Plenaries

Are black-boxed methods the future for genomic prediction in animal breeding?

---

Abstract

The explosion of genomic and phenotypic data has transformed animal breeding, enabling the use of increasingly complex statistical and machine learning approaches for prediction. Agnostic or "black-box" models -such as non-parametric, semi-parametric or deep learning models- have demonstrated strong predictive ability, often matching or surpassing classical approaches with biological assumptions. Yet, these biological assumptions are often simplistic, with limited added value to interpret the complexity of biological systems. The phenomics era bring new opportunities to incorporating biological mechanisms in the prediction of complex traits. It is reasonable to question whether black boxed methods will be able to accommodate this information to deliver more accurate predictions than biologically informed methods, particularly under complex genotype-by-environment (Gà—E) interactions. This paper discusses examples from genomic prediction using black-boxed methods, including machine learning methods as well as other more classical methods such as mixed models or structural equation models, which are still low level black-boxed approaches. For instance, AI-based predictive models have been used to predict the role of feed additives in the rumen microbial community, and their effect on methane production. New AI algorithms can find candidate regions for gene editing after predicting expected effects on the trait. Mediation and moderation analysis have gained attention to accommodate additional -omic and biological information in a low-level agnostic manner to predict complex traits.The talk will highlighting the trade-offs between predictive accuracy, interpretability, and biological relevance. The goal will be to illustrate how agnostic - type models can combine biologic information in methods or algorithms that can both provide accurate genomic prediction and insights on the underlying biological mechanisms of complex traits.

Keywords: 2026

How to Cite:

Gonzalez-Recio, O., (2026) “Are black-boxed methods the future for genomic prediction in animal breeding?”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2247837. doi: https://doi.org/10.31274/wcgalp.23381

Rights: 1