Improving the accuracy of genomic prediction in dairy cattle using the biologically annotated neural networks framework.

BACKGROUND: Biologically annotated neural networks (BANNs) are feedforward Bayesian neural network models that utilize partially connected architectures based on SNP-set annotations. As an interpretable neural network, BANNs model SNP and SNP-set effects in their input and hidden layers, respectively. Furthermore, the weights and connections of the network are regarded as random variables with prior distributions reflecting the manifestation of genetic effects at various genomic scales. However, its application in genomic prediction has yet to be explored. RESULTS: This study extended the BANNs framework to the area of genomic selection and explored the optimal SNP-set partitioning strategies by using dairy cattle datasets. The SNP-sets were partitioned based on two strategies-gene annotations and 100 kb windows, denoted as BANN_gene and BANN_100kb, respectively. The BANNs model was compared with GBLUP, random forest (RF), BayesB and BayesCπ through five replicates of five-fold cross-validation using genotypic and phenotypic data on milk production traits, type traits, and one health trait of 6,558, 6,210 and 5,962 Chinese Holsteins, respectively. Results showed that the BANNs framework achieves higher genomic prediction accuracy compared to GBLUP, RF and Bayesian methods. Specifically, the BANN_100kb demonstrated superior accuracy and the BANN_gene exhibited generally suboptimal accuracy compared to GBLUP, RF, BayesB and BayesCπ across all traits. The average accuracy improvements of BANN_100kb over GBLUP, RF, BayesB and BayesCπ were 4.86%, 3.95%, 3.84% and 1.92%, and the accuracy of BANN_gene was improved by 3.75%, 2.86%, 2.73% and 0.85% compared to GBLUP, RF, BayesB and BayesCπ, respectively across all seven traits. Meanwhile, both BANN_100kb and BANN_gene yielded lower overall mean square error values than GBLUP, RF and Bayesian methods. CONCLUSION: Our findings demonstrated that the BANNs framework performed better than traditional genomic prediction methods in our tested scenarios, and might serve as a promising alternative approach for genomic prediction in dairy cattle.

Cross-Species Comparative DNA Methylation Reveals Novel Insights into Complex Trait Genetics among Cattle, Sheep, and Goats.

The cross-species characterization of evolutionary changes in the functional genome can facilitate the translation of genetic findings across species and the interpretation of the evolutionary basis underlying complex phenotypes. Yet, this has not been fully explored between cattle, sheep, goats, and other mammals. Here, we systematically characterized the evolutionary dynamics of DNA methylation and gene expression in 3 somatic tissues (i.e. brain, liver, and skeletal muscle) and sperm across 7 mammalian species, including 3 ruminant livestock species (cattle, sheep, and goats), humans, pigs, mice, and dogs, by generating and integrating 160 DNA methylation and transcriptomic data sets. We demonstrate dynamic changes of DNA hypomethylated regions and hypermethylated regions in tissue-type manner across cattle, sheep, and goats. Specifically, based on the phylo-epigenetic model of DNA methylome, we identified a total of 25,074 hypomethylated region extension events specific to cattle, which participated in rewiring tissue-specific regulatory network. Furthermore, by integrating genome-wide association studies of 50 cattle traits, we provided novel insights into the genetic and evolutionary basis of complex phenotypes in cattle. Overall, our study provides a valuable resource for exploring the evolutionary dynamics of the functional genome and highlights the importance of cross-species characterization of multiomics data sets for the evolutionary interpretation of complex phenotypes in cattle livestock.

Preselecting Variants from Large-Scale Genome-Wide Association Study Meta-Analyses Increases the Genomic Prediction Accuracy of Growth and Carcass Traits in Large White Pigs.

