PlasGUN

PlasGUN predicts genes from plasmid metagenomic short-read datasets using a deep learning framework to improve detection of plasmid-encoded functions such as antibiotic resistance.

Key Features:

• Deep learning framework: Employs a deep learning framework for gene prediction on plasmid metagenomic short reads.
• Multiple-input Convolutional Neural Networks (CNNs): Uses a multiple-input CNN architecture to model sequence features in plasmid-derived short reads.
• Targeted for plasmids: Specifically optimized for plasmid sequences and plasmid-derived short-read metagenomic data rather than chromosome-derived sequences.
• Benchmark performance: Demonstrated superior performance on benchmark datasets of artificial short reads and provided more reliable results for real plasmid metagenomic data compared with conventional gene prediction tools.

Scientific Applications:

• Plasmid metagenomics: Facilitates gene discovery and annotation in plasmid-centric metagenomic studies relevant to microbial ecology and evolution.
• Antibiotic resistance research: Supports analysis of plasmid-encoded antibiotic resistance genes to study their dissemination.

Methodology:

Implements multiple-input Convolutional Neural Networks for gene prediction on plasmid metagenomic short reads and was evaluated using benchmark datasets of artificial short reads.