Genome-enabled prediction of reproductive traits in Nellore cattle using parametric models and machine learning methods

This study aimed to assess the predictive ability of different machine learning (ML) methods for genomic prediction of reproductive traits in Nellore cattle. The studied traits were age at first calving (AFC), scrotal circumference (SC), early pregnancy (EP) and stayability (STAY). The numbers of genotyped animals and SNP markers available were 2342 and 321 419 (AFC), 4671 and 309 486 (SC), 2681 and 319 619 (STAY) and 3356 and 319 108 (EP). Predictive ability of support vector regression (SVR), Bayesian regularized artificial neural network (BRANN) and random forest (RF) were compared with results obtained using parametric models (genomic best linear unbiased predictor, GBLUP, and Bayesian least absolute shrinkage and selection operator, BLASSO). A 5-fold cross-validation strategy was performed and the average prediction accuracy (ACC) and mean squared errors (MSE) were computed. The ACC was defined as the linear correlation between predicted and observed breeding values for categorical traits (EP and STAY) and as the correlation between predicted and observed adjusted phenotypes divided by the square root of the estimated heritability for continuous traits (AFC and SC). The average ACC varied from low to moderate depending on the trait and model under consideration, ranging between 0.56 and 0.63 (AFC), 0.27 and 0.36 (SC), 0.57 and 0.67 (EP), and 0.52 and 0.62 (STAY). SVR provided slightly better accuracies than the parametric models for all traits, increasing the prediction accuracy for AFC to around 6.3 and 4.8% compared with GBLUP and BLASSO respectively. Likewise, there was an increase of 8.3% for SC, 4.5% for EP and 4.8% for STAY, comparing SVR with both GBLUP and BLASSO. In contrast, the RF and BRANN did not present competitive predictive ability compared with the parametric models. The results indicate that SVR is a suitable method for genome-enabled prediction of reproductive traits in Nellore cattle. Further, the optimal kernel bandwidth parameter in the SVR model was trait-dependent, thus, a fine-tuning for this hyper-parameter in the training phase is crucial.

Contemporary group alternatives for genetic evaluation of milk yield in small populations of dairy cattle

In the present study, different random regression models, focussed on the nature of the effect of the contemporary group, fixed or random, were compared for genetic evaluation of test-day milk yield in dairy Gir breed, whose herds are characterised by relatively reduced numbers of cows. Cows were assigned to the same contemporary group if they were tested in the same herd, year and month. In one of the evaluated models, the contemporary group was investigated as a fixed effect and a clustering procedure was adopted to group herd-year subclasses in order to, at the end of the process, all contemporary groups contained at least three cows. The models were compared by the quality of fit, the accuracy of the predicted breeding values and their estimates of genetic parameters. The clustering procedure did not significantly improve the accuracy of predicted breeding values. Moreover, the fit to the data for this model was negatively affected. Therefore, this strategy should not be further implemented. The model including the herd-calving period effect and contemporary group treated as random effect showed similar characteristics to its equivalent, where the contemporary group was modelled as a fixed effect. However, the fit to the data for this model was slightly worse. Thus, the results suggest a random regression model including the herd-calving period effect and the fixed effect of contemporary group for the genetic evaluations of production traits in dairy Gir cattle. These findings could be extended to small dairy cattle populations whose herds are small-sized.

Pedigree analysis and inbreeding effects over morphological traits in Campolina horse population

Genetic improvement, without control of inbreeding, can go to loss of genetic variability, reducing the potential for genetic gains in the domestic populations. The aim of this study was to analyze the population structure and the inbreeding depression in Campolina horses. Phenotype information from 43 465 individuals was analyzed, data provided by the Campolina Breeders Association. A pedigree file containing 107 951 horses was used to connected the phenotyped individuals. The inbreeding coefficient was performed by use of the diagonal of the relationship matrix and the genealogical parameters were computed using proper softwares. The effective population size was estimated based on the rate of inbreeding and census information, and the stratification of the population was verified by the average relationship coefficient between animals born in different regions of Brazil. The effects of inbreeding on morphological traits were made by the use of inbreeding coefficient as a covariate in the model of random regression. The inbreeding coefficient increased from 1990 on, impacting effective population size and, consequently, shrinking genetic variability. The paternal inbreeding was greater than maternal, which may be attributed to the preference for inbred animals in reproduction. The average genetic relationship coefficient of animals born in different states was lower than individuals born within the same state. The increase in the inbreeding coefficient was negatively associated with all studied traits, showing the importance to avoid genetic losses in the long term. Although results do not indicate a severe narrowing of the population until the present date, the average relationship coefficient shows signs of increase, which could cause a drastic reduction in genetic variability if inbred mating is not successfully controlled in the Campolina horse population.

Genomic study and Medical Subject Headings enrichment analysis of early pregnancy rate and antral follicle numbers in Nelore heifers

