TECHNICAL SUMMARY October 2013 GeneSTAR MVPs Moleculr Vlue Predictions for beef feed efficiency, 1 mrbling 2 nd tenderness Key Points GeneSTAR is DNA-mrker test for importnt production trits in ll breeds of beef cttle. GeneSTAR MVPs employ pnel of 56 DNA mrkers to produce GeneSTAR Moleculr Vlue Predictions (MVPs). GeneSTAR MVP relibility vlues verge 30, 26 nd 39 percent for feed efficiency, mrbling nd tenderness, respectively. GeneSTAR MVPs hve been internlly nd externlly vlidted in vriety of beef cttle popultions globlly. Introduction GeneSTAR is DNA test for qulity nd production trits in ll breeds of beef cttle. It tests DNA mrkers known to be ssocited with feed efficiency, mrbling nd tenderness. GeneSTAR ws initilly relesed in 2000 s single-mrker test for n importnt gene for mrbling. Advnces in niml genomics hve llowed GeneSTAR to expnd to include dditionl mrkers for mrbling nd the trits listed bove. The rte t which new mrkers/trits re incorported into the GeneSTAR product rnge is ever incresing, ided over the pst few yers by n exponentil growth in our knowledge of the bovine genome. Until now, four DNA mrkers ech for feed efficiency nd mrbling nd three mrkers for tenderness hve been utilized in the GeneSTAR pltform. Previous GeneSTAR results hve been presented on scle of 0 to 8 strs for ech trit (ech mrker hving result of zero, one or two strs with the overll trit result clculted by dding the results for ech individul mrker together). Under this system, the higher the number of strs, the better the niml ws likely to perform for tht trit. As the true vlue of gene mrker tests lies in pnels of mrkers, Pfizer Animl Genetics ims to incorporte dditionl mrkers into these pnels s they re discovered through rnge of ongoing gene mpping nd sequencing projects. These mrker pnels become tools for predicting the true breeding vlue of n niml nd better enble customers to utilize individul niml precision mngement prctices nd fully exploit genome-enbled genetic improvement progrms. This comprehensive pproch is the cornerstone of Pfizer Animl Genetics long-term gol of providing full service genomic solutions to the globl cttle industry.
GeneSTAR MVPs GeneSTAR Moleculr Vlue Predictions (MVPs) s described in this summry hve been in development since 2006 bsed on exclusively licensed technology ugmented with publicly vilble mrkers. The pltform utilizes pnel of 56 DNA mrkers to clculte MVPs for three core mngement trits. These trits include: Feed efficiency s defined by net (or residul) feed intke (NFI or RFI) Mrbling (mesured s USDA degree of mrbling) Tenderness (mesured s Wrner-Brtzler sher force t 14-dy postmortem ging) The commercil development of GeneSTAR MVPs hs involved n extensive nd rigorous four-step process: 1) Initil discovery of DNA mrkers tht hve sttisticlly importnt effects on performnce in economiclly relevnt trits of interest. 2) Development of sttisticl methodology for the inclusion of individul mrkers into genome-wide pnels for prediction of MVPs nd ssocited prediction ccurcies. 3) Evlution of the reltionship between MVPs nd phenotypic performnce in rnge of popultions within the Pfizer Animl Genetics reserch progrm. 4) Third-prty vlidtion of the effects of the mrker pnel predictions of MVPs. Inclusion of Mrkers in GeneSTAR MVPs The first gene mrker distributed under GeneSTAR ws the mrbling mrker (M1). This product ws offered commercilly in 2000. Since 2004 three dditionl mrbling nd three tenderness mrkers hve been dded. The ddition of ech mrker ws mjor undertking with n extensive mount of externl nd internl discovery nd vlidtion work required. As the technology nd the field of niml genetics dvnced, four feed efficiency mrkers were incorported in 2006. This cme s result of the first bovine genome-wide nlysis using publicly vilble 10,000 single nucleotide polymorphism (SNP) rry. GeneSTAR MVPs now incorporte 56 DNA mrkers for use in multiple-trit prediction of feed