In synthetic biology, artificially designed cis-regulatory elements (CREs), such as enhancers, can be used to precisely control the yield of target products, playing a critical role in cost reduction and efficiency enhancement. However, our understanding of CRE nucleic acid syntax remains limited, and de novo design of these elements is still in its infancy. To address this challenge, we are developing CisEncoder, a platform that integrates Massively Parallel Reporter Assays (MPRAs), which provide high-quality, large-scale quantification of CREs, with DREAM (DNA cis-Regulatory Elements with controllable Activity design platforM), an innovative deep learning framework designed to unravel the nucleic acid syntax of CREs. To demonstrate the capabilities of CisEncoder, we achieved state-of-the-art sequence-based enhancer activity prediction in Drosophila S2 cells and identified key sequence features that are crucial for strong enhancer activity. Leveraging this predictive power, we designed DreaMer001, a synthetic enhancer with 3.6 times the activity of the strongest natural enhancer in the Drosophila genome. Remarkably, DreaMer001 not only showed high activity in Drosophila S2 cells but also demonstrated significant activity across multiple species' cell lines. In mammals like humans, mice, and pigs, DreaMer001 averaged over twice the activity of the CMV enhancer. In SF9 cells, its activity was 15.7 times higher than the Hr5 enhancer, and it exhibited 7.6 times and 26.6 times higher activity than the CMV enhancer in chicken DF1 cells and fish spermatogonial cells, respectively. Additionally, using MPRA-derived data, we developed the ultra-strong silencer DreaMer002, which reduced gene expression by 44.7-fold. Our study not only introduces an efficient platform for enhancer design but also establishes a general framework applicable to other CRE types, offering significant potential for designing gene expression circuits in synthetic biology. 

暂无