Rider framework boosts RNA virus detection in metatranscriptomes using deep learning
Researchers from Westlake University have introduced Rider, a novel two-stage computational framework designed to enhance the discovery of RNA viruses within complex metatranscriptomic data. The system addresses the significant challenges posed by extreme sequence divergence and variable viral lengths, which often hinder traditional detection methods. By combining a compact pretrained protein…
Key points
- Rider combines deep learning with structural validation to detect divergent RNA viruses in metatranscriptomes.
- The framework recalled 99% of known RNA viruses across 10,000 metatranscriptomes, matching LucaProt sensitivity.
- Rider identified thousands of new, deeply divergent viral sequences and clades missed by existing methods.
The first stage of the framework uses the language model to broadly detect potential viral candidates, overcoming the limitations of input length constraints. The second stage employs an overlapping sliding-window strategy for structure prediction, allowing for precise alignment against a curated database of non-redundant RNA-dependent RNA polymerase (RdRp) domains. This domain-centric design ensures accurate detection even in truncated fragments or massive polyproteins.
In validation across 10,000 metatranscriptomes from diverse ecosystems, Rider demonstrated sensitivity comparable to the established LucaProt tool, successfully recalling 99% of previously cataloged RNA viruses. Furthermore, the framework identified thousands of new, deeply divergent sequences and clades that existing methods missed. In a specific application to a human inflammatory bowel disease cohort, Rider extended detection capabilities to divergent lineages, offering a scalable pipeline for global virome exploration.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us.
More in Research
All →- Retrieval‑Augmented LLM Boosts Intersection Safety Recommendations from Crash Narratives · 1 src
- Comment on arXiv:2607.01233: Survivorship Bias Concerns · 1 src
- Typos Disrupt Prompt‑Injection Probes in LLMs, Study Finds · 1 src
- New framework optimizes LLM inference costs via adaptive model activation · 1 src
- Functionalizer cuts vocab size and boosts code modeling with lossless tokenization · 1 src
Comments
via GitHub Discussions