# Multi-Trait Genomic Prediction of Meat Yield in Pacific Whiteleg Shrimp (Penaeus vannamei)

## TL;DR

This study shows that multi-trait genomic models significantly improve the prediction of meat yield in Pacific whiteleg shrimp when auxiliary trait data are available.

## Contribution

The study demonstrates that multi-trait genomic models outperform single-trait models for predicting meat yield in shrimp, especially with auxiliary traits in the validation set.

## Key findings

- Multi-trait models improved meat yield prediction accuracy by up to 58.8% when auxiliary traits were included in the validation set.
- Excluding auxiliary traits limited accuracy gains to 8.6%.
- Incorporating abdominal segment length increased heritability estimates for other traits by 5.4–7.6%.

## Abstract

Enhancing the meat yield in Pacific whiteleg shrimp (Penaeus vannamei) is vital for aquaculture profitability. In this study, we compared single-trait and multi-trait genomic models using 899 shrimp from 63 families to predict meat yields and related traits. Results showed that multi-trait genomic models were superior to single-trait ones. When auxiliary traits (like net meat weight) were added to the validation set to build multi-trait genomic models, the prediction accuracy of the meat yield could increase by 58.8%. But if auxiliary traits were not in the validation set, such models only slightly boosted the meat yield prediction accuracy by up to 8.6%. These findings underscore the value of multi-trait genomic models in shrimp breeding programs, particularly when supplementary trait data are accessible, and highlight a practical route to accelerate improvements in commercial aquaculture.

The meat yield (MY) is a key economic trait in Pacific whiteleg shrimp (Penaeus vannamei) breeding, necessitating accurate genomic prediction for efficient genetic improvement. In this study, we investigated single-trait (STGMs) and multi-trait genomic models (MTGMs) for predicting MY and related traits, using two cross-validation strategies reflecting different data-availability scenarios. A total of 899 individuals from 63 full-sibling families were phenotyped for MY, net meat weight (MW), body weight (BW), body length (BL), and abdominal segment length (AL). We estimated the genomic heritability and genetic correlations of MY and related traits in P. vannamei, followed by comparing the prediction accuracy of STGMs and MTGMs for MY and MW. Two validation approaches were then applied: CV1 retained auxiliary traits in the validation sets, and CV2 excluded both target and auxiliary traits. Heritability estimates indicated that MY had low heritability (STGM: 0.160; MTGMs: 0.145–0.156), whereas MW, BW, BL, and AL showed low-to-moderate heritability (0.099–0.204). Genetic correlations revealed strong associations between MY and MW/BW/BL (rg = 0.605–0.783), yet a low positive correlation with AL (rg = 0.286). Across all comparisons, MTGMs consistently surpassed STGMs. For MY, MTGMs improved the accuracy by 4.8–58.8% relative to STGM (0.187), with the MY-MW model achieving the highest accuracy (0.297) under CV1. Similarly, MTGMs enhanced MW prediction by 36.6–138.2% over STGM (0.254), with the MW-BW model reaching the highest accuracy (0.605) under CV1. Notably, retaining auxiliary traits (CV1) boosted accuracy gains substantially (up to 138.2%), whereas excluding them (CV2) yielded only marginal improvements (≤8.6%). Moreover, incorporating AL as an auxiliary trait increased heritability estimates for MW, BW, and BL by 5.4–7.6%, indicating its synergistic value in MTGMs. Overall, these results demonstrate that MTGMs markedly enhance genomic prediction for carcass traits compared to STGMs, particularly when auxiliary trait data are accessible (CV1). The findings underscore the importance of maintaining auxiliary trait records in breeding populations, offering a robust framework for improving P. vannamei through multi-trait genomic prediction models.