A New Beginning of Rational Design in the Age of Large Language Models of Synthetic Biology

Synthetic biology aims to design predictable functional modules by integrating high-dimensional multi-omics data with large-scale computational models. However, the inherent complexity of biological regulation and cellular stochasticity continue to challenge traditional engineering principles. This review systematically examines the evolution of biological sequence modeling, tracing the trajectory from early convolutional neural networks (CNNs) to contemporary genome-scale foundation models and sub-quadratic architectures. Moving from isolated component optimization to system-level synthesis, we evaluate current progress in generative genetic design and analyze the persistent representation gap arising from unaccounted systemic variables and the non-linear dynamics of cellular environments. Furthermore, we discuss the role of mechanistic interpretability in transitioning from statistical correlation to engineering causality, alongside the emergence of autonomous, closed-loop laboratories. Finally, we address technical safety and sequence authentication, providing a perspective on the realization of programmable biological systems grounded in biological logic.