In Increasing accuracy of genomic predictions: fr from SNP chips - PowerPoint PPT Presentation

In Increasing accuracy of genomic predictions: fr from SNP chips to sequence data Daniela Lourenco Shogo Tsuruta and Ignacy Misztal 6/12/2020 52nd Beef Improvement Federation Research Symposium and Convention

Genomics in in animal and pla lant breeding Use of DNA markers • Construct genetic relationships • Parentage determination • Identification of genes/QTL 2

SNP - Sin ingle nucleotide poly lymorphisms Individual 1 “SNPs have become the bread-and-butter of DNA sequence variation” (Stonecking, 2001) Individual 2 http://www.thinnergene.com/about-thinnergene/genetics-101/ 3

What are SNPs doin ing? gene gene gene • Identify genes • Track relationships • Parentage determination 4

How many SNP in in the genomic test? • Livestock, poultry, aquaculture: 3,000 - 800,000 SNP https://www.sciencelearn.org.nz /images/2543-dna-precipitate https://www.lihs.cuhk.edu.hk/en-us/corefacilitiesandresources/corefacilities/geneexpressionprofiling,analysisandgenotyping.aspx 5

Genomic Selection Pedigree Performance Genomic Progeny data GEPD

Adoption of genomic testing 460,000 3,400,000 780,000 http://www.angus.org/AGI/default.aspx https://www.usjersey.com/AJCA-NAJ- http://www.holsteinusa.net/programs_services/backgrounds.html JMS/AJCA/AnimalIdentificationServices/HerdRegister.aspx ~50,000/line 50,000 - 70,000 ~50,000/line 2,000 - 10,000 www.sheepcrc.org.au/ http://sesenfarm.com/raising-pigs/ 7 https://www.cobb-vantress.com/

Benefits of genomic testing in in beef cattle Average gain in accuracy 52k animals = 25% Angus 335k animals = 36% 0.47 0.40 0.39 0.38 Genotyped for 50k SNP 0.35 0.34 0.29 0.29 0.23 The gain in accuracy is because genomics is a non-redundant piece of information! BW WW PWG BLUP ssGBLUP14 ssGBLUP17 Lourenco et al., 2018 8

Benefits of genomic testing Cole, 2020 https://www.angus.org/AGI/GenomicEnhancedEPDs.pdf 9

Benefits of genomic testing Genetic gain = (selection intensity * accuracy * genetic SD) / generation interval 10

Benefits of genomic testing in in dairy ry cattle Genotyped for ~ 60k SNP • Increase in genetic gain • ~50 to 100% for yield traits • 3- to 4-fold for traits with low heritability (fertility) 11

Benefits of genomic testing in in pig igs Genotyped for ~ 60k SNP ssGBLUP Selection Index 12 2019 2006 2013/2014

Benefits of genomic testing • Increase in accuracy • Number of genotyped animals • Amount of performance records • Heritability of the trait • How informative the genomic test is • 50k SNP in beef cattle Is the genomic test informative enough? Do we have enough SNPs in the test? Are we looking at the right SNPs? 13

About the rig ight SNP • Is the genomic testing informative enough? • Do we have enough SNPs in the test? • Are we looking at the right SNPs? 38k SNP 54k SNP cattle 44k SNP 54k SNP pigs 39k SNP 54k SNP chicken 14

Sequence in information http://mtc.science/playing-around-with-ngs-step-by-step 50k SNP 30M SNP 15

SNPs capture rela lationships at genes • Expected relationship between a bull and its grandparents? • Observed relationships based on proportion of alleles shared between a bull and its paternal grandsire and maternal grandsire? With sequence data, we may find SNPs that give more precise information about genes because they may be 25.8% 15.4% closer to genes Jared Decker, University of Missouri 16

Sequence in information ~ 50k SNP ~ 30M SNP • Trying to get a more precise information about the genes • More SNPs to cover larger areas of the DNA • Select SNP that are closer to genes for traits of interest 17

