Deep Learning-Enhanced Plant Genomic Prediction

Genomic prediction (GP) has become a crucial component of modern plant breeding, enabling early selection of high-performance genotypes based on molecular markers. While traditional statistical models, such as genomic best linear unbiased prediction (GBLUP) and Bayesian approaches, have facilitated genetic gain, they often fall short in capturing the non-linear and high-dimensional interactions inherent in plant genomes. Deep learning (DL), with its hierarchical architecture and representation learning capabilities, has emerged as a powerful alternative, offering improved accuracy, scalability, and adaptability in GP. This chapter provides a comprehensive overview of DL applications in plant GP, discussing model architectures (convolutional neural networks (CNNs), recurrent neural networks (RNNs), transformers), data integration strategies, and performance comparisons with conventional methods. It further explores case studies across key crops (e.g. corn, rice, wheat), examines challenges such as overfitting and interpretability, and discusses future prospects, including multi-omics integration and explainable artificial intelligence (XAI). The integration of DL and genomics holds the potential to revolutionize data-driven breeding for climate-resilient and high-yielding crops.