Utilizing machine learning with phenotypic and genotypic data to enhance effective breeding in agricultural and horticultural crops

Authors: Barnaby, Jinyoung; CORTÉS, ANDRÉS - Swedish University Of Agricultural Sciences

Submitted to: Frontiers in Plant Science
Publication Type: Review Article
Publication Acceptance Date: 6/19/2026
Publication Date: 7/21/2026
Citation: Barnaby, J.Y., Cortés, A.J. 2026. Utilizing machine learning with phenotypic and genotypic data to enhance effective breeding in agricultural and horticultural crops. Frontiers in Plant Science. 17. Article 1869724. https://doi.org/10.3389/fpls.2026.1869724.
DOI: https://doi.org/10.3389/fpls.2026.1869724

Interpretive Summary: Artificial intelligence (AI) and machine learning (ML) are rapidly advancing plant breeding by improving the speed, accuracy, and scale of data-driven decision-making. This editorial highlights 14 recent studies that showcase how these tools are transforming breeding strategies through high-throughput phenotyping, trait forecasting, genomic mapping, and integration of complex biological datasets. ML models are enabling precise trait extraction from imaging platforms and improving predictions of yield, disease resistance, and environmental performance across crops like soybean, barley, maize, and lychee. In parallel, AI-assisted genomic studies are enhancing the discovery of genes linked to key traits in wheat, sorghum, and other species. Advanced ML frameworks, including deep learning and ensemble models, are increasing the accuracy of genomic selection in crops such as coffee and almonds. Meanwhile, the integration of multi-omics data using AI-driven approaches is revealing genetic networks behind important traits like protein content and flower color. Collectively, these innovations demonstrate how AI and ML are reshaping plant breeding by enabling more informed, scalable, and targeted selection decisions across a wide range of crops.

Technical Abstract: Machine learning (ML) and artificial intelligence (AI) promise transforming all disciplines, plant breeding not being the exception. As part of this Editorial, encompassing 14 scientific contributions, we explore how these recent developments can leverage high-throughput phenotyping platforms, trait forecasting, genomic-enabled trait mapping and prediction, the integration of omics data, and their overall joint impact on modern plant breeding strategies. High-throughput phenotyping (HTP) has been powered ML models like convolutional neural networks (CNNs), random forests, and gradient boosting, which enable precise trait extraction from imaging platforms for yield prediction, disease assessment, and trait inheritance studies in crops like soybean, barley, and buckwheat, respectively. In turn, phenotypic forecasting has been assisted by latent feature models such as compositional autoencoders, which improve trait prediction accuracy and cultivar clustering across environments, as respectively demonstrated in maize and lychee. Meanwhile, ML-assisted genome-wide association studies (GWAS), including multi-locus and mixed linear models, have advanced trait mapping and candidate gene detection in sorghum and wheat for in vitro regeneration and plant height. In terms of ML-enabled genomic selection, deep learning models (e.g., parallel CNNs) and stacking ensemble learning frameworks (SEL) have increased prediction accuracy for agronomic traits in staple crops and coffee, while explainable AI tools enhanced interpretability in almond breeding. Finally, ML-driven multi-omics integration, leveraging approaches like weighted gene co-expression network analysis (WGCNA) and clustering algorithms, has uncovered gene networks and regulatory pathways influencing key traits such as protein content and flower color in soybean, wheat, Brassica napus, and Impatiens uliginosa. Overall, this impressive compilation of studies demonstrate that ML and AI offer powerful tools to accelerate crop improvement as never thought before by integrating high-dimensional phenotyping, environmental and multi-omics data, and refining breeding decisions, ultimately enabling more precise and scalable selection of complex traits across diverse crop species.