|Leeds, Timothy - Tim|
|Welch, Timothy - Tim|
|Wiens, Gregory - Greg|
Submitted to: Plant and Animal Genome Conference
Publication Type: Abstract Only
Publication Acceptance Date: 1/10/2015
Publication Date: 1/14/2015
Citation: Vallejo, R.L., Leeds, T.D., Liu, S., Gao, G., Welch, T.J., Wiens, G.D., Palti, Y. 2015. Accuracy of genomic prediction for BCWD resistance in rainbow trout using different genotyping platforms and genomic selection models. Plant and Animal Genome Conference. P726.
Technical Abstract: In this study, we aimed to (1) predict genomic estimated breeding value (GEBV) for bacterial cold water disease (BCWD) resistance by genotyping training (n=583) and validation samples (n=53) with two genotyping platforms (24K RAD-SNP and 49K SNP) and using different genomic selection (GS) models (Bayes B/C and single-step GBLUP); and (2) compare reliability of GS models for BCWD resistance with pedigree-based model (PED). Survival days (DAYS) and survival status (STAT) were recorded in training animals, and each validation animal had estimated breeding value (EBV) records. Reliability of GS models was assessed through predictive ability (PAGEBV) estimated in validation animals. Reliability was estimated as: R2GEBV = R2EBV,GEBV / R2EBV; where R2EBV,GEBV is squared correlation between GEBV and EBV, while R2EBV is EBV reliability. We present results from GS models with 49K SNP; RAD-SNP results will be presented later. Reliability of PED model was higher for DAYS (0.33) than for STAT (0.27). All GS models (0.37-0.54) had higher reliability than PED model; with exception of ssGBLUP (0.22) which was outperformed by PED for DAYS. For DAYS and STAT, Bayes B (64% and 56% increase) and C (58% and 38% increase) outperformed PED. For STAT, ssGBLUP (33% increase) outperformed PED. For DAYS, Bayes B (0.42) and C (0.41) had higher PAGEBV than ssGBLUP (0.30). For STAT, ssGBLUP (0.34) and Bayes B had similar PAGEBV, and both had higher PAGEBV than Bayes C (0.32). The accuracy of GS for BCWD resistance in rainbow trout is expected to increase with larger training and validation samples.