Host innate immune response profiling reveals hidden viral infections across diverse animal species

This study introduces a computationally efficient, host-response-based framework that quantifies interferon-stimulated gene expression to rapidly identify hidden viral infections across approximately 210,000 diverse animal RNA-seq datasets, effectively detecting highly divergent viruses that conventional homology-based methods often miss.

The Gist

The Big Idea: Finding Viruses by Listening to the "Alarm Bells"

Imagine you are trying to find a thief in a massive, dark warehouse filled with thousands of different people. The traditional way to do this is to check everyone's ID card against a "Wanted List." If the person isn't on the list, you assume they are innocent.

The Problem:

1. The List is incomplete: New thieves (viruses) are invented every day. If a thief wears a mask or a disguise (a highly mutated virus), the ID check fails, and they walk right past you.
2. It's too slow: Checking 200,000 people one by one against a giant list takes forever and requires a supercomputer.
3. False Alarms: Sometimes, a person is holding a prop that looks like a weapon (contamination), and you mistakenly arrest them.

The New Solution:
Instead of checking ID cards, this paper suggests listening for the sirens.

When a virus enters a body, the body's immune system doesn't just sit there; it screams. It turns on a specific set of "alarm genes" called Interferon-Stimulated Genes (ISGs). These are like the body's internal fire alarms. Even if the virus is wearing a perfect disguise, the alarm still goes off.

The Tools: The "Siren Detector" and the "Smart Filter"

The researchers built two main tools to solve the problem:

1. ISG Profiler: The Universal Siren Detector

Think of this as a high-tech microphone that can hear the specific frequency of a fire alarm, no matter what language the building is in.

• How it works: It scans RNA data (the genetic instructions inside cells) from animals ranging from chickens to bears. It doesn't need to know the specific species of the animal beforehand. It just looks for the "alarm genes" turning up the volume.
• The Magic: It works even if you don't have a reference map for that specific animal. It's like a translator that understands the concept of an alarm, regardless of the dialect.

2. ISG-VIP: The Smart Security Guard

This is a computer program (an AI) trained to look at the "Siren Detector" data and say, "Is this a real infection, or just a noisy day?"

• Training: It learned by looking at millions of samples. It knows that if the alarms are blaring, there's likely a virus.
• The Filter: It acts as a prescreen. Instead of checking every single animal sample with the slow, heavy-duty ID check, ISG-VIP quickly scans them. If the alarms are quiet, it says, "Safe, move on." If the alarms are loud, it flags the sample for a deep dive.

What They Found: The Hidden World

The team used these tools to scan 210,000 RNA samples from public databases (a massive library of genetic data). Here is what they discovered:

• The "Invisible" Viruses: They found viruses that the old "ID check" method completely missed. These were highly mutated viruses that looked nothing like the ones on the "Wanted List." Because the body's alarm went off, the new method caught them.
• Chicken Hepatitis: They found a new virus in chickens that causes liver inflammation (hepatitis). It was hiding in plain sight in liver samples, but the body's immune response gave it away.
• The "Rodent Connection": They found a virus in rats that looks very similar to the deadly viruses that infect pigs and dogs (like Parvovirus). This suggests that these dangerous animal viruses might have actually jumped from rats to other animals in the past.
• The "Bathtub" Virus: They found a virus in reed voles (a type of rodent) that is closely related to a virus that causes massive outbreaks in pig farms. This is a red flag for future pandemics.

Why This Matters: The "Fire Drill" Strategy

The most important takeaway is efficiency.

Imagine you have a library with 200,000 books. You need to find the ones that contain secret codes.

• Old Way: Read every single page of every book. (Takes years, costs a fortune).
• New Way: Use a metal detector (ISG Profiler) to scan the books. If the detector beeps, then you open the book and read it.
  • The new method only requires you to do the heavy reading on about 7% of the books.
  • Yet, it still finds 45% of the secret codes (viruses) that the old method missed.

The Bottom Line

This paper introduces a smarter, faster way to hunt for dangerous viruses in wildlife and livestock. Instead of waiting for a virus to look familiar, we now listen for the body's reaction to it. This helps us find "hidden" viruses before they jump to humans, acting like an early warning system for the next pandemic.

In short: They stopped looking for the thief's face and started listening for the alarm bells. And thanks to that, they found a lot of thieves who were wearing perfect masks.

Technical Summary

1. Problem Statement

The discovery of novel viruses in wildlife and livestock is critical for pandemic preparedness. Public repositories like the Sequence Read Archive (SRA) contain hundreds of thousands of RNA-seq datasets from diverse animal species. However, current virus discovery methods face three major limitations:

• Computational Intensity: Conventional de novo assembly and homology-based searches (e.g., BLAST, geNomad) are too slow to scale for the exponentially growing volume of RNA-seq data.
• Limited Sensitivity for Divergent Viruses: Homology-based methods struggle to detect highly divergent viruses or those with few conserved sequences, leading to false negatives.
• Inability to Distinguish Infection from Contamination: Detected viral sequences may arise from environmental contamination or endogenous viral elements rather than active infection. Traditional methods cannot easily distinguish true infections from artifacts.

2. Methodology

The authors developed a host-response-centric framework comprising two main tools: ISG Profiler and ISG-VIP.

