TIMER is a Siamese neural network-based framework for identifying both general and species-specific bacterial...

Area of Science:

• Microbiology and Molecular Biology
• Bioinformatics and Computational Biology
• Genomics and Gene Regulation

Background:

• Bacterial promoters are crucial DNA regions initiating gene transcription, essential for bacterial growth and adaptation.
• Existing machine learning predictors often lack generalizability, being species-specific with limited performance.
• Accurate promoter identification is vital for understanding bacterial gene expression and function.

Purpose of the Study:

• To develop a robust computational tool for identifying both general and species-specific bacterial promoters.
• To enhance the predictive performance beyond existing methods for bacterial promoter recognition.
• To provide a user-friendly web server for accessible promoter prediction.

Main Methods:

• Development of TIMER, a Siamese neural network-based approach for promoter identification.
• Utilizing DNA sequences as input for three Siamese neural networks with attention layers.
• Training and optimization on 13 species-specific and general bacterial promoter datasets.

Main Results:

• TIMER demonstrated competitive performance in identifying general and species-specific bacterial promoters.
• Extensive cross-validation and independent tests confirmed TIMER's superior predictive accuracy.
• TIMER outperformed several existing methods in bacterial promoter prediction tasks.

Conclusions:

• TIMER offers a significant advancement in bacterial promoter identification accuracy and scope.
• The developed Siamese neural network approach effectively addresses limitations of current predictors.
• The publicly accessible TIMER web server facilitates broader research and application in bacterial genomics.