High-resolution phage-host assignment through key proteins using large language models

2026-03-20 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Mathematics & Computational Sciences, Health & Medical Research · Depth: Expert, long

Summary

The VirHost Hunter framework, published in Nature Communications on March 20, 2026, introduces an innovative method for high-resolution phage-host assignment. This framework utilizes Protein Language Models and Vision Transformers to analyze phage tails and lysins, thereby eliminating the need for full viral genomes. This approach significantly enhances prediction accuracy by capturing protein functional homology despite sequence dissimilarity. In a study focusing on disease-associated gut bacteria, VirHost Hunter outperformed existing methods, doubling phage host assignments and expanding taxonomic reach. It successfully identified previously uncharacterized phages targeting gut bacteria like Akkermansia and Prevotella. The research also established a Gut Phage Lysin Database (GPLD) and demonstrated the synthesis of a lysin specifically targeting an obesity-promoting bacterium, marking a substantial advancement in virome research and microbiome therapies.

Key takeaway

For AI Researchers and Research Scientists working on microbiome therapies, VirHost Hunter offers a robust, scalable tool for high-resolution phage-host assignment. Your team should consider integrating this framework, particularly its use of Protein Language Models and Vision Transformers, to accelerate the discovery of novel phages and lysins, potentially leading to targeted interventions against disease-associated bacteria like those linked to obesity.

Key insights

VirHost Hunter uses AI to assign phage hosts from key proteins, improving accuracy and enabling new therapeutic applications.

Principles

• Protein functional homology can be captured despite sequence dissimilarity.
• Phage tails and lysins are sufficient for high-resolution host assignment.

Method

The VirHost Hunter framework employs Protein Language Models and Vision Transformers to analyze phage tail and lysin proteins for host assignment, bypassing full genome sequencing. It identifies functional homology to boost prediction accuracy.

In practice

• Use VirHost Hunter for efficient phage-host assignment in virome studies.
• Explore the GPLD for lysin sequences targeting gut bacteria.
• Apply protein language models for functional homology detection.

Topics

• Phage-Host Prediction
• Protein Language Models
• Vision Transformers
• Microbiome Therapeutics
• Gut Virome

Code references

• YuehuaOu/Viral-Host-Hunter

Best for: AI Researcher, AI Scientist, Research Scientist

Related on AIssential

• Predicting phage–host specificity — Protein language models enhance phage–host specificity prediction by capturing biological information from viral sequences.
• ‘Treasure trove’ of antiviral proteins could inspire powerful molecular tools — Machine learning reveals a vast, previously unknown bacterial antiviral defense system, offering new biotechnological tools.
• Rethinking Genomic Modeling Through Optical Character Recognition — OpticalDNA reframes genomic modeling as OCR-style document understanding, achieving efficiency and high-fidelity compression via visual layouts.
• Programmable RNA translation through deep learning-driven IRES discovery and de novo generation — An AI framework unifies IRES identification, optimization, and de novo generation for programmable RNA translation.
• What Makes a Bacterial Model a Good Reservoir Computer? Predicting Performance from Separability and Similarity — Bacterial metabolic models can function as effective substrates for nonlinear reservoir computing.
• 🔬ESM: The Bitter Lesson is Coming for Proteins - Alex Rives, BioHub — Scaling protein language models with vast, diverse data reveals emergent biological capabilities and enables programmable design.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.