Advances and critical evaluation of autonomous protein engineering: towards transparent, accessible, and reproducible platforms

Protein engineering aims to enhance enzymatic properties such as activity, selectivity, stability, and solvent tolerance by restructuring protein frameworks beyond natural performance limits. This process relies on the iterative Design-Build-Test-Learn cycle, where experimental feedback guides progressive improvements. Advancements in artificial intelligence have transformed both the Design and Learn phases, with zero-shot protein language models predicting beneficial mutations directly from sequence data and supervised models integrating assay results to refine subsequent variant designs. These approaches reduce the dependence on structural insights while enabling the discovery of synergistic effects across mutations. Automation technologies, including robotic liquid handlers and integrated platforms, have become central to modern protein engineering by reducing errors, ensuring reproducibility, and enabling large-scale variant screening. Emerging autonomous platforms demonstrate closed-loop optimization that couples protein library design, automated plasmid transformation/protein expression and corresponding assays, and machine learning-driven decision-making. These systems achieve significant accelerations in the research process, reducing multi-round engineering cycles from months to days while successfully improving diverse proteins/enzymes to a targeted objective. Beyond single-lab platforms, orchestration frameworks adhering to FAIR data principles and leveraging knowledge graphs promise distributed 'self-driving' laboratories capable of coordinating workflows across facilities. While high setup costs and proprietary systems remain challenging, open-source and modular alternatives highlight a path toward transparent, flexible automation. Collectively, these innovations are redefining protein engineering as an increasingly autonomous, data-driven discipline.