Predicting novel mosquito-associated viruses from metatranscriptomic dark matter.

Area of Science:

• Virology
• Bioinformatics
• Genomics

Background:

• Metatranscriptomic studies generate vast amounts of unclassified viral sequences (virome dark matter).
• Mosquito-associated viruses are categorized as Mosquito-specific viruses (MSV) or Arboviruses, differing in host range.
• Accurate classification is crucial for arboviral surveillance and understanding virus-host interactions.

Purpose of the Study:

• To develop a robust computational pipeline (MosViR) for classifying novel mosquito-associated viruses.
• To identify complex genomic patterns for distinguishing between MSV and Arboviruses.
• To analyze existing metatranscriptomic data to discover new viruses.

Main Methods:

• Developed the MosViR pipeline utilizing machine learning models to predict viral classification.
• Incorporated genomic discriminatory patterns for classifying viral contigs as short as 500 bp.
• Validated pipeline performance using Area Under ROC (AUC) values > 0.99 on test datasets.

Main Results:

• Analyzed 24 published mosquito metatranscriptomic datasets.
• Identified 605 novel mosquito-associated viruses.
• Discovered eight putative novel Arboviruses with high confidence scores.

Conclusions:

• The MosViR pipeline provides a highly accurate and powerful tool for classifying mosquito-associated viruses.
• Current homology-based methods have limitations in identifying novel viruses.
• MosViR significantly advances arboviral surveillance and the study of the mosquito RNA virome.