Predicting expected progeny difference for marbling score in Angus cattle using artificial neural networks and Bayesian regression models

BACKGROUND

Artificial neural networks (ANN) mimic the function of the human brain and are capable of performing massively parallel computations for data processing and knowledge representation. ANN can capture nonlinear relationships between predictors and responses and can adaptively learn complex functional forms, in particular, for situations where conventional regression models are ineffective. In a previous study, ANN with Bayesian regularization outperformed a benchmark linear model when predicting milk yield in dairy cattle or grain yield of wheat. Although breeding values rely on the assumption of additive inheritance, the predictive capabilities of ANN are of interest from the perspective of their potential to increase the accuracy of prediction of molecular breeding values used for genomic selection. This motivated the present study, in which the aim was to investigate the accuracy of ANN when predicting the expected progeny difference (EPD) of marbling score in Angus cattle. Various ANN architectures were explored, which involved two training algorithms, two types of activation functions, and from 1 to 4 neurons in hidden layers. For comparison, BayesCπ models were used to select a subset of optimal markers (referred to as feature selection), under the assumption of additive inheritance, and then the marker effects were estimated using BayesCπ with π set equal to zero. This procedure is referred to as BayesCpC and was implemented on a high-throughput computing cluster.

RESULTS

The ANN with Bayesian regularization method performed equally well for prediction of EPD as BayesCpC, based on prediction accuracy and sum of squared errors. With the 3K-SNP panel, for example, prediction accuracy was 0.776 using BayesCpC, and ranged from 0.776 to 0.807 using BRANN. With the selected 700-SNP panel, prediction accuracy was 0.863 for BayesCpC and ranged from 0.842 to 0.858 for BRANN. However, prediction accuracy for the ANN with scaled conjugate gradient back-propagation was lower, ranging from 0.653 to 0.689 with the 3K-SNP panel, and from 0.743 to 0.793 with the selected 700-SNP panel.

CONCLUSIONS

ANN with Bayesian regularization performed as well as linear Bayesian regression models in predicting additive genetic values, supporting the idea that ANN are useful as universal approximators of functions of interest in breeding contexts.

A novel analytical method, Birth Date Selection Mapping, detects response of the Angus (Bos taurus) genome to selection on complex traits

BACKGROUND

Several methods have recently been developed to identify regions of the genome that have been exposed to strong selection. However, recent theoretical and empirical work suggests that polygenic models are required to identify the genomic regions that are more moderately responding to ongoing selection on complex traits. We examine the effects of multi-trait selection on the genome of a population of US registered Angus beef cattle born over a 50-year period representing approximately 10 generations of selection. We present results from the application of a quantitative genetic model, called Birth Date Selection Mapping, to identify signatures of recent ongoing selection.

RESULTS

We show that US Angus cattle have been systematically selected to alter their mean additive genetic merit for most of the 16 production traits routinely recorded by breeders. Using Birth Date Selection Mapping, we estimate the time-dependency of allele frequency for 44,817 SNP loci using genomic best linear unbiased prediction, generalized least squares, and BayesCπ analyses. Finally, we reconstruct the primary phenotypes that have historically been exposed to selection from a genome-wide analysis of the 16 production traits and gene ontology enrichment analysis.

CONCLUSIONS

We demonstrate that Birth Date Selection Mapping utilizing mixed models corrects for time-dependent pedigree sampling effects that lead to spurious SNP associations and reveals genomic signatures of ongoing selection on complex traits. Because multiple traits have historically been selected in concert and most quantitative trait loci have small effects, selection has incrementally altered allele frequencies throughout the genome. Two quantitative trait loci of large effect were not the most strongly selected of the loci due to their antagonistic pleiotropic effects on strongly selected phenotypes. Birth Date Selection Mapping may readily be extended to temporally-stratified human or model organism populations.