Preselected variants associated with the trait of interest from genome-wide association studies (GWASs) are available to improve genomic prediction in pigs. The objectives of this study were to use preselected variants from a large GWAS meta-analysis to assess the impact of single-nucleotide polymorphism (SNP) preselection strategies on genome prediction of growth and carcass traits in pigs. We genotyped 1018 Large White pigs using medium (50k) SNP arrays and then imputed SNPs to sequence level by utilizing a reference panel of 1602 whole-genome sequencing samples. We tested the effects of different proportions of selected top SNPs across different SNP preselection strategies on genomic prediction. Finally, we compared the prediction accuracies by employing genomic best linear unbiased prediction (GBLUP), genomic feature BLUP and three weighted GBLUP models. SNP preselection strategies showed an average improvement in accuracy ranging from 0.3 to 2% in comparison to the SNP chip data. The accuracy of genomic prediction exhibited a pattern of initial increase followed by decrease, or continuous decrease across various SNP preselection strategies, as the proportion of selected top SNPs increased. The highest level of prediction accuracy was observed when utilizing 1 or 5% of top SNPs. Compared with the GBLUP model, the utilization of estimated marker effects from a GWAS meta-analysis as SNP weights in the BLUP|GA model improved the accuracy of genomic prediction in different SNP preselection strategies. The new SNP preselection strategies gained from this study bring opportunities for genomic prediction in limited-size populations in pigs.

Improving Genomic Prediction Accuracy in the Chinese Holstein Population by Combining with the Nordic Holstein Reference Population.

The size of the reference population is critical in order to improve the accuracy of genomic prediction. Indeed, improving genomic prediction accuracy by combining multinational reference populations has proven to be effective. In this study, we investigated the improvement of genomic prediction accuracy in seven complex traits (i.e., milk yield; fat yield; protein yield; somatic cell count; body conformation; feet and legs; and mammary system conformation) by combining the Chinese and Nordic Holstein reference populations. The estimated genetic correlations between the Chinese and Nordic Holstein populations are high with respect to protein yield, fat yield, and milk yield-whereby these correlations range from 0.621 to 0.720-and are moderate with respect to somatic cell count (0.449), but low for the three conformation traits (which range from 0.144 to 0.236). When utilizing the joint reference data and a two-trait GBLUP model, the genomic prediction accuracy in the Chinese Holsteins improves considerably with respect to the traits with moderate-to-high genetic correlations, whereas the improvement in Nordic Holsteins is small. When compared with the single population analysis, using the joint reference population for genomic prediction in younger animals, results in a 2.3 to 8.1 percent improvement in accuracy. Meanwhile, 10 replications of five-fold cross-validation were also implemented in order to evaluate the performance of joint genomic prediction, thereby resulting in a 1.6 to 5.2 percent increase in accuracy. With respect to joint genomic prediction, the bias was found to be quite low. However, for traits with low genetic correlations, the joint reference data do not improve the prediction accuracy substantially for either population.

Metabolite Profiling of Wheat Response to Cultivar Improvement and Nitrogen Fertilizer.

Both genetic improvement and the application of N fertilizer increase the quality and yields of wheat. However, the molecular kinetics that underlies the differences between them are not well understood. In this study, we performed a non-targeted metabolomic analysis on wheat cultivars from different release years to comprehensively investigate the metabolic differences between cultivar and N treatments. The results revealed that the plant height and tiller number steadily decreased with increased ears numbers, whereas the grain number and weight increased with genetic improvement. Following the addition of N fertilizer, the panicle numbers and grain weights increased in an old cultivar, whereas the panicle number and grain number per panicle increased in a modern cultivar. For the 1950s to 2010s cultivar, the yield increases due to genetic improvements ranged from -1.9% to 96.7%, whereas that of N application ranged from 19.1% to 81.6%. Based on the untargeted metabolomics approach, the findings demonstrated that genetic improvements induced 1.4 to 7.4 times more metabolic alterations than N fertilizer supply. After the addition of N, 69.6%, 29.4%, and 33.3% of the differential metabolites were upregulated in the 1950s, 1980s, and 2010s cultivars, respectively. The results of metabolic pathway analysis of the identified differential metabolites via genetic improvement indicated enrichment in 1-2 KEGG pathways, whereas the application of N fertilizer enriched 2-4 pathways. Our results provide new insights into the molecular mechanisms of wheat quality and grain yield developments.

Comparative epigenomics reveals the impact of ruminant-specific regulatory elements on complex traits.

