Strategies to find predictive variants that improve multi-breed genomic prediction in dairy cattle

Abstract: The  lack  of  large  reference  populations  has  limited  the  accuracy  of  genomic  prediction  for  dairy  cattle  breeds with small populations. We review strategies to improve multi-breed genomic prediction. Several studies  show  that  multi-breed  reference  populations  can  increase  accuracy  for  minor  breeds.  Evidence  suggests sequence and functional genomics can get us closer to causal variants, and thus improve multi-breed prediction. However, we know little about the sharing of causal variants across breeds. We carried out a case study comparing direction of effect of variants selected from within-breed genome-wide association studies (GWAS), a meta-GWAS, or a combination of functional, evolutionary, and pleiotropic information. Variants  selected  from  a  meta-GWAS  were  most  likely  to  have  the  same  direction  of  effect  in  different  populations.  Selecting  variants  from  a  meta-GWAS  with  the  same  direction  of  effect  in  multiple  breeds  may  be  a  suitable  strategy  to  select  sequence  variants  that  increase  the  accuracy  of  multi-breed  genomic  prediction.