Rapid technological progress has enabled plant biologists to accumulate unprecedented volumes of multi-scale, multi-modal data, yet this abundance of data has intensified the challenge of translating complexity into biological understanding. Foundation models (FMs), large-scale artificial intelligence (AI) systems pretrained on millions of sequences, structures, or images and adaptable to diverse tasks are breaking through this barrier. Across plant science, these FMs are already making an impact: genomic FMs decode regulatory grammar, protein FMs enable rational protein engineering, vision FMs score phenotypes at breeding-population scale, single-cell FMs annotate cell types across species, and FM-powered AI agents accelerate knowledge retrieval and automate research workflows. While experimental validation remains indispensable, foundation models empower plant scientists to accelerate scientific discovery.
Haopeng Yu (Tue,) studied this question.