Accelerating promoter identification and design by deep learning.

Area of Science:

• Molecular Biology
• Synthetic Biology
• Bioinformatics

Background:

• Promoters are crucial DNA sequences regulating gene transcription, impacting cellular growth and lifespan.
• Engineered promoters offer precise control over gene expression for applications like natural product biosynthesis.
• Traditional promoter engineering methods include rational design and directed evolution.

Purpose of the Study:

• To review the application of deep learning (DL) techniques in promoter engineering.
• To highlight DL-driven methods for promoter identification, strength prediction, and de novo design.
• To discuss the influence of data quality, feature extraction, and model architecture on DL model performance.

Main Methods:

• Review of recent literature on deep learning applications in promoter engineering.
• Analysis of DL techniques including generative models for promoter design.
• Discussion of factors affecting predictive accuracy, such as database quality and feature engineering.

Main Results:

• Deep learning models show significant potential for accurate promoter identification and strength prediction.
• Generative DL models enable the de novo design of novel promoters with desired characteristics.
• Database quality, feature extraction methods, and model architecture critically impact DL model performance.

Conclusions:

• Deep learning is transforming promoter engineering, offering powerful tools for biological control.
• Further development of robust DL models requires attention to data quality and methodological choices.
• Future perspectives include advancing DL for more sophisticated and reliable promoter design and engineering.