CDBProm

CDBProm predicts bacterial promoter sequences and annotates their genomic associations for regulatory and functional genomics applications.

Key Features:

• Machine learning model: employs an Extreme Gradient Boosting (XGBoost) classifier to distinguish promoter regions from random downstream sequences with reported 87% accuracy.
• Dual-classifier strategy: a secondary XGBoost classifier is trained on instances misclassified by the primary model to capture distinctive promoter signals.
• Genome-scale application: applied to over 55 million upstream regions extracted from more than 6,000 bacterial genomes.
• Promoter catalogue: contains over 24 million predicted promoter sequences from bacterial genomes.
• Genomic mapping: maps predicted promoters to corresponding genomic data and links predicted promoters with their coding DNA sequences and facilitates the identification of gene functions regulated by these promoters.
• In-silico prediction: performs computational promoter prediction from extracted upstream genomic regions.

Scientific Applications:

• Functional genomics: supports identification of promoter–gene associations to study gene regulation in bacteria.
• Regulatory network modeling: provides predicted promoter locations to inform models of transcriptional regulation and network topology.
• Comparative genomics: enables comparison of promoter sequences and regulatory elements across bacterial genomes.
• Quantitative bacterial genomics: supplies large-scale promoter prediction data for statistical and genome-wide analyses.

Methodology:

Upstream regions were extracted from bacterial genomes and classified using an XGBoost classifier; a secondary XGBoost classifier was trained on instances misclassified by the first model; predictions were mapped to genomic coordinates and linked to coding DNA sequences to associate promoters with gene functions.