Assessing runs of Homozygosity: a comparison of SNP Array and whole genome sequence low coverage data

Abstract

Background: Runs of Homozygosity (ROH) are genomic regions where identical haplotypes are inherited from each parent. Since their first detection due to technological advances in the late 1990s, ROHs have been shedding light on human population history and deciphering the genetic basis of monogenic and complex traits and diseases. ROH studies have predominantly exploited SNP array data, but are gradually moving to whole genome sequence (WGS) data as it becomes available. WGS data, covering more genetic variability, can add value to ROH studies, but require additional considerations during analysis.

Results: Using SNP array and low coverage WGS data from 1885 individuals from 20 world populations, our aims were to compare ROH from the two datasets and to establish software conditions to get comparable results, thus providing guidelines for combining disparate datasets in joint ROH analyses. By allowing heterozygous SNPs per window, using the PLINK homozygosity function and non-parametric analysis, we were able to obtain non-significant differences in number ROH, mean ROH size and total sum of ROH between data sets using the different technologies for almost all populations.

Conclusions: By allowing 3 heterozygous SNPs per ROH when dealing with WGS low coverage data, it is possible to establish meaningful comparisons between data using SNP array and WGS low coverage technologies.

Keywords: ROH; Runs of Homozygosity; SNP array data; WGS low coverage data.

Fig. 1

Effect of allowing heterozygous SNPs per window evaluated by ep(P,h) as a measure of the empirically observed actually number of heterozygous SNPs found in population P when we allow h heterozygous SNP. (See Materials and methods for the definition)

Fig. 2

Violin plots of the mean number of ROH longer than 1 Mb. Populations are colored by 5 biogeographical groups by admixture analysis. Admixed (Hispanic-American: CLM, MXL; African-American: ACB, ASW) – blue, Native Americans (PEL) – green, East (CHS, CDX, JPT) and South (KHV) Asia – tan, North (FIN, GBR, CEU) and South (IBS, TSI) Europe – violet, South (ZUL), East (BAG, LWK) and West (YRI) Africa – red. Four distributions per population are shown, array data with 1 heterozygous SNP allowed per window and WGS with 1 to 3 heterozygous SNPs allowed

Fig. 3

Violin plots of mean ROH size longer than 1 Mb (in Mb). Different biogeographical groups have different x-axis scales in an attempt to maximize the difference between distributions within populations. See Fig. 2 legend for population codes

Fig. 4

Violin plots of mean total sum of ROH longer than 1 Mb (in Gb). See fig. 2 legend for population codes

Fig. 5

Heatmaps of correlations and MWW tests of mean number of ROH, mean ROH size and mean total sum of ROH between array data allowing 1 heterozygous SNP per window and WGS data allowing 1 to 5 heterozygous SNPs per window (y-axis). a to c Pearson correlations. d to f P-values of Mann-Whitney-Wilcoxon non-parametrical test (MWW), red shows significant difference between array and WGS while blue shows distributions that cannot be considered different. See Fig. 2 legend for population codes

Fig. 6

Mean sum of ROH in different length categories. The light colored lines represent WGS with 3 heterozygous SNP allowed per window and dark colored lines represent array data with 1 heterozygous SNP allowed per ROH. See Fig. 2 legend for population codes