iProm-phage: A two-layer model to identify phage promoters and their types using a convolutional neural network

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Increasing number of sequenced phage genomes requires efficient annotation tools.
• Promoters are crucial DNA sequences for phage genome annotation.
• Existing tools may lack accuracy in distinguishing phage and host promoters.

Purpose of the Study:

• To develop a user-friendly bioinformatics tool, iProm-phage, for predicting and classifying phage promoters.
• To improve the accuracy of distinguishing between phage and host promoters.
• To provide a web server for easy access by researchers.

Main Methods:

• Proposed a two-layer prediction model: promoter vs. non-promoter, then phage vs. host promoter.
• Utilized a challenging negative dataset comprising promoter sequences.
• Investigated 10 feature encoding methods and machine learning algorithms, including a 1-D convolutional neural network (CNN) with one-hot encoding.

Main Results:

• The iProm-phage model demonstrated high potential and accuracy through 5-fold cross-validation.
• The one-hot encoding approach combined with the CNN model yielded superior performance.
• The approach improved discrimination and reduced false positive predictions.

Conclusions:

• iProm-phage is an effective tool for accurate phage promoter prediction and classification.
• The developed web server offers accessible functionality for the scientific community.
• This tool supports the advancement of phage-based antibacterial research through improved genome annotation.