Genetic evaluation of Angus cattle for carcass marbling using ultrasound and genomic indicators

The objectives were to estimate genetic parameters needed to elucidate the relationships of a molecular breeding value (MBV) for marbling, intramuscular fat (IMF) of yearling bulls measured with ultrasound, and marbling score (MRB) of slaughtered steers, and to assess the utility of MBV and IMF in predicting the breeding value for MRB. Records for MRB (n = 38,296) and IMF (n = 6,594) were from the American Angus Association database used for national cattle evaluation. A total of 1,006 records of MBV were used in this study. (Co)variance components were estimated with ASREML, fitting an animal model with fixed contemporary groups for MRB and IMF similar to those used in the Angus national genetic evaluation. The overall mean was the only fixed effect included in the model for MBV. Heritability estimates for carcass measures were 0.48 +/- 0.03, 0.31 +/- 0.03, and 0.98 +/- 0.05 for MRB, IMF, and MBV, respectively. Genetic correlations of IMF and MBV with MRB were 0.56 +/- 0.09 and 0.38 +/- 0.10, respectively. The genetic correlation between IMF and MBV was 0.80 +/- 0.22. These results indicate the MBV evaluated may yield a greater genetic advance of approximately 20% when used as an indicator trait for genetic prediction of MRB compared with IMF. However, neither of these indicators alone provides sufficient information to produce highly accurate prediction of breeding value for the economically relevant trait MRB. Given that the goal is a highly accurate prediction of true breeding value for MRB, results of this work point to the need to 1) continue progeny testing, and 2) continue increasing the genetic correlation between the MBV and MRB.

Impact of single nucleotide polymorphisms in leptin, leptin receptor, growth hormone receptor, and diacylglycerol acyltransferase (DGAT1) gene loci on milk production, feed, and body energy traits of UK dairy cows

The impact of 9 single nucleotide polymorphisms (SNP) in the leptin (LEP), leptin receptor (LEPR), growth hormone receptor (GHR), and diacylglycerol acyltransferase (DGAT1) gene loci on daily milk production, feed intake, and feed conversion, and weekly measures of live weight, BCS, and body energy traits was evaluated using genetic and phenotypic data on 571 Holstein cows raised at the Langhill Dairy Cattle Research Center in Scotland. Six SNP were typed on the LEP gene and 1 on each of the other 3 loci. Of the 6 LEP SNP, 3 were in very high linkage disequilibrium, meaning there is little gain in typing all of them in the future. Seven LEP haplotypes were identified by parsimony-based analyses. Random-regression allele-substitution models were used to assess the impact of each SNP allele or haplotype on the traits of interest. Diacylglycerol acyltransferase had a significant effect on milk yield, whereas GHR significantly affected feed intake, feed conversion, and body energy traits. There was also evidence of dominance in allelic effects on milk yield and BCS. The LEP haplotype CCGTTT (corresponding to leptin SNP C207T, C528T, A1457G, C963T, A252T, and C305T, respectively) significantly affected milk yield and feed and dry matter intake. Animals carrying this haplotype produced 3.13 kg more milk daily and consumed 4.64 kg more feed. Furthermore, they tended to preserve more energy than average. Such results may be used to facilitate genetic selection in animal breeding programs.

Heterozygosity for genes influencing a quantitative trait

Choosing families to sample for a quantitative trait locus mapping experiment is a critical component of experimental design because only heterozygous families contribute information to the analysis. Additive genetic variance of a paternal half-sib family can be partitioned into two parts: a variance component of maternal source that is constant across different families and a variance component of paternal source that is defined as an index of heterozygosity of a sire. This index is shown to be an upper limit of variance among marker genotypes of a half-sib family and can be used to identify highly heterozygous sires, thus improving the power of detecting QTL in detection studies. Simulated progeny phenotypic data were used to estimate sire's heterozygosity index via an ANOVA method, and accuracy of the estimation was evaluated with the correlation coefficient between the true and estimated index summarized both as the correlation and by the correct ranking of results as measured by the ratio of the true average heterozygosity index of experimentally selected parents to average heterozygosity of all sires. Positive but small correlation can be achieved in the estimation of a sire's heterozygosity when based on the daughters' phenotypic data, and accuracy was improved when progeny-tested sons were used to estimate their grandsire's heterozygosity index, depending on the genetic model of a trait and the size and structure of families.

