent genotyping efforts using the same DNA extract and SNP platform: This panel illustrates the association between NROH and genotype concordance in Franches-Montagnes (FM) and Lusitano horses. (b) Overlap of discordant SNPs: The Venn diagram shows the common and unique discordant SNPs among the three re-genotyped FM horses.

Runs of homozygosity (ROH) are continuous homozygous segments that arise through the transmission of haplotypes that are identical by descent (IBD). The length and distribution of ROH segments provide insight into the genetic diversity of populations and are useful to detect selection signatures. Here, we analysed pooled whole-genome sequencing data from 265 Western honey bee colonies from the two subspecies Apis mellifera mellifera and Apis mellifera carnica. Integrating individual ROH patterns and admixture levels in a high-resolution population network visualization allowed us to ascertain major differences between the two subspecies. Within A. m. mellifera, we identified well-defined substructures according to the genetic origin of the colonies and a fair amount of admixed colonies, despite the current applied conservation efforts. In contrast, A. m. carnica colonies were more inbred and could not be differentiated according to the geographical origin. We identified 29 coding genes in overlapping ROH segments within the two subspecies. Genes embedded in A. m. carnica specific homozygosity islands suggested a strong selection for production and behavioural traits, whilst the identified cuticula protein-coding genes (CPR3 and CPR4) were associated with their breed-specific stripe pattern. Local adaption of the two subspecies could be confirmed by the identification of two genes involved in the response to ultraviolet (UV) light. We demonstrated that colony genotypes derived from pooled honey bee workers are reliable to unravel the population dynamics in A. mellifera and provide fundamental information to conserve native honey bees.