190 likes | 287 Views
Methods to Include Foreign Information in National Evaluations. Introduction. Animal model (1989 to present) Single-trait milk, fat, and protein Others (PL, SCS, DPR) added later All-breed animal model – 2007 Multiple-step genomic model - 2009 One step evaluation (future)
E N D
Methods to Include Foreign Information in National Evaluations
Introduction • Animal model (1989 to present) • Single-trait milk, fat, and protein • Others (PL, SCS, DPR) added later • All-breed animal model – 2007 • Multiple-step genomic model - 2009 • One step evaluation (future) • All breeds, traits, and genotypes might be evaluated together
One Step Evaluation Tests • Include foreign phenotypes using MACE EBVs of bulls • Single (ST) or multi-trait (MT) national evaluations • 7-trait model: yield, SCS, daughter pregnancy rate (DPR), MT and ST productive life (PL) • 14-trait model with conformation
Adding Foreign Information • Foreign portion of deregressed MACE EBV added as “daughters” • Adapted from Bonaiti, Boichard 1995 • Subtract domestic portion of MACE • Add 1 observation weighted by EDC instead of adding n “daughters” • For multi-trait, multiply by D.5 T-1 D.5 • Mean included to adjust foreign data to genetic base of solutions
Data (Yield and Health) • Foreign deregressed PTAs for 118,741 bulls (foreign dams not yet included) • Deregession: pure foreign or mixed PAIB + ( PTAIB – PAIB ) / RELIB-PA PTAdom + ( PTAIB – PTAdom) / RELIB-dom
Data (Yield and Health) • One step model includes: • 72 million lactation phenotypes • 50 million animals in pedigree • 29 million permanent environment • 7 million herd mgmt groups • 11 million herd by sire interactions • Traits: M, F, P, SCS, PL, DPR • Genotypes not yet included
Previous Features Retained • Official animal model • Data edits and adjustments • Separate weight for each observation • All breeds and crossbreds included • Inbreeding and heterosis correction • Herd by sire interaction • Convert PTAs to within-breed base • Heritability of yield decreased • Mimic effect of cow adjustments
New Features Added • Model options now include: • Multi-trait models • Multiple class and regress variables • Suppress some factors / each trait • Random regressions • Foreign data • Parallel processing • Genomic information (Paul’s talk) • Renumber factors in same program
Traditional, ST, and MT EvaluationDomestic bulls (≥ 10 USA daughters)
Foreign Yield, Health Data: Results • Test bulls with no USA daughters • Corr (National EBV, MACE EBV) • 0.77 before adding foreign data • ~ 0.99 after adding foreign data • Interbull uses only ST longevity • USA computes both ST and MT in the same one-step program with genetic correlations set = 0 for ST
MACE Data in National EvaluationForeign proven bulls (no USA daughters)
Adding April 2012 MACE Data Domestic bulls (≥50 USA daughters)
Computation Required • CPU for all-breed model (7 traits) • ST: 4 min / round with 7 processors and ~1000 rounds • MT: 15 min / round and ~1000 rounds • ~200 rounds for updates using priors • Little extra cost to include foreign • Memory required • ST or MT: 32Gbytes (256 available)
Jersey Conformation Evaluation • 716,645 cows, most with 14 traits • ST required 1.75 hours / 1000 rnds • MT required 5 hours / 1000 rnds • Foreign data not added yet • Conformation traits currently use a multi-trait system.
Conclusions • Foreign data can add to national evaluations • In one step model instead of post-process • High correlations of national with MACE • Multi-trait all-breed model developed • Replace software used since 1989 • Many new features added • Correlations ~.99 with traditional AM • Tested with 7 yield and health traits • Also tested with 14 JE conformation traits
Future Research • Adding genomics • Adding foreign dams • Adding supporting variables • Improving reliability calculations • Integrating one-step output for routine delivery
Acknowledgments • George Wiggans provided advice on algorithms and modeling