Evaluation of calpastatin activity measures in ante- and postmortem muscle from half-sib bulls and steers

Calpastatin activity measured at 24 h postmortem in bovine longissimus muscle (PMLD24) is correlated with Warner-Bratzler shear force (WBS) measurements, an objective measure of tenderness. A live-animal measurement of calpastatin activity that correlates with 24-h postmortem activity would provide information for selection programs without the expense of progeny testing. The purpose of this study was to evaluate the effectiveness of calpastatin activity measurements obtained on tissue samples from live animals and to determine the relationship among various calpastatin activity measures and tenderness determined by WBS and sensory panel. Biopsies (approximately 10 g) were obtained surgically 2 d before slaughter from the supraspinatus muscle on the anterior surface of the scapula (LISH0) from contemporary purebred Angus bulls (n = 12) and steers (n = 17). Biopsies from a subset of these cattle (n = 12) were refrigerated at 4 degrees C to simulate the postmortem cooling process for 24 h (LISH24) prior to extraction. A rib section anterior to the 12 and 13th rib interface was collected from all animals at the commercial abattoir between 22 and 23 h postmortem for PMLD24, sensory panel, and WBS measurements. A postmortem shoulder muscle sample (PMSH24) was collected at the same time. Calpastatin was extracted from all muscle samples using a heated calpastatin activity protocol. Sensory panel tenderness, WBS, LISH0, LISH24, and PMSH24 were not different between bulls and steers. However, PMLD24 values were significantly different. Significant partial correlations were found between WBS and sensory panel tenderness (-.55), between WBS and PMLD24 (-.43), and between LISH24 and PMLD24 (.78). Therefore, similar calpastatin activity values are possible with ante- and postmortem tissue samples, suggesting the possibility of using measurements from live-tissue biopsies from other than the longissimus muscle to predict end product tenderness.

Comparison of the persistent efficacy of the injectable and pour-on formulations of doramectin against artificially-induced infection with Dictyocaulus viviparus in cattle

The persistent efficacy of the injectable and topical formulations of doramectin was compared against experimental challenges with infective larvae of Dictyocaulus viviparus in two separate studies. Four groups of 10 randomly-assigned calves, negative for lungworm larvae by the Baermann technique, were used in each study. Calves were treated subcutaneously in the midline of the neck or poured down the midline of the back with saline (1 ml/50 kg. injection: 1 ml/10 kg. pour-on) on Day 0 or doramectin (200 microg/kg = 1 ml/50 kg. injection: 500 microg/kg = 1 ml/10 kg. pour-on) on Day 0, 7, or 14. Two additional calves from the same pool of animals were randomly assigned as larval-viability monitors and received no treatment. Calves were inoculated daily with a gavage of approximately 100 larvae of D. viviparus from days 35 to 49 for the injectable study and days 28 to 42 for the pour-on study. The two larval viability monitor calves received approximately 3000 infective larvae in the same manner on Day 49 or 42 for the injectable and pour-on studies, respectively. Equal numbers of calves from each treatment group as well as the larval viability monitor calves were necropsied on days 14 and 15 after the last lungworm inoculation to enumerate the worm burden. The worms recovered were quantified and identified. For each study, geometric mean worm recoveries for each treatment group were back transformed from the natural log-transformed data (worm count +1) and were used to estimate percentage reduction. Doramectin injectable solution was 100.0% efficacious against lungworms for up to 49 days and the pour-on formulation was 100.0%, 93.1% and 81.5% effective in reducing lungworm infection resulting from challenge infection for up to 28, 35, and 42 days post-treatment, respectively.