efficiency, mrbling nd tenderness. The previously offered 11 mrkers re included in the new pnel long with n dditionl 45 mrkers from more recent discovery nd vlidtion efforts. These dditionl 45 mrkers hve been identified from rnge of sources including recent Pfizer Animl Genetics whole genome scn, public domin informtion nd series of collbortive reserch projects. More detiled informtion on the discovery effort involved in the production of both the originl 11 mrkers nd the recently relesed dditionl 45 mrkers cn be found t www.pfizernimlgenetics.com. All 56 mrkers contined in GeneSTAR MVPs re SNPs, DNA sequence vritions tht occur when single nucleotide (A, T, C or G) in the genome sequence is ltered t prticulr loction in the genome. All SNPs used in the GeneSTAR MVPs re di-llelic, mening they hve two possible lleles which cn be combined to form three possible genotypes (i.e., if we ssume the two lleles of SNP re Q nd q, then the three possible genotypes for tht mrker re QQ, Qq nd qq). The mrkers included in this new pnel hve ll been found to hve sttisticlly significnt effects on one or more trits of interest. GeneSTAR MVP Methodology The reporting nd interprettion system for the GeneSTAR pltform hs undergone significnt expnsion in order to ccommodte the dditionl informtion explined by GeneSTAR MVPs. In the er of incresingly lrge mrker pnels, the im is to predict the true breeding vlue (nd resulting phenotype) of n individul s opposed to simple genotype result. The previous process simply counted the number of fvorble lleles zero, one or two for prticulr mrker, which ws then trnslted into number of strs. With the reliztion tht not ll mrkers crry similr effects 2
nd tht individul mrkers my contribute to more thn one trit simultneously, simple str rting my fil to best represent the true genetic vlue of n individul. GeneSTAR MVPs employ much lrger number of mrkers s shown in Tble 1. The pproch tken to improve both the relibility nd utility of GeneSTAR test results hs been to develop Moleculr Vlue Prediction (MVP) of n niml s true breeding vlue. The ccurcy of the MVP is mximized by fitting mrker dditive effects s regression on the number of fvorble lleles. Under this pproch, ll individul mrker effects re ssumed to be dditive, but not necessrily equl. The resulting MVP is then produced by dding the individul mrker effects of the genotypes observed for n niml cross ll mrker loci in the pnel, ccording to the model shown in Tble 2. In order to ensure tht the MVP is not only ccurte but lso conservtive in the fce of complex predictive equtions deling with genotypic informtion, shrinking methodology (Kinghorn B. 2008) ws developed for Pfizer Animl Genetics. This pproch llows for the shrinking of individul mrker effects bsed on the stndrd error of the mrker genotype effect before contributing to the overll MVP. The shrinking pproch pproximtes the results tht would be expected from whole genome nlysis using thousnds of mrkers; however, s only 56 mrkers re currently used, the potentil vrince ccounted for is less thn would be possible from full-scle whole genome nlysis. As with ll predictions of genetic vlue, ssessment of relibility nd ccurcy is importnt. The primry stndrd for ssessing the ccurcy nd predictive power of the MVP for trit is the relibility vlue. Relibility is bsed on the correltion (R) between the MVP nd the niml s genetic breeding vlue if ll informtion were known. Becuse MVPs re predictions of genetic vlue (i.e., breeding vlue) for the trit, the mximum vlue for R is the squre root of the heritbility of the trit in the popultion. The relibility vlue describes this rtio nd is expressed s percentge of the mximum ccurcy ttinble. The relibility is highly useful indiction of how much dditionl informtion my be dded in the future s greter numbers of mrkers re dded to the pnels used to clculte the MVP. Tble 1 From STARS to MVPs. Number of Mrkers Results Expressed Relibility Vlue Vlidted