Daniela Lourenco, Un. of Georgia 6/1/17. Pedigree. Individual 1. Individual 2

Transcription

1 Traditional evaluation The promise of genomics for beef improvement Daniela Lourenco Keith Bertrand Pedigree Performance EPD BLUP Progeny Heather Bradford, Steve Miller, Ignacy Misztal BIF - Athens, GA - 06/01/2017 EPD sum of gene effects 2 What if we could know the genes that affect the traits? Individual 1 Genomic information Could we have more accurate EPD? Individual 2 Genomics in livestock breeding hnp:// thinnergene/genercs- 101/ Errors in the DNA Most are repaired Some are transmined Some influence performance Some are beneficial Some are harmful Genomic information SNP used as markers for GENES or regions in the genome that contain genes = QTL Genomic information Marker Assisted SelecRon - MAS SNP gene SNP Expensive!!! Few SNPs Meat quality Meat tenderness Feed efficiency Disease 2017 BIF Symposium, Athens, Ga. 1

2 Why MAS did not quite work? Traits of interest are polygenic gene gene gene gene What if we could use thousands of SNPs? Meuwissen Hayes Goddard 2001 Thousands of genes Thousands of SNP The promises Cost of genotyping What is 100,000 cheaper NOW than in 2001? 2 Rmes cheaper We can use thousands of SNPs Genotyping thousands of SNPs will become cheap We can calculate EPD based on SNPs (e.g., MBV) Without own performance or progeny records Accuracy of predicbng EPD more than double (0.40 vs. 0.85) Increase in accuracy for traits with low h 2 and hard to measure We can select animals earlier (reducing generaron interval) 100,000 Rmes cheaper hnps:// 5 Rmes cheaper Peak of excitement Dairy Industry Cheaper genotyping Illumina 50k SNP chip First genomic evaluaron in EvaluaRon in 2012 Who would go first? K SNP + parent informaron No daughters with records 1 Parent informaron + 100s of daughters with records Net merit = $792 7 bulls > $ BIF Symposium, Athens, Ga. 2

3 The promises Genomics in beef cattle : Angus We can use thousands of SNPs Genotyping thousands of SNPs will become cheap We can calculate EPD based on SNPs (e.g., MBV) Without own performance or progeny records Accuracy of predicrng EPD more than double (0.40 vs. 0.85) Increase in accuracy for traits with low h 2 and hard to measure We can select animals earlier (reducing generaron interval) 2012: Simmental, Hereford, Red Angus, Limousin : Charolais, Santa Gertrudis, Shorthorn, Brangus, Guelbvieh How is genomic incorporated? How is genomic incorporated? Performance 1.SNP effects Pedigree Progeny A A B B Genomic Info GE- EPD Extra source of informaron A B B B More informaron means greater accuracy of EPD MBV = = 5.7 How is genomic incorporated? 2. BeNer relaronships Which methods? MulRstep first method implemented Full- sibs I do not trust EPD, but I do trust MBV because they come straight from the DNA 2017 BIF Symposium, Athens, Ga. 3

4 Which methods? Trend Single- step Developed at UGA: MulRstep Single- step Simplicity Which software to use for single-step? UGA: blupf90 family Theta SoluRons: Bolt The biggest promise Increase (double) accuracy of predicrng EPD Meuwissen et al., 2001 Keith Ignacy Daniela Bruce Dorian Daniel Beef canle? Trait Accuracy gains in beef Breed number of genotyped animals EPD accuracy GE- EPD or MBV accuracy Gain % Author Simulated - 2, Meuwissen et al., 2001 Carcass Hereford 1, Saatchi et al., 2013 Feed intake Nellore 1, Silva et al., 2016 Growth Angus 2, Lourenco et al., 2015 Small gain due to small number of genotyped animals ~ 2,000 You should genotype more animals Number of genotyped animals Number of genotyped animals You should genotype more animals 1,600,000 1,400,000 1,200,000 1,000, , , , , , , , , Females Males US Holsteins ,000 American Angus , , , , , BIF Symposium, Athens, Ga. 4