Persistent efficacy of doramectin injectable against artificially induced infections with Cooperia punctata and Dictyocaulus viviparus in cattle

Three studies were conducted to evaluate the persistent efficacy of doramectin injectable solution against experimental challenges with infective larvae of Cooperia punctata and Dictyocaulus viviparus. In each study, four groups of ten randomly-assigned calves, negative for trichostrongyle-type eggs on fecal examination, were treated subcutaneously in the midline of the neck with saline (1 ml/50 kg) on Day 0 or doramectin (200 microg/kg = 1 ml/50 kg) on Day 0, 7, or 14. Two additional calves from the same pool of animals were randomly assigned as larval-viability monitors and received no treatment. On Days 14-28, approximately 1000 and 50 infective larvae of Cooperia spp. and D. viviparus, respectively, were administered daily by gavage to each animal in Groups T1-T4. On Day 28, the two larval-viability monitor calves were inoculated in a similar manner with a single dose of approximately 30000 and 2000 larvae of Cooperia spp. and D. viviparus, respectively. Equal numbers of calves from each treatment group were killed on Days 42-45, as well as the two viability monitor animals to enumerate worm numbers. A 2% or 5% aliquot of small intestinal contents and washings were examined for worm quantification and identification, while 100% of the lung recoveries were quantified and identified. For each study and across the three studies, geometric mean worm recoveries for each treatment group were calculated from the natural log transformed data (worm count + 1) and were used to estimate percentage reduction. In the three studies, doramectin injectable solution was 97.5% efficacious against lungworms for up to 28 days and was 99.8% efficacious in reducing infection resulting from challenge with infective larvae of C. punctata for at least 28 days post-treatment.

Haplotype construction of sires with progeny genotypes based on an exact likelihood

A maximum likelihood method is presented that can be used to construct parental haplotypes based on their progeny genotypes. The exact error rate and choice of family size in haplotype construction were evaluated through mathematical expressions and numerical examples. Numerical results suggest that, if two markers are tightly linked (< or = 10 cM) and each has intermediate allele frequencies, a difference of one between progeny receiving parental and recombinant gametes is sufficient for constructing sire linkage phase; a difference of two or more progeny is required with two markers 30 cM apart. When each of two adjacent markers has two alleles with equal allelic frequencies, genotyping 10 and 50 progeny are needed to achieve a power of 0.85 for constructing a sire linkage phase of two tightly (10 cM) and moderately tightly linked (30 cM) markers, respectively. The family size is reduced by approximately half when both markers have three alleles with equal frequencies. Results suggest that, when an experiment requiring haplotype determination of a parent is being designed, researches should choose the appropriate threshold and family size in the context of marker allelic frequencies and recombination fractions.

The journey to professional excellence: a matter of priorities

Today and for the foreseeable future, it appears that we will confront continual challenges and almost overwhelming change that could easily disillusion and distract us from our primary purpose of patient care. To avoid that possibility and to ensure our continuing journey to excellence, let us remember and teach: It is always a matter of priorities.

A two-stage half-sib design for mapping quantitative trait loci in food animals

Daughter and granddaughter half-sib designs for mapping quantitative trait loci were modified to increase experimental power. This new design includes a two-stage procedure, in contrast to conventional one-step half-sib designs. In stage 1, a few progeny of each sire are genotyped for marker loci. Based on the analyses of stage 1 data, some sires are chosen to continue genotyping more progeny for stage 2. When multiple chromosomes are under investigation, chromosomes and sires for stage 2 are selected based on the analysis of stage 1 data. Sire selection results in increased frequency of heterozygous genotypes of interest in stage 2 if the markers are linked to those genes. Chromosome selection can increase the proportion of chromosomes with segregating quantitative trait loci in stage 2 if not all of the chromosomes evaluated in stage 1 have segregating quantitative trait loci. Numerical results indicated that two-stage half-sib designs are generally more powerful than conventional designs when 1) the noncentrality parameter is moderate or larger, 2) larger quantitative trait loci are mapped using tightly linked markers in larger families, and 3) variation is large in numbers and sizes of segregating quantitative trait loci among the chromosomes evaluated in stage 1.

