Exploring the Promoter Generation and Prediction of Halomonas spp. Based on GAN and Multi-Model Fusion Methods.

Area of Science:

• Synthetic biology
• Genetic engineering
• Extremophile microbiology

Background:

• Promoter strength is crucial for gene expression regulation in genetic engineering and synthetic biology.
• Accurate prediction and optimization of promoter strength are essential for advancing these fields.
• Existing methods lack tailored solutions for extremophilic microorganisms like Halomonas.

Purpose of the Study:

• To establish the first promoter strength database for Halomonas.
• To propose a novel AI-driven method for promoter design and prediction.
• To enhance the efficiency of genetic engineering in extremophiles.

Main Methods:

• Developed a generative adversarial network (GAN) model to learn Halomonas promoter sequence features.
• Integrated deep learning (BiLSTM, CNN) and machine learning models for multi-model fusion prediction.
• Utilized k-mer, PSSM, and engineered string/non-string features for comprehensive analysis.
• Experimentally validated newly generated and predicted promoter sequences.

Main Results:

• The GAN model successfully generated biologically plausible Halomonas promoter sequences.
• The multi-model fusion framework significantly improved promoter strength prediction accuracy, especially in top quantiles.
• Experimental validation confirmed the functional validity of predicted promoters.
• The approach reduced the experimental validation space through an intersection-based strategy.

Conclusions:

• Introduced an innovative AI-based strategy for promoter design and strength prediction in Halomonas.
• Laid a foundation for advancing industrial biotechnology using extremophiles.
• Demonstrated the versatility of GANs and multi-model fusion for promoter engineering in other extremophiles.
• Highlighted the synergy between artificial intelligence and synthetic biology for academic and practical implications.