BACKGROUND: Insights into the genetic basis of complex traits and disease in both human and livestock species have been achieved over the past decade through detection of genetic variants in genome-wide association studies (GWAS). A majority of such variants were found located in noncoding genomic regions, and though the involvement of numerous regulatory elements (REs) has been predicted across multiple tissues in domesticated animals, their evolutionary conservation and effects on complex traits have not been fully elucidated, particularly in ruminants. Here, we systematically analyzed 137 epigenomic and transcriptomic datasets of six mammals, including cattle, sheep, goats, pigs, mice, and humans, and then integrated them with large-scale GWAS of complex traits. RESULTS: Using 40 ChIP-seq datasets of H3K4me3 and H3K27ac, we detected 68,479, 58,562, 63,273, 97,244, 111,881, and 87,049 REs in the liver of cattle, sheep, goats, pigs, humans and mice, respectively. We then systematically characterized the dynamic functional landscapes of these REs by integrating multi-omics datasets, including gene expression, chromatin accessibility, and DNA methylation. We identified a core set (n = 6359) of ruminant-specific REs that are involved in liver development, metabolism, and immune processes. Genes with more complex cis-REs exhibited higher gene expression levels and stronger conservation across species. Furthermore, we integrated expression quantitative trait loci (eQTLs) and GWAS from 44 and 52 complex traits/diseases in cattle and humans, respectively. These results demonstrated that REs with different degrees of evolutionary conservation across species exhibited distinct enrichments for GWAS signals of complex traits. CONCLUSIONS: We systematically annotated genome-wide functional REs in liver across six mammals and demonstrated the evolution of REs and their associations with transcriptional output and conservation. Detecting lineage-specific REs allows us to decipher the evolutionary and genetic basis of complex phenotypes in livestock and humans, which may benefit the discovery of potential biomedical models for functional variants and genes of specific human diseases.

Using machine learning to improve the accuracy of genomic prediction of reproduction traits in pigs.

BACKGROUND: Recently, machine learning (ML) has become attractive in genomic prediction, but its superiority in genomic prediction over conventional (ss) GBLUP methods and the choice of optimal ML methods need to be investigated. RESULTS: In this study, 2566 Chinese Yorkshire pigs with reproduction trait records were genotyped with the GenoBaits Porcine SNP 50 K and PorcineSNP50 panels. Four ML methods, including support vector regression (SVR), kernel ridge regression (KRR), random forest (RF) and Adaboost.R2 were implemented. Through 20 replicates of fivefold cross-validation (CV) and one prediction for younger individuals, the utility of ML methods in genomic prediction was explored. In CV, compared with genomic BLUP (GBLUP), single-step GBLUP (ssGBLUP) and the Bayesian method BayesHE, ML methods significantly outperformed these conventional methods. ML methods improved the genomic prediction accuracy of GBLUP, ssGBLUP, and BayesHE by 19.3%, 15.0% and 20.8%, respectively. In addition, ML methods yielded smaller mean squared error (MSE) and mean absolute error (MAE) in all scenarios. ssGBLUP yielded an improvement of 3.8% on average in accuracy compared to that of GBLUP, and the accuracy of BayesHE was close to that of GBLUP. In genomic prediction of younger individuals, RF and Adaboost.R2_KRR performed better than GBLUP and BayesHE, while ssGBLUP performed comparably with RF, and ssGBLUP yielded slightly higher accuracy and lower MSE than Adaboost.R2_KRR in the prediction of total number of piglets born, while for number of piglets born alive, Adaboost.R2_KRR performed significantly better than ssGBLUP. Among ML methods, Adaboost.R2_KRR consistently performed well in our study. Our findings also demonstrated that optimal hyperparameters are useful for ML methods. After tuning hyperparameters in CV and in predicting genomic outcomes of younger individuals, the average improvement was 14.3% and 21.8% over those using default hyperparameters, respectively. CONCLUSION: Our findings demonstrated that ML methods had better overall prediction performance than conventional genomic selection methods, and could be new options for genomic prediction. Among ML methods, Adaboost.R2_KRR consistently performed well in our study, and tuning hyperparameters is necessary for ML methods. The optimal hyperparameters depend on the character of traits, datasets etc.

Incorporation of Trait-Specific Genetic Information into Genomic Prediction Models.