Disease management: opportunity for pharmaceutical care?

Most of the efforts to control health care costs in the U.S. have been based upon control of the more expensive components of health care. These efforts have typically failed or fallen short, and some studies even indicate that certain efforts to reduce costs of drugs have significantly increased other medical costs, resulting in overall increases in cost of care. A dramatic new concept called disease management advocates a much broader approach to appropriate treatment of the entire disease process. This often involves shifting care and necessary dollars away from expensive inpatient and acute care to areas such as preventive medicine, patient counseling and education, and outpatient care. This concept has great potential impact regarding drugs and drug therapy because of the implications of appropriate versus inappropriate therapy on the overall cost and clinical outcome of a particular disease. Managed care organizations, prescription benefit management companies, and pharmaceutical manufacturers all are very interested in this promising new paradigm for patient care. Likewise the concept possibly has great implications for pharmacists' successful implementation of pharmaceutical care. In this regard, disease management may have the potential to prove both clinical and economic value of appropriate drug therapy and related pharmaceutical care far beyond the cost of the drug. Some suggestions for pharmacists interested in pursuing the concept of disease management are included at the end of this article.

Development of an animal model for across-herd genetic evaluation of number born alive in swine

An animal model and computer software were developed to conduct across-herd genetic evaluations using data from producers participating in the Sow Productivity Index program of the American Yorkshire Club. The final data set consisted of 61,596 litter records from 1986 to early 1990. The animal model included fixed contemporary group effects and random additive direct, service sire, permanent environmental, and residual effects. Additive genetic relationships among animals were included. A separate relationship matrix for service sires and their sires was also included. A data set similar to the Yorkshire field data was simulated to use in testing the animal model. The simulated data set consisted of 40 herds, each with 120 reproducing dams and either four or five sires. Six generations of simulated data were produced, resulting in 20,605 litter records. These records were then evaluated using the animal model for number of pigs born alive. Finally, correlations between the true breeding values from the simulation and the predicted breeding values were computed. The correlation between the 918 true and predicted sire breeding values was considerably lower for the animal model without a service sire effect than when it was included (.53 vs .74, respectively). However, the difference was cut in half (.66 vs .77) when only sires with greater than five daughter records were included. The high accuracy of the animal model with a random service sire effect indicates that the proposed model adequately accounts for the variation found in records for number of pigs born alive.

Parameter estimation for carcass traits including growth information of Simmental beef cattle using restricted maximum likelihood with a multiple-trait model

(Co)variance component estimates were computed for retail cuts per day of age (kilograms per day), cutability (percentage of carcass weight), and marbling score (1 through 11) using a multiple-trait sire model. Restricted maximum likelihood estimates of (co)variance components were obtained via an expectation-maximization algorithm. Carcass data consisted of 8,265 progeny records collected by U.S. Simmental producers. Growth trait information (birth weight, weaning weight, and[or] postweaning gain) for those progeny with carcass data and an additional 5,405 contemporaries formed the complete data set for analysis. A total of 420 sires were represented. Three models differing in number of traits were investigated: 1) carcass traits with growth traits, 2) carcass traits only, and 3) single trait. The final models did not include postweaning gain because of convergence problems. Parameter estimates for all three models were essentially the same. Heritability estimates were .30, .18, and .23 for retail cuts per day, cutability, and marbling score, respectively. Correlations between growth and carcass traits were low except for those with retail cuts per day, which were moderate and positive. The additional information gained by adding growth traits to the carcass-traits-only evaluation lowered prediction error variances most for retail cuts per day. Little change in prediction error variances was found for cutability and marbling score. Inclusion of growth traits in future sire evaluations for carcass traits will benefit the evaluation of retail cuts per day but have considerably less effect on cutability and marbling score.