Harnessing artificial intelligence in genomic selection: A new era in plant breeding

Abstract

Artificial intelligence (AI) and genomic selection (GS) is transforming modern breeding of plants whereby the intricate characteristics can be precisely forecasted using abundant genomic and phenomic and environmental data sets. This review provides an overview of the current advance in deep learning, machine learning and high-throughput phenotyping (HTP) of genomic prediction on crops of key crops. The AI models, CNNs, LSTM and Transformers, are better models of the non-linear interactions between genes and gene and environment interactions compared to the classical statistical models. The integration with UAV-based phenotyping, ground sensors and multi-omics data will further go a step further to enhance the accuracy of the prediction and minimise the breeding cycles. Besides, the novel solutions like multi-task learning, federated learning, and explainable AI improve the model performance, its scalability, and comprehensibility. No matter whether the problem of data quality, computational needs, and model transparency is present, AI-based genomic selection is a prospective model to increase the genetic gain and contribute to climate-resilient crop production. Such innovations are a significant advancement towards Breeding 5.0 and sustainable global food security which involve data-led approaches.