5 Accuracy gains in beef Accuracy gains 50k vs. 500k VanRaden et al., 2011 Trait Breed number of genotyped animals EPD accuracy GE- EPD or MBV accuracy Gain % Simulated - 2, Growth Angus 2, Growth Angus 33, Author Meuwissen et al., 2001 Lourenco et al., 2015 Lourenco et al., 2015 Gain = 0.02 You should genotype more animals You are using only 50k SNP not enough you should use over 300,000 You are using only 50k SNP not enough you should use over 300,000 Sequence the whole genome Sequence information for predictions Sequence vs. genotyping > 30M 50k VanRaden et al., 2017 Small gain with more SNP Why Meuwissen et al. (2001) got it, but we did not? 1. BeNer relaronships: already accurate with 50k 2. SNP effects: only more SNP to esrmate effects without increasing phenotypes A A B A B B Assumed SNP with large effect Large SNP explained large proporron of generc variance Traits of interest are polygenic: several genes with small effect 2017 BIF Symposium, Athens, Ga. 5

6 The promises Increase in accuracy for traits with low h 2 and hard to measure We can use thousands of SNPs Genotyping thousands of SNPs will become cheap We can calculate EPD based on SNPs (e.g., MBV) Without own performance or progeny records Accuracy of predicrng EPD more Increases than double (0.40 vs. 0.85) Increase in accuracy for traits with low h 2 and hard to measure We can select animals earlier (reducing generaron interval) Increase is small Need more genotypes and phenotypes Kor Oldenbroek and Liesbeth van der Waaij, 2015 The promises We can select animals earlier Acc = 0.25 Acc = 0.40 Acc = 0.40 We can use thousands of SNPs Genotyping thousands of SNPs will become cheap We can calculate EPD based on SNPs (e.g., MBV) Without own performance or progeny records Accuracy of predicrng EPD more Increases than double (0.40 vs. 0.85) Increase in accuracy for traits with low h 2 and hard to measure We can select animals earlier (reducing generaron interval) Parent info Does it mean we do not need to collect phenotypes? There is no magic here MulRstep Single- step I like GE-EPD, but I want to see MBV MulRstep Single- step Genotype Phenotype MBV = Sum of SNP effects GE- EPD = MBV + Parent Average + Performance + Progeny More informaron = higher accuracy 2017 BIF Symposium, Athens, Ga. 6

7 Millions of genotyped animals How is it possible? Millions of genotyped animals How is it possible? More informaron = higher accuracy DNA inherited in blocks More genotypes, phenotypes, pedigree Challenge If we trace those blocks few ancestors 4,000 blocks = 4,000 animals = 90% of variance +10,000 blocks = +10,000 animals = 99% of variance Millions of genotyped animals How is it possible? New single-step = single-step with APY Single- step before 2015 Single- step now Max. load 150,000 Max. load Keep genotyping! American Angus! 300k genotyped animals! 10M pedigree! All traits! US Holsteins! >250k genotyped animals! Millions in pedigree! Several traits! US Holsteins! 760k genotyped animals! 23M pedigree! 37M phenotypes! ProducRon traits Keep in mind Promises were higher than the realized We can use genomic informaron to bener esrmate EPD Accuracy of EPD increases (not astronomically) We can select animals earlier reducing generaron interval Keep genotyping and phenotyping if you want more reliable GE- EPD Keep in mind GeneRc gain = (selecron intensity * accuracy * generc SD) / generaron interval Expected trend if no changes in 2013 δ x How much you trust in the generc advisor 2017 BIF Symposium, Athens, Ga. 7

8 Acknowledgements Shogo Tsuruta Yutaka Masuda Breno Fragomeni Ivan Pocrnic Ignacio Aguilar Andres Legarra 2017 BIF Symposium, Athens, Ga. 8