Automated annotation and validation of human respiratory virus sequences using VADR.

Area of Science:

• Virology
• Bioinformatics
• Genomics

Background:

• Accurate viral genome annotation is crucial for research and public health data sharing.
• The National Center for Biotechnology Information's (NCBI) Viral Annotation DefineR (VADR) pipeline offers standardized annotation but currently supports limited viral groups.

Purpose of the Study:

• To develop and validate new VADR models for 12 key human respiratory viruses.
• To improve the accuracy and scope of viral genome annotation using reference sequence-based approaches.

Main Methods:

• Constructed genotype-specific VADR models based on comprehensive genomic and phylogenetic diversity analysis.
• Trained models on over 5,300 public viral genomes and tested on 372 novel genomes.
• Validated model performance in identifying genomic features and misassemblies.

Main Results:

• Achieved high VADR pipeline performance, passing 96.3% of public and 98.1% of novel viral genomes.
• Successfully identified complex genomic features like overlapping ORFs and transcriptional slippage.
• Discovered novel viral biology, including an HCoV-OC43 NS2 knockout and new coding sequence lengths in human metapneumovirus.

Conclusions:

• The developed VADR models significantly enhance viral genome annotation capabilities for key respiratory viruses.
• These models are publicly available and integrated into the NCBI GenBank submission pipeline, supporting scalable and high-quality annotation.
• The findings contribute to improved viral surveillance, research, and public health preparedness.