Correlation scan: identifying genomic regions that affect genetic correlations applied to fertility traits

Abstract

Although the genetic correlations between complex traits have been estimated for more than a century, only recently we have started to map and understand the precise localization of the genomic region(s) that underpin these correlations. Reproductive traits are often genetically correlated. Yet, we don’t fully understand the complexities, synergism, or trade-offs between male and female fertility. In this study, we used reproductive traits in two cattle populations (Brahman; BB, Tropical Composite; TC) to develop a novel framework termed correlation scan (CS). This framework was used to identify local regions associated with the genetic correlations between male and female fertility traits. Animals were genotyped with bovine high-density single nucleotide polymorphisms (SNPs) chip assay. The data used consisted of ~1000 individual records measured through frequent ovarian scanning for age at first corpus luteum (AGECL) and a laboratory assay for serum levels of insulin growth hormone (IGF1 measured in bulls, IGF1b, or cows, IGF1c). The methodology developed herein used correlations of 500-SNP effects in a 100-SNPs sliding window in each chromosome to identify local genomic regions that either drive or antagonize the genetic correlations between traits. We used Fisher’s Z-statistics through a permutation method to confirm which regions of the genome harboured significant correlations. About 30% of the total genomic regions were identified as driving and antagonizing genetic correlations between male and female fertility traits in the two populations. These regions confirmed the polygenic nature of the traits being studied and pointed to genes of interest. For BB, the most important chromosome in terms of local regions is often located on bovine chromosome (BTA) 14. However, the important regions are spread across few different BTA’s in TC. Quantitative trait loci (QTLs) and functional enrichment analysis revealed many significant windows co-localized with known QTLs related to milk production and fertility traits, especially puberty. In general, the enriched reproductive QTLs driving the genetic correlations between male and female fertility are the same for both cattle populations, while the antagonizing regions were population specific. Moreover, most of the antagonizing regions were mapped to chromosome X. These results suggest regions of chromosome X for further investigation into the trade-offs between male and female fertility. We compared the CS with two other recently proposed methods that map local genomic correlations. Some genomic regions were significant across methods. Yet, many significant regions identified with the CS were overlooked by other methods.

Stacking haplotypes from the Vavilov wheat collection to accelerate breeding for multiple disease resistance

“…The genome-wide 1,756 haploblocks each harbouring >3 markers were used to calculate local correlations between all 15 pairs of disease resistance traits (Figure S14; Table S5). The results from a sliding-window-based approach, correlation scan (Olasege et al, 2022), was used to compare with that of the haploblock-based method, and similar estimates of local genetic correlations for three resistance trait pairs were observed on most chromosomes between the two methods (Figure S15).…”

“…Considering the overall genetic relationships between disease resistance traits, we hypothesised that there may be local genomic regions underpinning the global genetic correlations, which can either be in the same direction as the global correlation estimate (‘drivers’), or less commonly in the opposite direction (‘antagonizers’) (Olasege et al, 2022). To explore local genetic correlations, we adopted a novel haploblock-based approach that correlated genetic effects of markers in strong LD.…”

“…Similar genome-wide patterns were observed for genetic correlation estimates from the CS method (Figure S15), implying the true existence of genomic regions that are genetically correlated. A study in cattle also showed that the marker effects were not biased by LD, that is, for the LD pruned markers, no substantial differences were found for the genome-wide pattern of the r estimates (Olasege et al, 2022). Some genes might be involved in multiple traits that are not correlated, leading to identification of significantly correlated resistance haploblocks for uncorrelated diseases (Figure 1).…”

“…Linear regression was performed in R package 'ggplot2' between mean correlations of significant haploblocks and global genetic correlations (see above) using all 15 trait pairs. Finally, the 'driver' and 'antagonizing' haploblocks were defined as the haploblocks driving or antagonizing the global genetic correlations, respectively (Olasege et al, 2022). If the global genetic correlation between two traits is positive/negative, the 'driver' regions are the haploblocks with significant positive/negative correlations, while the 'antagonizing' haploblocks are with significant negative/positive correlations.…”

“…If the global genetic correlation between two traits is positive/negative, the 'driver' regions are the haploblocks with significant positive/negative correlations, while the 'antagonizing' haploblocks are with significant negative/positive correlations. We also used correlation scan, a recently developed method (Olasege et al, 2022) to identify local genomic correlations in livestock, to compare with the haploblock-based method.…”