Zebu animals () are known to take longer to reach puberty compared with taurine animals (), limiting the supply of animals for harvest or breeding and impacting profitability. Genomic information can be a helpful tool to better understand complex traits and improve genetic gains. In this study, we performed a genomewide association study (GWAS) to identify genetic variants associated with reproductive traits in Nelore beef cattle. Heifer pregnancy (HP) was recorded for 1,267 genotyped animals distributed in 12 contemporary groups (CG) with an average pregnancy rate of 0.35 (±0.01). Disregarding one of these CG, the number of antral follicles (NF) was also collected for 937 of these animals, with an average of 11.53 (±4.43). The animals were organized in CG: 12 and 11 for HP and NF, respectively. Genes in linkage disequilibrium (LD) with the associated variants can be considered in a functional enrichment analysis to identify biological mechanisms involved in fertility. Medical Subject Headings (MeSH) were detected using the MESHR package, allowing the extraction of broad meanings from the gene lists provided by the GWAS. The estimated heritability for HP was 0.28 ± 0.07 and for NF was 0.49 ± 0.09, with the genomic correlation being -0.21 ± 0.29. The average LD between adjacent markers was 0.23 ± 0.01, and GWAS identified genomic windows that accounted for >1% of total genetic variance on chromosomes 5, 14, and 18 for HP and on chromosomes 2, 8, 11, 14, 15, 16, and 22 for NF. The MeSH enrichment analyses revealed significant ( < 0.05) terms associated with HP-"Munc18 Proteins," "Fucose," and "Hemoglobins"-and with NF-"Cathepsin B," "Receptors, Neuropeptide," and "Palmitic Acid." This is the first study in Nelore cattle introducing the concept of MeSH analysis. The genomic analyses contributed to a better understanding of the genetic control of the reproductive traits HP and NF and provide new selection strategies to improve beef production.

Parameters of a dynamic mechanistic model of cattle growth retain enough biological interpretation for genotype-to-phenotype mapping

This study aimed to investigate the predictability of a phenotype when using a dynamic model of cattle growth. Genotypic and phenotypic information on Nellore (Bos indicus) cattle were used in a genome-wide association analysis designed to contrast the biological interpretation of core parameters [conversion efficiency of metabolizable energy to net energy for gain (k_g) and adjusted final shrunk body weight (AFSBW)] to their associated genomic regions and nearby quantitative trait loci (QTLs). Single nucleotide polymorphisms (SNPs) were used to develop prediction equations for k_g and AFSBW, which enter the model for simulative prediction purposes. QTLs and genes, one related to mature body weight and another to growth efficiency, are consistent with the model equations. Significantly associated SNPs were used to compute parameters, which yielded reasonable model outcomes when compared with regular parameter computations. Our results provide evidence of the biological validity of using such parameters as component traits of higher phenotypes and the possibility of using genomic data for genotype-to-parameter mapping.

Systems genetics and genome-wide association approaches for analysis of feed intake, feed efficiency, and performance in beef cattle

Feed intake, feed efficiency, and weight gain are important economic traits of beef cattle in feedlots. In the present study, we investigated the physiological processes underlying such traits from the point of view of systems genetics. Firstly, using data from 1334 Nellore (Bos indicus) cattle and 943,577 single nucleotide polymorphisms (SNPs), a genome-wide association analysis was performed for dry matter intake, average daily weight gain, feed conversion ratio, and residual feed intake with a Bayesian Lasso procedure. Genes within 50-kb SNPs, most relevant for explaining the genomic variance, were annotated and the biological processes underlying the traits were inferred from Database for Annotation, Visualization and Integrated Discovery (DAVID) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases. Our results indicated several putative genomic regions associated with the target phenotypes and showed that almost all genomic variances were in the SNPs located in the intergenic and intronic regions. We further identified five main metabolic pathways related to ion transport, body composition, and feed intake control, which influenced the four phenotypes simultaneously. The systems genetics approach used in this study revealed novel pathways related to feed efficiency traits in beef cattle.

A genomewide association mapping study using ultrasound-scanned information identifies potential genomic regions and candidate genes affecting carcass traits in Nellore cattle

The aim of this study was to identify candidate genes and genomic regions associated with ultrasound-derived measurements of the rib-eye area (REA), backfat thickness (BFT) and rumpfat thickness (RFT) in Nellore cattle. Data from 640 Nellore steers and young bulls with genotypes for 290 863 single nucleotide polymorphisms (SNPs) were used for genomewide association mapping. Significant SNP associations were explored to find possible candidate genes related to physiological processes. Several of the significant markers detected were mapped onto functional candidate genes including ARFGAP3, CLSTN2 and DPYD for REA; OSBPL3 and SUDS3 for BFT; and RARRES1 and VEPH1 for RFT. The physiological pathway related to lipid metabolism (CLSTN2, OSBPL3, RARRES1 and VEPH1) was identified. The significant markers within previously reported QTLs reinforce the importance of the genomic regions, and the other loci offer candidate genes that have not been related to carcass traits in previous investigations.

Impact of reference population on accuracy of imputation from 6K to 50K single nucleotide polymorphism chips in purebred and crossbreed beef cattle

Genotyping with lower density but lower cost panels enables more animals to be genotyped for genomic selection. Imputation enables the determination of missing SNP genotypes in animals genotyped with a low-density panel by using information from a reference population genotyped with a higher density panel, which should increase accuracy of genomic EBV. In this study, population imputation, using linkage disequilibrium among markers, was implemented using the software BEAGLE, FIMPUTE 2.2, and IMPUTE2 in a multibreed, crossbred taurine beef cattle population genotyped with the Illumina SNP50. Different combinations of reference populations and imputed animals were defined based on breed composition. Number of animals (n = 250 to 4,932) and the presence of closer relatives in the reference population (only for Angus animals) were investigated. The overall average imputation accuracy for purebred animals ranged from 94.20 to 97.93% using FIMPUTE, from 95.35 to 98.31% using IMPUTE2, and from 90.02 to 96.38% when BEAGLE software was used. Imputation accuracy of crossbred animals ranged from 54.15 to 97.53% (FIMPUTE), from 57.04 to 97.46% (IMPUTE2), and from 54.35 to 95.64% (BEAGLE). Higher imputation accuracies were obtained when closer relatives along with the breed composition of imputed animals was well represented in the reference population. Within breed imputation from 6K to 50K did not improve when an additional purebred population was added to the reference population. FIMPUTE reduced the run time by 13 to 52 times compared to BEAGLE and 51 to 108 times compared to IMPUTE2.