Originl GeneSTAR 11 STARS b NA c New GeneSTAR MVP 56 Trit Prediction d 26 39% e Four mrkers ech for feed efficiency nd mrbling nd three for tenderness b Number of fvorble lleles c Not pplicble d Estimte of moleculr breeding vlue e % Relibility vlue rnges cross trits Tble 2 Moleculr breeding vlues t single-mrker locus. Moleculr Breeding Genotype Number of Fvorble Alleles Vlues s Devition from Popultion Men Error Vrince QQ 2 (2 2p) (2 2p) 2 V Qq 1 (1 2p) (1 2p) 2 V qq 0 (0 2p) (0 2p) 2 V Moleculr breeding vlue s the devition from popultion men nd its error vrince for genotypes with zero, one or two copies of the fvorble llele of mrker ( is the substitution effect of the mrker, p is the frequency of the fvorble llele, V is the squre of the stndrd error of substitution effect estimte). 3
Reltionship of GeneSTAR MVPs to Phenotypic Performnce GeneSTAR Moleculr Vlue Predictions hve been evluted in reserch popultions in Austrli (AU) nd the United Sttes (U.S.) selected to provide representtive rnge of commercil breeds nd situtions. Tble 3 describes the popultions. Feed efficiency is mesured s net feed intke (NFI/RFI); mrbling is mesured s mrbling score (MS) nd qulity grde (USDA QG); nd tenderness is mesured s Wrner-Brtzler sher force (WBSF) nd sensory pnel tenderness score (ST). GeneSTAR MVPs were developed using popultions A through E (N = 2,866) with ll other popultions (N = 4,455) used to evlute nd further refine the clibrtion of the prediction equtions. Summry results of the reltionship of GeneSTAR MVPs with phenotypic performnce re presented in Tble 4 (on the next pge). The criteri for this reltionship re the sttisticl significnce of the prediction on the trit of interest (where P<0.05 is the minimum threshold of significnce) nd the correltion (R) between the MVP nd djusted phenotype. While individul popultions cn vry in level of heritbility for trits, verge levels for the trits of interest result in mximl vlues of R of 0.50, 0.52 nd 0.55 for feed efficiency, mrbling nd tenderness, respectively. Predictions from the new 56-mrker pnel were compred in ll popultions with predictions from the previously vilble GeneSTAR mrkers. The verge improvement in R observed rnged from 45 percent for mrbling to 250 percent for feed efficiency. The GeneSTAR MVPs re significntly improved cross ll popultions studied s compred to the previous mrker pnels vilble. While results vried cross popultions studied internlly (summrized in Tble 4), the overll trend for the MVP relibility vlues ws on the order of 30 percent for feed efficiency, 26 percent for mrbling nd 39 percent for tenderness. Tble 3 Description of evlution popultions, mngement regimens s dys on feed (DOF) nd verge trit performnce. Popultion Breed Composition Regimens (DOF) N Phenotype Exmined A (AU) Wgyu 300 450 605 MS B (AU) Angus 250 920 MS C (AU) Angus 150 180 235 332 MS d, NFI D (AU) Composite b <150 755 MS E (AU) Shorthorn Progeny Test 100 180 114 254 MS, NFI F (AU) British-X <180 508 MS G (US) Angus-X 150 872 MS e H (AU) Angus x Hereford <180 387 MS I (US) Multi-breed c 150 180 828 1037 MS e, WBSF, ST J (AU) Angus 150 180 348 MS f K (US) L (US) M (US) Angus Brngus Snt Gertrudis Bulls: 70 d Feed Intke Heifers: 70 d Feed Intke Steers: 70 d Feed Intke 4 605 NFI 468 NFI 206 230 MS e, WBSF, NFI Mrbling is AUS-Met Mrble Score (0 9) or USDA Mrbling Score (4.0 = Smll 0 degree of mrbling); Net Feed Intke is the difference between n niml s ctul nd expected feed intke bsed on its body weight nd growth rte; W-B Sher Force (kg); Sensory Pnel Tenderness (1 8). b Composite contining Bos Indicus nd Bos Turus breeds. c This popultion includes Angus, Brhmn, Chrolis, Hereford, Limousin, Mine-Anjou, Red Angus, Shorthorn nd South Devon breeds. d Number of observtions vried by trit within the rnge given. e Mrbling score ssessment of degree of intrmusculr ftness under the USDA grding system (4.0 = Smll 0 degree of mrbling). f Mrbling mesured s % IMF.
