Estimation & Prediction

Evaluating Modelling Strategies and Genotype by Environment for Mortality in Crossbred Layers

---

Abstract

In laying hens, the long‑term genetic emphasis on egg number and persistency has extended productive cycles but also introduced challenges such as skeletal fragility, alongside behavioral causes of loss (pecking and feather cover damage) that can elevate mortality under cage housing. The objective of this study was to first, benchmark three genetic modelling strategies for mortality/survival in commercial crossbred laying hens and second, quantify genotype by environment (G×E) interaction for end-of-cycle mortality across a cage occupancy gradient to inform breeding and management strategies. Data comprised 18,510 reciprocal crossbred hens (A×B: 10,297; B×A: 8,213) with 26,843 pedigree records (≤6 generations) housed in cage-based systems across 10 houses on six farms (overall mortality 12%; survival time recorded in weeks, with right-censoring at 100 weeks). For model benchmarking, analyses were restricted to one farm (6,095 hens in 1,303 cages; 4-6 hens/cage). Linear, threshold (probit liability), and Weibull proportional-hazards frailty models were fitted with fixed effects of cross and generation/house (confounded by design), and random cage and additive animal effects (pedigree-based relationship matrix). Linear and threshold models were estimated by Gibbs sampling (gibbsf90+), and the survival model by Bayesian inference (R-INLA). Model choice materially shifted variance partitioning and ranking. The linear model produced h² = 0.10 ± 0.03 and cage share = 0.14 ± 0.01; the threshold model increased estimates to h² = 0.25 ± 0.07 and cage share = 0.29 ± 0.04; and the Weibull model yielded h² = 0.24 ± 0.02 with a larger cage component (0.35 ± 0.02). EBV concordance was high between linear and threshold (Spearman ρ = 0.98) but moderate versus Weibull (ρ = 0.70-0.71). For G×E, a threshold random-regression animal model using cage occupancy as a gradient (3-13 hens/cage) identified a substantial cage contribution (0.43 ± 0.05) and non-zero genetic slope variance (0.36 ± 0.23), indicating heterogeneous genetic sensitivity to number of hens per cage. Estimated breeding values agreement between low and high cage occupancy was moderate (ρ = 0.54) with limited overlap among top 5% animals (193/938). In conclusion, linear modelling compresses genetic and cage signals for binary mortality, whereas threshold and survival models better capture underlying liability and time-to-event information; additionally, cage occupancy drives actionable G×E interaction, supporting selection tailored to housing conditions and targeted cage-level management to mitigate shared social and micro-environmental risk.

Keywords: 2026

How to Cite:

BIADA, I., Casto Rebollo, C., Cavero Pintado, D., Ibáà±ez Escriche, N. & Rodriguez Juan, M., (2026) “Evaluating Modelling Strategies and Genotype by Environment for Mortality in Crossbred Layers”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2285381. doi: https://doi.org/10.31274/wcgalp.23669

Rights: 1