Identifying hard and soft sweeps has been basically a solved problem for almost 15 years. This is because loci drifting (a) in the population have very different frequency histories than loci being swept to fixation (b). 
However, selection on complex, quantitative traits does not leave signatures of sweeps. When selecting quantitative traits, the trait changes rapidly, but this is due to hundreds or thousands of loci having small changes in allele frequency (c).
This has made it difficult to identify polygenic selection loci. However, in our recent preprint, we report results which basically solve this problem.
1) We have large datasets. 2) We use appropriate models.
1) We have large datasets. 2) We use appropriate models.
The key is to fit generation number (or a proxy such as birth date) as the dependent variable in a linear mixed model. A SNP which is significantly associated with birth date is changing in freq over time due to directional selection.
However, controlling for dependency between samples with a genomic relationship matrix is essential to control for population structure, relatedness, and inbreeding. Basically, demography is the bane of selection mapping, but the GRM helps control for demography.
So what do we find?
We identified 207 directional selection loci in 3 populations of cattle.
Analysis is well powered, with -log10(q-values) larger than 100 and loci identified across the genome.
Some of these selected loci are changing relatively slowly (1x10-4 per generation)
We identified 207 directional selection loci in 3 populations of cattle.
Analysis is well powered, with -log10(q-values) larger than 100 and loci identified across the genome.
Some of these selected loci are changing relatively slowly (1x10-4 per generation)
These selected loci make sense in the context of the different breeds. In Red Angus, ovarian steroidogenesis & reproductive tissues are under selection. Red Angus has been selected as a highly fertile, maternal breed. In Gelbvieh, muscle development & muscle tissues are enriched.
Further, the immune system is under selection. Cattle that are more resilient to disease are more productive.
However, not only can we identify loci responding to directional, polygenic selection, we can also identify local adaptation!
Instead of birth date, we fit environmental variables as the "trait" in a mixed model GWAS. Power is improved by using multivariate models.
Instead of birth date, we fit environmental variables as the "trait" in a mixed model GWAS. Power is improved by using multivariate models.
Again, we find environmental adaptation loci across the genome.
Here, expression of genes involved in neural signaling pathways and brain tissues are enriched.
Although different genes were identified in different populations, they identified the same pathways.
Here, expression of genes involved in neural signaling pathways and brain tissues are enriched.
Although different genes were identified in different populations, they identified the same pathways.
We hypothesize that these neural signaling pathways are important for local adaptation across mammals, which makes sense in the context of environmental physiology (sensing temperature, day length, etc.).
This also helps interpret selection scans from other datasets/species.
KHDRBS2 previously reported as under selection for domestication (nature.com/articles/ng.37…). Our results clarify that this locus is likely under balancing selection for local adaptation.
KHDRBS2 previously reported as under selection for domestication (nature.com/articles/ng.37…). Our results clarify that this locus is likely under balancing selection for local adaptation.
(Yellow is Arid Prairie, Magenta is Mountains)
For my cattle breeding tweeps, vasoconstriction and vasodilation are important, which makes sense in terms of high altitude disease (brisket disease) and fescue toxicosis.
Finally, we performed our directional selection mapping method (GPSM) in different ecoregions & did a meta-analysis across these ecoregion-specific selected loci. Meta-analysis looked for loci under strong selection in one or more regions and not under selection in other regions.
One of the striking things we see in this analysis is that previously different ecoregion allele freqs are converging to the population mean. This means that, likely due to the use of artificial insemination, cattle populations are losing local adaptation.
Tools to help farmers and ranchers selected for locally adapted cattle is important for #sustainability, especially in the context of climate variability. More on that from my lab soon!
Link to #bioRxiv #preprint:
biorxiv.org/content/10.110…
Congrats to @TroyNRowan for great work.
Thanks to all of our collaborators and to @USDA_NIFA for funding.
Thanks to @RedAngusAmerica @ASASimmental & @GelbviehNews for sharing large datasets!
biorxiv.org/content/10.110…
Congrats to @TroyNRowan for great work.
Thanks to all of our collaborators and to @USDA_NIFA for funding.
Thanks to @RedAngusAmerica @ASASimmental & @GelbviehNews for sharing large datasets!
I.e. preferentially introgressed between cattle and yak because of environmental advantage, not domestication advantage.
Thanks to @ChrisGaynor11 and @HickeyJohn for helping @TroyNRowan with simulations to prove this during his time at the Roslin Institute as a visiting scientist.
Check out their #rstats package AlphaSimR!
Check out their #rstats package AlphaSimR!