Tble 4 Associtions between GeneSTAR MVPs nd trit phenotypes from internl popultions. Fitting 56-mrker pnel,b,c Correltion % Genetic Trit Rnge in N (R) d % Relibility e Vrition f P-vlue g Feed Efficiency Overll 1,749 0.13 29.8 12.1 Rnge 114 605 0.02 0.28 3.4 63.1 0.1 39.8 0.87 0.001 Mrbling Overll 6,127 0.12 25.5 9.2 Rnge 206 1,037 0.02 0.28 3.1 61.5 0.1 37.8 0.84 0.01 Tenderness Overll 2,030 0.30 39.0 24.6 Rnge 206 996 0.17 0.35 30.5 53.1 9.3 28.2 0.02 0.01 Prediction of MVPs bsed on the 56-mrker pnel. b Overll vlue for R, % relibility, nd % genetic vrition is clculted s the weighted verge over ll popultions in which the trit ws evluted. c Rnge in vlues cross popultions is shown in the second row. d The mximum vlue possible for R is the squre root of the heritbility of the trit. e Rtio of R to the squre root of the heritbility of the trit. f Percentge of dditive genetic vrition ccounted for by vrition in the MVPs. g Rnge in P-vlues cross the popultions evluted is given to mtch the rnge in N. Results of Third-Prty Independent Vlidtion Anlyses The finl step in the development of GeneSTAR MVPs ws third-prty vlidtion studies to confirm commercil relevnce. One group tht plyed n importnt role in this process ws the U.S. Ntionl Beef Cttle Evlution Consortium (NBCEC, http://www.nsci.cornell.edu/nbcec/nbcec. html). The purpose of NBCEC vlidtion is to independently verify ssocitions between genetic tests nd trits climed by commercil genotyping compny using phenotypes nd DNA from different cttle popultions. In the NBCEC system, experimentl popultions with well-defined phenotypic records re used in the vlidtion process. In this cse, Pfizer Animl Genetics provided MVPs of the cttle to the NBCEC for nlysis of the reltionship to trit performnce in ech popultion. The results of this vlidtion pproch re confirmtion of the clims mde for ech trit in the mrker test. Until recently, the NBCEC service ws the only publicly vilble model for DNA-mrker test vlidtion. Relizing importnt differences in the production nd mrketing environments between Austrli, New Zelnd nd North Americ, similr vlidtion process ws crried out in Austrli. These vlidtions were commissioned by Pfizer Animl Genetics with the Coopertive Reserch Centre for Beef Genetic Technologies (Beef CRC, http://www.beefcrc.com.u/us-beefdn-results), who commissioned the Animl Genetics nd Breeding Unit (AGBU) of the University of New Englnd (Armidle, NSW) to conduct the nlysis following the sme guidelines used in the NBCEC model. Bsed on Pfizer Animl Genetics cumultive internl dt, the verge relibility vlues were on the order of 30 percent for feed efficiency, 26 percent for mrbling nd 39 percent for tenderness. 5