Due to the rapid development of high-throughput sequencing technology, we can easily obtain not only the genetic variants at the whole-genome sequence level (e.g., from 1000 Genomes project and 1000 Bull Genomes project), but also a wide range of functional annotations (e.g., enhancers and promoters from ENCODE, FAANG, and FarmGTEx projects) across a wide range of tissues, cell types, developmental stages, and environmental conditions. This huge amount of information leads to a revolution in studying genetics and genomics of complex traits in humans, livestock, and plant species. In this chapter, we focused on and reviewed the genomic prediction methods that incorporate external biological information into genomic prediction, such as sequence ontology, linkage disequilibrium (LD) of SNPs, quantitative trait loci (QTL), and multi-layer omics data (e.g., transcriptome, epigenome, and microbiome).

Genomic Prediction Using Bayesian Regression Models With Global-Local Prior.

Bayesian regression models are widely used in genomic prediction for various species. By introducing the global parameter τ, which can shrink marker effects to zero, and the local parameter λ k , which can allow markers with large effects to escape from the shrinkage, we developed two novel Bayesian models, named BayesHP and BayesHE. The BayesHP model uses Horseshoe+ prior, whereas the BayesHE model assumes local parameter λ k , after a half-t distribution with an unknown degree of freedom. The performances of BayesHP and BayesHE models were compared with three classical prediction models, including GBLUP, BayesA, and BayesB, and BayesU, which also applied global-local prior (Horseshoe prior). To assess model performances for traits with various genetic architectures, simulated data and real data in cattle (milk production, health, and type traits) and mice (type and growth traits) were analyzed. The results of simulation data analysis indicated that models based on global-local priors, including BayesU, BayesHP, and BayesHE, performed better in traits with higher heritability and fewer quantitative trait locus. The results of real data analysis showed that BayesHE was optimal or suboptimal for all traits, whereas BayesHP was not superior to other classical models. For BayesHE, its flexibility to estimate hyperparameter automatically allows the model to be more adaptable to a wider range of traits. The BayesHP model, however, tended to be suitable for traits having major/large quantitative trait locus, given its nature of the "U" type-like shrinkage pattern. Our results suggested that auto-estimate the degree of freedom (e.g., BayesHE) would be a better choice other than increasing the local parameter layers (e.g., BayesHP). In this study, we introduced the global-local prior with unknown hyperparameter to Bayesian regression models for genomic prediction, which can trigger further investigations on model development.

Weighted Single-Step Genome-Wide Association Study of Semen Traits in Holstein Bulls of China.

Efficient production of high-quality semen is a crucial trait in the dairy cattle breeding due to the widespread use of artificial insemination. However, the genetic architecture (e.g., distributions of causal variants and their corresponding effects) underlying such semen quality traits remains unclear. In this study, we performed genome-wide association studies to identify genes associated with five semen quality traits in Chinese Holstein population, including ejaculate volume, progressive sperm motility, sperm concentration, number of sperm, and number of progressive motile sperm. Our dataset consisted of 2,218 Holstein bulls in China with full pedigree information, representing 12 artificial insemination centers, with 1,508 genotyped using the Illumina BovineSNP50 BeadChip. We used a weighted single-step genome-wide association method with 10 adjacent Single nucleotide polymorphisms (SNPs) as sliding windows, which can make use of individuals without genotypes. We considered the top 10 genomic regions in terms of their explained genomic variants as candidate window regions for each trait. In total, we detected 36 window regions related to one or multiple semen traits across 19 chromosomes. Promising candidate genes of PSMB5, PRMT5, ACTB, PDE3A, NPC1, FSCN1, NR5A2, IQCG, LHX8, and DMRT1 were identified in these window regions for these five semen traits. Our findings provided a solid basis for further research into genetic mechanisms underlying semen quality traits, which may contribute to their accurate genomic prediction in Chinese Holstein population.

Comparative Transcriptomic and Proteomic Analyses Identify Key Genes Associated With Milk Fat Traits in Chinese Holstein Cows.

Milk fat is the most important energy substance in milk and contributes to its quality and health benefits. However, the genetic mechanisms underlying milk fat synthesis are not fully understood. The development of RNA sequencing and tandem mass tag technologies has facilitated the identification of eukaryotic genes associated with complex traits. In this study, we used these methods to obtain liver transcriptomic and proteomic profiles of Chinese Holstein cows (n = 6). Comparative analyses of cows with extremely high vs. low milk fat percentage phenotypes yielded 321 differentially expressed genes (DEGs) and 76 differentially expressed proteins (DEPs). Functional annotation of these DEGs and DEPs revealed 26 genes that were predicted to influence lipid metabolism through insulin, phosphatidylinositol 3-kinase/Akt, mitogen-activated protein kinase, 5' AMP-activated protein kinase, mammalian target of rapamycin, and peroxisome proliferator-activated receptor signaling pathways; these genes are considered as the most promising candidate regulators of milk fat synthesis. The findings of this study enhance the understanding of the genetic basis and molecular mechanisms of milk fat synthesis, which could lead to the development of cow breeds that produce milk with higher nutritional value.

