TL;DR
The dbAMP 2.0 database has been significantly updated to provide a more comprehensive and user-friendly resource for antimicrobial peptides (AMPs). It now includes an expanded collection of AMPs, enhanced prediction tools for identifying AMPs from genomic and proteomic data, and a novel 3D structure viewer for both experimentally validated and predicted structures, accelerating AMP research and drug development.
What Problem Does This Paper Address?
The increasing threat of antibiotic resistance necessitates the discovery and development of novel antimicrobial agents. Antimicrobial peptides (AMPs) are a promising class of such agents due to their broad-spectrum activity and ability to overcome existing resistance mechanisms. However, the rapid growth of AMP-related data makes manual curation and analysis challenging. Existing databases, while valuable, often lack comprehensive scanning capabilities for large genomic and proteomic datasets or integrated tools for structure prediction and functional annotation. This paper addresses the need for a centralized, up-to-date, and functionally enhanced resource that simplifies the identification, analysis, and structural understanding of AMPs, thereby accelerating their therapeutic development.
Methods
The dbAMP 2.0 update introduces several key methodological advancements:
- Enhanced Text Mining: A new text mining system employing natural language processing (NLP) has been implemented to automatically extract and curate AMP-related information from scientific literature, significantly improving the efficiency of data collection.
- AMPfinder Pipeline: This is a novel, integrated computational pipeline designed for identifying AMPs from large genomic, metagenomic, and proteomic datasets. It combines open reading frame (ORF) prediction (using Prodigal) with accurate AMP classification. The prediction is based on homology searching against existing AMP databases and machine learning models trained on established AMP datasets.
- Two-Stage AMP Target Prediction: An enhanced prediction scheme uses a machine learning-based model to identify specific AMP targets. This involves a first stage to distinguish AMPs from regular peptides and a second stage to characterize their specific functions against various microbes (bacteria, viruses, fungi) and cell types (cancer, mammalian). This utilizes peptide descriptors and gradient boosting decision trees (GBDT) with imbalanced learning strategies.
- 3D Structure Prediction and Visualization: For AMPs lacking experimentally determined structures in the Protein Data Bank (PDB), 3D structures are predicted using the I-TASSER server. This allows for automated protein structure prediction and structure-based functional annotation, providing valuable insights into AMP mechanisms. A new 3D structure viewer integrates these predicted and experimental structures.
Datasets
- dbAMP Database: The core dataset comprises curated AMP sequences, antimicrobial proteins, and associated data. dbAMP 2.0 has accumulated 26,447 AMPs and 2,262 antimicrobial proteins from 3044 organisms, curated from over 4579 research articles.
- PDB: Experimentally validated AMP structures were sourced from the Protein Data Bank (PDB).
- Training Data for Machine Learning Models: The AMP prediction module in AMPfinder was trained on a dataset adapted from previous work, containing 2399 AMPs and 26,850 non-AMPs. The two-stage AMP prediction model utilized datasets for various tasks, including AMP identification, anti-Gram-positive, anti-Gram-negative, antiviral, antifungal, anticancer, and mammalian inhibition.
Key Insights and Ideas
- AMPs as a Crucial Alternative to Antibiotics: The paper strongly emphasizes the growing importance of AMPs as a viable alternative to traditional antibiotics, especially in the face of escalating antibiotic resistance and the ongoing COVID-19 pandemic.
- The Power of Integrated Bioinformatics Tools: The development of AMPfinder and the enhanced prediction models highlights the critical role of computational tools in accelerating AMP discovery and characterization from vast biological datasets.
- Structure-Function Relationship is Key: The integration of 3D structure prediction and visualization underscores the understanding that structural characteristics are fundamental to AMP function and mechanism of action, crucial for drug design.
- The Need for Comprehensive, Up-to-Date Resources: The continuous expansion and improvement of dbAMP demonstrate the ongoing need for centralized, well-annotated databases to support the research community.
Key Results
- Database Expansion: dbAMP 2.0 now contains 26,447 AMPs and 2,262 antimicrobial proteins from 3044 organisms, a significant increase from its previous version.
- Enhanced Prediction Accuracy: The AMPfinder pipeline and the two-stage prediction model demonstrate high performance, with Mean Accuracy (MeanACC) values often exceeding 80% for various prediction tasks (e.g., 89.78% for general AMP prediction, 84.96% for antiviral prediction).
- Structural Insights: Over 458 predicted 3D structures have been added to the database, complementing the existing experimentally validated structures.
- Efficient Genome Scanning: AMPfinder can efficiently identify AMPs from large genomes and proteomes, offering a significant advantage over tools focused on smaller datasets or specific species.
Limitations and Future Work
- Limitations of Predicted Structures: While I-TASSER provides valuable predicted structures, they are not a substitute for experimentally determined ones and may have inherent limitations in accuracy. The paper notes that for stapled peptides, structural prediction is an alternative due to a lack of validated 3D structures.
- Ongoing Data Curation: The paper acknowledges that the database is continuously updated through manual curation and text mining, implying that the data is a snapshot in time and will evolve.
- Future Work: The authors aim to maintain dbAMP in real-time, incorporating cutting-edge developments and studies of novel therapeutic candidates. They aspire to establish dbAMP as a comprehensive encyclopedia of AMPs for scientific research. They also plan to continue improving the prediction models and expanding the database's content and functionalities.
Summary updated