A. ISG Profiler (Quantification Engine)

• Concept: Viral infections trigger the host's innate immune system, specifically the upregulation of Interferon-Stimulated Genes (ISGs). This response is evolutionarily conserved across vertebrates.
• Workflow:
  1. Reference Database: Constructed from orthologous sequences of 59 core ISGs (experimentally validated, conserved across 10 mammalian/avian species) and 100 Internal Control Genes (ICGs) (stable, non-responsive genes) derived from 398 amniote species.
  2. Mapping: RNA-seq reads are mapped to this ortholog database using Salmon (selected for highest correlation with conventional pipelines).
  3. Normalization: Raw counts are normalized by total ICG counts, log-transformed, and standardized using precomputed means/SDs.
  4. Scoring: An ISG Score is calculated as the mean of standardized counts across all core ISGs for each sample.
• Advantage: Does not require species-specific reference genomes, making it applicable to species lacking genomic resources. It processes a standard 150-bp paired-end dataset in ~3.9 minutes.

B. ISG-VIP (Prediction Model)

• Concept: A machine learning model to predict viral infection status based on ISG expression profiles and host taxonomy.
• Training Data: Used ~168,000 RNA-seq datasets ("170k dataset") with pre-assembled contigs from the Logan project. Ground truth labels were generated using geNomad (a state-of-the-art homology-based tool), focusing on "ISG-inducing" viral families.
• Architecture: A stacking ensemble model using LightGBM and Logistic Regression as base learners and Random Forest as a meta-learner.
• Features: Includes standardized counts of 59 ISGs, 100 ICGs, the ISG score, normalized read counts, and one-hot encoded host taxonomic data (species and order).
• Validation:
  • Internal: 5-fold cross-validation (grouped by BioProject to prevent data leakage).
  • External: Applied to ~40,000 new datasets ("41k dataset") released in 2024, where contigs were re-assembled and validated via BLASTx.

3. Key Contributions

1. Scalable Framework: Introduced a computationally efficient workflow that screens RNA-seq data in minutes rather than days, enabling the analysis of ~210,000 datasets.
2. Ortholog-Based Quantification: Developed a method to quantify immune responses across diverse species without needing species-specific reference genomes, overcoming a major barrier in wildlife virology.
3. Hybrid Discovery Workflow: Proposed a "prescreening" strategy where ISG-VIP filters samples, and only predicted-positive samples undergo computationally expensive assembly and homology searches.
4. Detection of "Hidden" Viruses: Demonstrated the ability to detect viral infections missed by conventional homology tools, particularly those caused by highly divergent or fragmented viral sequences.

4. Key Results

A. Performance and Validation

• Accuracy: ISG Profiler showed strong correlation with conventional differential expression analysis in IFN-stimulated cells and experimental infections.
• Predictive Power: ISG-VIP achieved an F1 score of ~0.40 in external validation, significantly outperforming simple logistic regression. It successfully identified 7.1% of the external dataset as infected, of which 37.3% were confirmed by geNomad/BLASTx (compared to only 3.4% in the predicted-negative group).
• Efficiency: The prescreening workflow reduces the number of samples requiring deep virome analysis to 7.1–8.5% while recovering 43–45% of virus-infected samples.

B. Biological Insights

• Basal ISG Levels: Basal ISG expression varies significantly by host lineage, with Primates, Artiodactyla, and Carnivora showing elevated baseline levels.
• Viral Family Effects: RNA viruses generally induce strong ISG responses, whereas many DNA viruses do not. Notable exceptions include Bornaviridae (persistent infection) and Togaviridae/Flaviviridae (strong IFN antagonism).
• False Positives: Some false positives in ISG-VIP were attributed to bacterial infections, which also induce ISGs, suggesting the tool could be repurposed for bacterial detection.

C. Novel Virus Discoveries

Using the ISG-VIP workflow, the authors identified novel viruses missed by geNomad:

1. Chaphamaparvovirus in Chickens: Identified in liver tissues of chickens and wild birds. Transcriptomic analysis revealed signatures of hepatitis (upregulation of antiviral/inflammatory pathways, downregulation of metabolic pathways), linking these novel viruses to liver disease.
2. Hepatovirus in Chickens: Detected a novel avian clade of Hepatovirus (related to Hepatitis A) in chicken livers, suggesting spillover from aquatic birds.
3. Protoparvovirus in Rats: Identified a novel Protoparvovirus in rats that phylogenetically clusters with highly pathogenic viruses like Feline Panleukopenia Virus (FPV) and Canine Parvovirus (CPV-2). Ancestral state reconstruction suggests a Muridae (mouse/rat) origin for this clade.
4. Betaarterivirus in Reed Voles: Discovered a novel Betaarterivirus (related to Porcine Reproductive and Respiratory Syndrome Virus, PRRSV) circulating at high prevalence (>75%) in reed vole populations in China, indicating a potential reservoir for future swine epidemics.

5. Significance

• Pandemic Preparedness: This framework provides a scalable, cost-effective method for global viral surveillance, capable of processing the massive influx of public RNA-seq data to identify emerging threats early.
• Overcoming Homology Limits: By leveraging the host's immune response rather than viral sequence similarity, the method detects highly divergent viruses that traditional tools miss.
• Functional Context: The approach adds a layer of biological validation (immune activation) to viral detection, helping to distinguish true infections from contamination or endogenous elements.
• Future Directions: The authors suggest expanding the framework to include other host pathways (e.g., apoptosis, stress) to detect viruses that actively suppress ISG responses, and applying it to environmental RNA (e.g., wastewater) for early outbreak detection.

In summary, this study presents a paradigm shift in virus discovery, moving from purely sequence-based homology searches to a host-response-driven approach that is faster, more scalable, and capable of uncovering "hidden" viral diversity in wildlife and livestock.