De novo promoter design method based on deep generative and dynamic evolution algorithm.

Area of Science:

• Synthetic biology
• Computational biology
• Molecular biology

Background:

• Promoters are essential for controlling gene expression, vital for metabolic engineering and gene therapy.
• Current promoter design methods often prioritize high expression, overlooking the need for tunable regulatory intensities.

Purpose of the Study:

• To introduce PromoDGDE, a novel deep learning framework for designing synthetic promoters with specific expression levels.
• To demonstrate the method's efficacy in *Escherichia coli* and *Saccharomyces cerevisiae*.

Main Methods:

• Utilizing Diffusion-GAN to learn natural promoter sequence features and generate novel candidates.
• Employing reinforcement learning and evolutionary algorithms for dynamic optimization of synthetic promoter sequences.

Main Results:

• PromoDGDE outperformed existing methods in silico, generating biologically relevant promoters with desired functions.
• In vivo experiments confirmed the expression activity of synthetic promoters, with over 60% achieving expected regulatory effects.

Conclusions:

• PromoDGDE offers an effective and flexible solution for designing synthetic promoters with precise regulatory control.
• The method has significant implications for advancing applications in synthetic biology, including metabolic pathway construction and gene therapy.