Sequencing animals • Becoming cheap • Sequence a small portion of animals • Imputation from 50k to sequence (filling the gaps) • 18,000 to 107,000 pigs • Sequenced 2% of the animals • Imputation to 20M to 30M • Imputation accuracy of 94% - 98% 18

Largest sequence data in in pig igs • Gain in accuracy • Multibreed • Persistency of accuracy • Up to 100,000 sequenced/imputed pigs • 20M to 30M SNP Do we work with all ~30M SNP? 19

Min ining sequence data Traits Adapted from: https://www.broadinstitute.org/visuals/explainer-genome-wide-association-studies 20

Min ining sequence data Birth weight in Angus • Causative SNP • SNP give information about birth weight 21 Aguilar et al., 2019

Selecting causative SNP • Issues • Effective population size • Number of animals with sequence and phenotypes • Large data – more confidence in finding causative SNP • Small data – more difficult to identify – more susceptible to errors 22

Selecting causative SNP 10 genes Pocrnic et al., 2020 of equal effect Small population 500 SNP to trace 10 genes 23

Selecting causative SNP Pocrnic et al., 2020 Ne = 60 Ne = 60 Ne = 600 Animals = 6000 Animals = 6000 Animals = 18000 24

Selecting causative SNP • Small data: not enough power to discover which SNP are truly causative • Methods are also important Mancin et al., 2020 Classical GWAS EMMAX ssGWAS 25

Usin ing selected SNP to compute GEPD Jang et al., 2020 • Hanwoo cattle from South Korea • 545k Marbling score records • 1.3M animals in pedigree • 1160 genotyped animals • Imputation: 50k to 777k to 11.1M SNP en.prnasia.com • 321k SNP selected out of 11.1M sequence SNP 26

Gain in in accuracy with selected SNP Jang et al., 2020 0.27 0.27 0.26 No gain in accuracy with sequence SNP 0.19 0.19 0.18 0.18 0.16 0.16 Small data: animals with SNP and records GBLUP/BayesR = ~ 1200 records ssGBLUP = 545k records BayesR GBLUP ssGBLUP 50k SEL 50k+SEL 27

Gain in in accuracy with selected SNP • VanRaden et al. (2017) • 27k US Holstein bulls • Selected 16.6k SNP based on effect size on 33 traits • Added to the 60k SNP chip Gain in accuracy by using selected sequence SNP 0.05 0.03 0.01 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 -0.01 -0.03 Milk Average Stature Fat -0.05 28 Protein

Gain in in accuracy with selected SNP • Data from VanRaden et al. (2017) – Bayes A • 27k US Holstein bulls • Selected 16.6k SNP based on effect size on 33 traits • Added to the 60k SNP chip • Fragomeni et al. (2019) – ssGBLUP Stature in US Holsteins 0.871 0.869 0.854 0.828 BayesA ssGBLUP 60k 60k+SEL 29

Sim imulated data – real genes • Simulated population • 180,000 animals • 30,000 genotyped • 60,000 SNP • 100 genes (QTN) • Single-step genomic predictions based on • SNP • SNP + QTN • QTN 30

Accuracy wit ith sim imulated genes 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.49 60k SNP - single-step 0.53 60k SNP + 100 QTN - single-step 0.52 0.89 60k SNP + 100 QTN - single-step "true" weights 0.75 0.99 100 QTN - single-step 100% 10% Fragomeni et al. (2017) 31

Accuracy wit ith sim imulated genes 60k + 100 QTN – ssGBLUP APY – True weights • Greatest accuracy if • all genes • locations are known • weights are optimal QTN Fragomeni et al. (2017) 32

SNP closely linked SNP that turn Amount of to genes genes on and off information they give Is it Is it possible to in increase accuracy of genomic EPD wit ith sequence data in in real cattle populations? We don’t know Knowing a bit Lots of work in all the genes may help a bit progress 33

Acknowledgements 34