Genomic Feature Model Analysis Of Complex Traits In Dairy Cattle

Understanding the genetic architecture of complex traits(e.g.,the distribution of causal variants and their respective effects)and predicting their future phenotypes are very important in human medicine,animal and plant breeding,and evolution trajectory.The genetic basis underlying complex traits is commonly due to many loci with small effects,which are individually undetectable by single-marker regression analyses.To overcome this limitation,assessing the connective association of multiple markers in a genomic feature with a complex trait could be a promising way to uncover the genetic architecture and thereby to improve genomic prediction accuracy for complex traits.We call this approach as genomic feature model analysis of genomic markers.Genomic features could be defined from various sources of biological knowledge,such as biological pathways and functional genome study.In the general introduction,we summarized and discussed the methods for defining genomic features,the approaches for analysing genomic features,and the potential factors influencing the performance,as well as highlighted the ongoing challenges.A better understanding of the various choices made in genomic feature analyses will improve the interpretability of findings and ensure the comparability across studies.In the following studies,we focused on mastitis and milk production traits in dairy cattle.In study ?,genome-wide expression of mRNAs and miRNAs in bovine mammary gland tissues were investigated at 24 h after intra-mammary infection(IMI)with high or low concentrations of S.aureus,followed by the integrated analysis with QTL database.This study aimed to identify the genes,miRNAs and pathways associated with S.aureus-induced mastitis in dairy cattle.We identified 194 DEGs and 2 differentially expressed miRNAs after IMI with high concentration of S.aureus compared to the negative controls,while we only found 21 DEGs after IMI with low concentration of S.aureus,demonstrating that the responses of the mammary gland to S.aureus were dosage-dependent.Several innate immune-relevant pathways were impacted during IMI with high concentration of S.aureus,such as cytokine-cytokine receptor interaction and inflammatory response.The integrated analysis of expression data with prior QTLs relevant with mastitis suggested 28 genes(e.g.CXCL14,KIT and SLC4A11)as the most promising candidates associated with S.aureus-induced mastitis.In Study ?,we combined sequence-based genome-wide association study(GWAS)with IMI-relevant transcriptome data by using SNP-set test to investigate the genetic and biological basis of mastitis and milk production traits in three dairy cattle breeds,Holstein(HOL),Jersey(JER),and Nordic red cattle(RDC).In the liver,we found that at 3 h post-IMI DEGs were more associated with mastitis,whereas at later stages DEGs were more associated with milk production.The up-and down-regulated genes were enriched for associated variants with mastitis and milk production,respectively.Similar patterns were observed in the three breeds.Using genomic feature linear mixed model,we partitioned genomic variance based on the most significant genomic features obtained from SNP-set test.We found that much larger proportions of genomic variance were accounted for by these genomic features relative to their SNP proportions over the whole genome,confirming the results from SNP-set test.In Study ?,we applied two genomic feature modelling approaches(i.e.,SNP-set test and GFBLUP model),using DEGs detected in the liver post-IMI with Escherichia coli(E.coli)as genomic features,to mastitis and milk production traits in both HOL and JER.The accuracy of genomic prediction with GFBLUP was marginally improved(approximate 3%)relative to standard GBLUP,and the most apparent increase(132.2%)for prediction accuracy was observed in across-breed prediction(i.e.,considering HOL as training population and JER as validation population).The degree of enrichment(determined by P-value)from SNP-set test was positively correlated(P<0.05)with the change in prediction accuracy with GFBLUP,suggesting that the SNP-set test could be used as the first-step to preselect the significant genomic features,potentially leading to more accurate GFBLUP or similar models.Study ? was similar to Study ?,but differed at the sources for defining genomic features:GO categories were used in Study IV.The functional annotation of genes in GO allows us to gain novel insights into the biological mechanisms underlying mastitis and milk production.We found that the immune-relevant GO terms were more associated with mastitis than milk production,whereas metabolism-relevant GO terms were more associated with milk production than mastitis.Compared to GBLUP,the accuracy of genomic prediction with GFBLUP was improved both.within and across breeds by several biologically meaningful GO terms.However,only a small fraction(approximate 20%)of bovine genes are annotated in GO,which potentially limits the findings of this study.To realize the potential of genomic feature analyses to discover the genetic and biological basis as well as to improve prediction accuracy for complex traits,a better knowledge about functional regions of the genome are required.With the increasing availability of biological knowledge,the genomic feature analysis of genomic markers,especially for sequence-level genomic markers,will become increasingly useful.