Comparative assessment of methods for estimating individual genome-wide homozygosity-by-descent from human genomic data

Abstract

BackgroundGenome-wide homozygosity estimation from genomic data is becoming an increasingly interesting research topic. The aim of this study was to compare different methods for estimating individual homozygosity-by-descent based on the information from human genome-wide scans rather than genealogies. We considered the four most commonly used methods and investigated their applicability to single-nucleotide polymorphism (SNP) data in both a simulation study and by using the human genotyped data. A total of 986 inhabitants from the isolated Island of Vis, Croatia (where inbreeding is present, but no pedigree-based inbreeding was observed at the level of F > 0.0625) were included in this study. All individuals were genotyped with the Illumina HumanHap300 array with 317,503 SNP markers.ResultsSimulation data suggested that multi-point FEstim is the method most strongly correlated to true homozygosity-by-descent. Correlation coefficients between the homozygosity-by-descent estimates were high but only for inbred individuals, with nearly absolute correlation between single-point measures.ConclusionsDeciding who is really inbred is a methodological challenge where multi-point approaches can be very helpful once the set of SNP markers is filtered to remove linkage disequilibrium. The use of several different methodological approaches and hence different homozygosity measures can help to distinguish between homozygosity-by-state and homozygosity-by-descent in studies investigating the effects of genomic autozygosity on human health.

Consanguinity around the world: what do the genomic data of the HGDP-CEPH diversity panel tell us?

“…Previous studies 9,24,25 have highlighted the risk of overestimating inbreeding coefficients when the studied markers are in LD. The strategy developed here, which consists of generating several sparse maps, seems to be a reliable strategy that avoids losing as much information as when a single sparse map is considered.…”

“…This approach provides a genomebased alternative to genealogy [2][3][4] and has been used so far in homozygosity mapping studies [5][6][7] or to study levels of inbreeding in isolated human populations. 8,9 Here, we propose to use the approach in population surveys to infer mating-type habits.…”

“…9 We propose to estimate by a maximum-likelihood method the proportion of some specific mating types (first cousin (1C), double first cousin (2Â1C), avuncular (AV), 2C etc) on the basis of the distribution of HBD segments over the genome of individuals from the studied populations. Indeed, the number and length of HBD segments in an individual genome depend on the relatedness of his/her parents and can thus be used to assess parental mating-type preferences.…”

“…FEstim requires that the markers in the map be in minimal linkage disequilibrium (LD), as, otherwise, inflation in inbreeding estimates has been demonstrated. 9 To avoid this bias, we developed an original strategy consisting of generating multiple sparse genome maps. This strategy has the advantage of not requiring any LD computation on the sample and of minimizing loss of information, as compared with a strategy that consists of using a single map of markers in minimal LD (as can be carried out with PLINK 4 or MASEL, 10 for instance).…”

Measuring individual inbreeding in the age of genomics: marker-based measures are better than pedigrees

“…We conducted 'LD pruning' because F E assumes all markers are in linkage equilibrium (Polašek et al, 2010). Additionally, F ROH is typically estimated after LD pruning to avoid detecting fixed or nearly fixed ROH arising due to ancient population processes (for example, selective sweeps) (for example, McQuillan et al, 2012).…”