Retrobiosynthesis-Guided Discovery of Cryptic Secondary Metabolites: An AI-Assisted Reconstruction of Plant Metabolic Pathways

Abstract

Plants synthesise an enormous chemical repertoire—more than 200,000 known specialised (secondary) metabolites and, by most estimates, a far larger number still undetected. Yet the biosynthetic pathways of the great majority remain unresolved, and many metabolites are “cryptic”: encoded in genomes but unexpressed under standard conditions, or predicted to exist but never isolated. This constitutes a vast reservoir of untapped chemistry of high therapeutic value, given that natural products underlie more than 70% of antibiotics and over half of anticancer agents. Conventional, gene-by-gene pathway elucidation cannot scale to this challenge—a difficulty compounded in plants, whose biosynthetic genes, unlike those of microbes, are frequently dispersed rather than clustered. This review examines an emerging solution: retrobiosynthesis, the application of retrosynthetic logic to enzyme-catalysed reactions, accelerated by artificial intelligence. We describe how AI-driven bio-retrosynthesis engines (READRetro, BioNavi-NP, graph-transformer models) predict biosynthetic routes backward from a target metabolite to primary-metabolite building blocks; how machine learning and genome mining (plantiSMASH, multi-omics ML) nominate the candidate genes and enzymes for each predicted step; and how protein-structure prediction (AlphaFold, ESMFold) and sequence-similarity networks refine enzyme-function assignment before heterologous expression validates the reconstructed pathway. We propose an integrated, iterative workflow uniting these components, illustrate it with a benzylisoquinoline-alkaloid example, and critically assess the principal limitations—enzyme promiscuity, data scarcity, plant genome complexity and the gap between prediction and experimental proof. Retrobiosynthesis-guided, AI-assisted reconstruction offers a scalable strategy for illuminating plant metabolic dark matter and unlocking cryptic secondary metabolites for drug discovery and synthetic biology.