Genome-Wide Association Study for Milk Protein Composition Traits in a Chinese Holstein Population Using a Single-Step Approach.

Genome-wide association studies (GWASs) have been widely used to determine the genetic architecture of quantitative traits in dairy cattle. In this study, with the aim of identifying candidate genes that affect milk protein composition traits, we conducted a GWAS for nine such traits (αs1-casein, αs2-casein, ß-casein, κ-casein, α-lactalbumin, ß-lactoglobulin, casein index, protein percentage, and protein yield) in 614 Chinese Holstein cows using a single-step strategy. We used the Illumina BovineSNP50 Bead chip and imputed genotypes from high-density single-nucleotide polymorphisms (SNPs) ranging from 50 to 777 K, and subsequent to genotype imputation and quality control, we screened a total of 586,304 informative high-quality SNPs. Phenotypic observations for six major milk proteins (αs1-casein, αs2-casein, ß-casein, κ-casein, α-lactalbumin, and ß-lactoglobulin) were evaluated as weight proportions of the total protein fraction (wt/wt%) using a commercial enzyme-linked immunosorbent assay kit. Informative windows comprising five adjacent SNPs explaining no < 0.5% of the genomic variance per window were selected for gene annotation and gene network and pathway analyses. Gene network analysis performed using the STRING Genomics 10.0 database revealed a co-expression network comprising 46 interactions among 62 of the most plausible candidate genes. A total of 178 genomic windows and 194 SNPs on 24 bovine autosomes were significantly associated with milk protein composition or protein percentage. Regions affecting milk protein composition traits were mainly observed on chromosomes BTA 1, 6, 11, 13, 14, and 18. Of these, several windows were close to or within the CSN1S1, CSN1S2, CSN2, CSN3, LAP3, DGAT1, RPL8, and HSF1 genes, which have well-known effects on milk protein composition traits of dairy cattle. Taken together with previously reported quantitative trait loci and the biological functions of the identified genes, we propose 19 novel candidate genes affecting milk protein composition traits: ARL6, SST, EHHADH, PCDHB4, PCDHB6, PCDHB7, PCDHB16, SLC36A2, GALNT14, FPGS, LARP4B, IDI1, COG4, FUK, WDR62, CLIP3, SLC25A21, IL5RA, and ACADSB. Our findings provide important insights into milk protein synthesis and indicate potential targets for improving milk quality.

Cell-bound lipases from Burkholderia sp. ZYB002: gene sequence analysis, expression, enzymatic characterization, and 3D structural model.

BACKGROUND: The whole-cell lipase from Burkholderia cepacia has been used as a biocatalyst in organic synthesis. However, there is no report in the literature on the component or the gene sequence of the cell-bound lipase from this species. Qualitative analysis of the cell-bound lipase would help to illuminate the regulation mechanism of gene expression and further improve the yield of the cell-bound lipase by gene engineering. RESULTS: Three predictive cell-bound lipases, lipA, lipC21 and lipC24, from Burkholderia sp. ZYB002 were cloned and expressed in E. coli. Both LipA and LipC24 displayed the lipase activity. LipC24 was a novel mesophilic enzyme and displayed preference for medium-chain-length acyl groups (C10-C14). The 3D structural model of LipC24 revealed the open Y-type active site. LipA displayed 96 % amino acid sequence identity with the known extracellular lipase. lipA-inactivation and lipC24-inactivation decreased the total cell-bound lipase activity of Burkholderia sp. ZYB002 by 42 % and 14 %, respectively. CONCLUSIONS: The cell-bound lipase activity from Burkholderia sp. ZYB002 originated from a multi-enzyme mixture with LipA as the main component. LipC24 was a novel lipase and displayed different enzymatic characteristics and structural model with LipA. Besides LipA and LipC24, other type of the cell-bound lipases (or esterases) should exist.