Related Experiment Video
Updated: Sep 8, 2025
A Virtual Machine Platform for Non-Computer Professionals for Using Deep Learning to Classify Biological Sequences of Metagenomic Data
Published on: September 25, 2021
DeephageTP: a convolutional neural network framework for identifying phage-specific proteins from metagenomic
Yunmeng Chu1,2, Shun Guo1, Dachao Cui1
1Shenzhen Key Laboratory of Synthetic Genomics, Guangdong Provincial Key Laboratory of Synthetic Genomics, CAS Key Laboratory of Quantitative Engineering Biology, Shenzhen Institute of Synthetic Biology, Shenzhen Institutes of Advanced Technology, Chinese, Shenzhen, Guangdong, P.R. China.
We developed DeephageTP, a CNN-based framework to identify specific phage proteins (Portal, TerL, TerS) in metagenomic data. This tool aids in discovering novel phages by accurately detecting conserved and divergent phage protein sequences.
Area of Science:
- Microbiology
- Bioinformatics
- Computational Biology
Background:
- Bacteriophages are abundant but have many unassigned genes, hindering identification and functional annotation.
- Identifying phage genomes and genes in metagenomic data is challenging due to low sequence conservation and lack of universal markers.
- Specific proteins like Portal, TerL, and TerS are key markers for Caudovirales phage identification.
Purpose of the Study:
- To develop a novel computational framework for identifying specific phage proteins from metagenomic data.
- To improve the efficiency and accuracy of detecting novel phage genomes and their functional elements.
- To address the challenge of identifying phage proteins with high complexity and low sequence conservation.
Main Methods:
- Developed DeephageTP, a convolutional neural network (CNN)-based framework for protein sequence analysis.
- Utilized one-hot encoding of protein sequences as input for automatic feature extraction.
- Implemented a cutoff-loss-value strategy to mitigate false positives and enhance precision.
Main Results:
- DeephageTP achieved high precision (94% for TerL, 90% for Portal) on a mimic metagenomic dataset.
- The framework demonstrated superior performance over alignment-based methods in identifying novel phage proteins with remote homology.
- Successfully identified specific phage proteins in three real-world metagenomic datasets.
Conclusions:
- DeephageTP is the first CNN-based framework for identifying complex, low-conservation phage-specific proteins.
- The framework significantly enhances the discovery of novel phages within metagenomic sequencing data.
- DeephageTP offers a powerful tool for advancing phage genomics and metagenomic research.
