AI-Guided Design and Predictive Modeling of Synthetic Escherichia coli Promoters through Comprehensive -10/-35 Box...

Area of Science:

• Molecular Biology
• Synthetic Biology
• Bioinformatics

Background:

• Promoters regulate gene transcription, with -10 and -35 boxes critical for strength.
• Quantitative relationships between promoter sequence and transcriptional output are poorly understood, hindering rational promoter design.

Purpose of the Study:

• To develop a quantitative framework linking promoter sequence composition to transcriptional activity.
• To engineer synthetic promoters with predictable and tunable expression levels using artificial intelligence.

Main Methods:

• Construction of a synthetic promoter library by varying RNA polymerase binding energies at -10 and -35 boxes.
• Fluorescence-activated cell sorting and sequencing to identify functional promoters.
• Development of a deep learning platform integrating convolutional neural networks and generative adversarial networks for promoter prediction and design.

Main Results:

• Identification of 20,799 distinct promoters with an 80-fold range of expression strengths.
• Discovery of specific -35 box sequences conferring high activity across various -10 partners.
• AI platform achieved high accuracy in predicting promoter strength (r=0.84) and designing promoters with user-defined strengths (r=0.85).

Conclusions:

• Established a bidirectional deep learning framework connecting promoter sequences (-10/-35 boxes) to transcriptional activity.
• Expanded sequence diversity of functional promoters in E. coli.
• Provided a predictive platform for rational promoter engineering and deciphered combinatorial motif interactions.