Tble 5 provides summry of the popultions used in externl vlidtion while Tble 6 presents the results from the individul nlyses of these popultions. Tble 5 Description of vlidtion popultions, mngement regimens s dys on feed (DOF) nd verge trit performnce. Phenotype Phenotypic Performnce Popultion Breed Composition Regimen (DOF) N Exmined Men SD e A Europen Mternl 462 WBSF 4.14 0.07 Bulls 100 130 (NBCEC) Line Composite 671 NFI -0.08 0.84 785 MS 4.08 0.55 B Europen Mternl Steers 150 180 785 IMF 2.99 0.68 (NBCEC) Line Composite 723 % Choice 0.57 0.49 392 MS 4.28 0.97 Bos Indicusinfluenced 392 % Choice 0.57 0.50 394 IMF 3.63 0.56 C Steers: 70 d (NBCEC) Feed Intke Composites b 390 WBSF f 3.61 0.98 395 NFI 0.03 0.93 1228 MS 1.00 0.88 D 1345 IMF 4.86 1.99 Multi-breed (CRC) Vrious d 1377 NFI 0.03 1.11 1244 WBSF 4.73 1.24 MS = AUS-Met Mrble Score (0 9) for popultion D nd USDA Mrbling Score (4.0 = Smll 0 degree of mrbling) for popultions B nd C; IMF = % intrmusculr ft estimted by ultrsound scnning for popultions B nd C, nd crcss intrmusculr ft estimted chemiclly for popultion D; % Choice is the percentge of nimls grding USDA Choice (CH) or higher; WBSF = Wrner- Brtzler sher force t 14-d ging (kg); NFI = net feed intke. b Equl percentge of Simbrh, Red Brngus nd Brford mternl line composites. c Breeds included Angus, Brhmn, Belmont Red, Hereford, Murry Grey, Snt Gertrudis nd Shorthorn. d Cttle were either grss finished or grin finished to either domestic specifictions (90 dys on feed), Koren (150 dys) or Jpnese (300 dys). e SD = Stndrd devition f Wrner-Brtzler sher force t one-dy ging. Tble 6 Associtions between pnel effects nd phenotypes in vlidtion popultions. MVP Trit Feed Efficiency Mrbling Mrbling Mrbling Tenderness Wrner-Brtzler sher force t one-dy ging. Vlidtion Phenotype Popultion N NFI MS IMF % Choice WBSF Regression Coefficient Fitting 56-mrker pnel Correltion (R) P-vlue A 671 0.400 0.340 0.020 C 395-0.020 0.010 >0.500 D 1377 0.340 0.080 0.001 B 785 0.120 0.040 0.146 C 392 0.370 0.070 0.096 D 1228-0.010 0.001 >0.500 B 785 0.310 0.090 0.011 C 394 0.660 0.130 0.001 D 1345 0.210 0.020 0.430 B 723-0.030 0.010 >0.500 C 392 0.276 0.110 0.028 A 462 0.670 0.310 0.021 C 390 0.360 0.210 0.007 D 1244 0.197 0.060 0.001 6
Overll Met-nlysis of Results In order to summrize the collective results from studies in the development nd vlidtion of the MVPs, met-nlysis ws designed nd conducted. The metnlysis effectively combined the sttisticl results into outcome groups by geogrphic region (Austrli nd North Americ) nd cttle subpopultions (Bos Turus nd Bos Indicus-influenced). The results of this metnlysis re shown in Tble 7 for North Americ, Austrli nd the combintion of the North Americn nd Austrlin dtsets, respectively. This nlysis reveled: Vlidtion for tenderness MVPs cross ll popultions nd mrkets Vlidtion in both North Americ nd Austrli for the influence of the feed efficiency MVP in Bos Turus popultions Vlidtion of feed efficiency MVPs for Bos Indicus-influenced popultions in Austrli with further substntition of this ssocition in internl North Americn dt For mrbling MVPs significnt ssocition ws substntited in NBCEC vlidtions with % IMF estimted by ultrsound scnning 60 dys prior to hrvest. Furthermore, significnt ssocition ws documented in North Americn Bos Indicus-influenced popultions with % Choice. Reporting Formt The GeneSTAR MVP pltform produces more informtion on n individul niml thn ny previous itertion of the GeneSTAR test. The outcome of this test is no longer reported ccording to