iPro-MP: a BERT-based model to predict multiple prokaryotic promoters.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Promoters are essential cis-regulatory elements in prokaryotes, controlling gene expression and adaptation.
• Accurate prokaryotic promoter identification is vital for understanding biological functions.
• Existing prediction tools often focus on model organisms and require improved accuracy.

Purpose of the Study:

• To develop a novel, accurate framework for prokaryotic promoter prediction.
• To evaluate the framework's performance across diverse prokaryotic species, including non-model organisms.
• To provide a superior tool for promoter identification compared to existing methods.

Main Methods:

• Development of iPro-MP, a transformer-based prediction framework.
• Utilizing a multi-head attention mechanism to analyze DNA sequence patterns.
• Systematic evaluation across 23 phylogenetically diverse species.

Main Results:

• iPro-MP effectively captures patterns in DNA sequences using multi-head attention.
• Cross-species prediction highlighted the need for species-specific models.
• iPro-MP achieved an AUC exceeding 0.9 in 18 out of 23 tested species, demonstrating outstanding performance.

Conclusions:

• iPro-MP offers superior performance in prokaryotic promoter prediction, particularly for non-model organisms.
• The framework advances the field of gene regulation analysis in prokaryotes.
• Source code and datasets are available for public use.