From codon optimization to AI-guided sequence design for engineered biological functions

Abstract

Engineering biological systems with intended sense-and-response functions requires precise control over the expression of their genetic elements.1,2 Achieving robust and durable protein production, however, remains a central challenge in living therapeutics. In addition to regulation through promoters, enhancers, and untranslated regions, coding sequences provide another layer of control over translation efficiency, transcript stability, and protein output. Conventional codon-optimization strategies typically rely on predefined metrics, such as codon usage, GC content, and predicted RNA structure,3 but may not fully capture the context-dependent effects of synonymous sequence variation.