From waste to wealth: AI-driven enzyme innovation for straw transformation

Straw is a renewable resource; however, its utilization poses challenges related to efficiency and environmental impact. Traditional enzyme engineering approaches face constraints arising from the scale of the protein sequence space and the limited availability of experimental data. Following an introduction to the structure of straw and its degrading enzymes, this review summarizes the historical development of straw-degrading enzymes from an experience-driven phase to an artificial intelligence (AI)-driven era. Within the context of AI, we highlight approaches for enzyme mining, directed evolution, and de novo generation, and provide a summary of seminal computational tools. In enzyme mining, protein large language models are employed to identify novel enzymes and distant homologs. For directed evolution, Design-Build-Test-Learn cycles are applied to improve enzyme performance. In case de novo generation, key catalytic motif design and function-based descriptions are applied to generate new protein sequences. Furthermore, we discuss the role of AI-powered enzyme engineering in enabling high-value straw-degrading enzymes applications across agriculture, green energy, and biomaterials, thereby supporting the development of a sustainable bio-based circular economy. This review provides a technical roadmap for utilizing AI to enable efficient straw utilization.