While classical statistics have long been the standard for data analysis in plant breeding, the authors explain that these methods often struggle with the complex, nonlinear nature of plant characteristics caused by the interaction between a plant’s genotype and its environment (G x E).
The researchers point out that as agricultural data expands into large-scale “big data”—including genomics, phenomics, and metabolomics—traditional models become less efficient at interpreting the results. According to the study, nonlinear and nonparametric machine learning (ML) techniques are more effective at handling these complex datasets, especially when dealing with multiple independent and dependent variables.
The authors highlight several specific models, such as neural networks, partial least square regression, random forest, and support vector machines, which have been successfully applied to both traditional breeding and lab-based biotechnology.
The researchers explain that the high interpretive power of ML allows for better classification of plant genotypes, more accurate modeling of quantitative traits, and the optimization of in vitro breeding methods. Furthermore, they note that precision agriculture is made possible by combining imaging techniques with ML to analyze high-throughput phenotyping data.
Ultimately, the authors suggest that these techniques will inspire researchers to apply machine learning to new layers of plant breeding in future studies.
Learn more about this study here: https://doi.org/10.3390/agriculture10100436
Reference:
Niazian, M., & Niedbała, G. (2020). Machine Learning for Plant Breeding and Biotechnology. Agriculture, 10(10), 436.
