Cis-regulatory elements (CREs) play a crucial role in regulating gene expression by controlling transcription, making the understanding and design of these elements essential for the advancement of biology. Traditional approaches often rely on empirical rules and iterative experimentation, which can be time-consuming and labor-intensive. Recent advances in deep learning have begun to influence this field by improving the accuracy of predictions for existing elements and offering preliminary strategies for designing synthetic CREs. Specialized design models can incorporate high-throughput experimental data, and DNA foundation models draw on pre-trained genomic representations to inform the design process. These approaches have shown encouraging progress in generating promoters, enhancers and more complex regulatory architectures. Nonetheless, substantial challenges remain, including limited data availability, gaps between computational predictions and experimental outcomes, and limited model interpretability. Moreover, although AI-driven methods hold considerable promise for CRE prediction and design, their generative capabilities are still constrained by data quality and by the tendency of current models to rely predominantly on sequence-level features without fully capturing broader regulatory context. In this review, we examine how emerging AI technologies may support more systematic and targeted design of synthetic CREs, and we discuss key challenges and future directions, including multimodal modeling, reinforcement learning (RL), and system-level regulatory network design.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号