the originl str system, but s GeneSTAR MVP. An exmple of the new GeneSTAR MVP reporting formt is shown below. Moleculr Vlue Prediction FE Mrb Tend MVP -0.73 0.01-0.93 % Rnk 26% 49% 7% The GeneSTAR MVP is expressed in units of the trit, devited on +/- scle from zero where the zero point is defined s the verge of ll nimls. To llow benchmrking of nimls within contemporry breed subpopultions, the overll breed verge nd individul niml percentile rnks re lso reported. Additionlly, verge nd rnge of relibility vlues re reported for ech trit in the report summry (Figure 1). Tble 7 Met-nlysis summry of ssocitions between MVPs nd phenotypes recorded s P-vlues. 3 NFI Mrb IMF Tend North Americn Popultions Bos Turus 0.023 0.175 0.008 <0.001 Bos Indicus-influenced 0.449 0.096 0.001 0.051 North Americn Combined 0.085 0.096 <0.001 <0.001 Austrlin Popultions Bos Turus 0.007 0.376 0.296 0.126 Bos Indicus-influenced 0.008 0.309 0.124 <0.001 Austrlin Combined <0.001 0.442 0.101 <0.001 Combined North Americn nd Austrlin Popultions Bos Turus <0.001 0.244 0.007 <0.001 Bos Indicus-influenced 0.053 0.125 0.001 <0.001 Externl Combined <0.001 0.111 <0.001 <0.001 7
Figure 1 Exmple of the summry pge in the new GeneSTAR MVP reporting formt. Moleculr Vlue Prediction FE Mrb Tend Breed: Angus Breed MVP Sttistics: Averge 0.48 0.17-0.17 Min -0.84-0.30-0.92 Mx 1.29 0.52 0.57 Herd MVP Sttistics: Averge 0.53 0.19-0.20 Min -0.38-0.25-0.83 Mx 1.21 0.52 0.57 Herd Distribution of Animls Reltive to Breed Qurtiles: Top 25% (76 100 qurtile) 5% 52% 29% 51 75% (qurtile) 21% 27% 25% 26 50% (qurtile) 25% 18% 25% Bottom 25% (1 25 qurtile) 48% 4% 21% Herd Relibility Sttistics: Averge 33.8% 21.8% 38.9% Min 26.0% 15.8% 25.9% Mx 35.9% 22.9% 41.4% Conclusions nd Implictions for Use The GeneSTAR MVP tool produces Moleculr Vlue Predictions bsed on pnel of 56 DNA mrkers for feed efficiency, mrbling nd tenderness. These MVPs hve verge relibility vlues of 30, 26 nd 39 percent for the three trits, respectively. These MVPs hve been further vlidted in thirdprty reserch in both Austrlin nd North Americn cttle popultions representing wide rnge of breeds nd production systems. GeneSTAR MVPs re directly pplicble to genetic improvement in these economiclly relevnt trits for both seedstock nd commercil cow-clf production systems. They re lso pplicble s tools for use in predicting phenotypic performnce of cttle in commercil production systems, mking them vluble tools for fcilitting precision mngement to enhnce profitbility cross the beef vlue chin. (Complete technicl reports summrizing the reserch nd development studies underlying the GeneSTAR MVP system re on file s Pfizer Animl Genetics Technicl Reports 57, 58 nd 59.) 1 Vlidted in Bos Turus popultions only, dditionl informtion vilble on file for Bos Indicus. Feed Efficiency is mesured s Net Feed Intke. 2 In NBCEC vlidtion, the mrbling MVP is significntly ssocited with Ultrsound s % IMF t 60 dys prior to hrvest. 3 Full met-nlysis dt on file. 2012 Pfizer Inc. All rights reserved. GeneSTAR nd MVP re registered trdemrks of Pfizer Inc, its ffilites or relted compnies. 8 GST09002