TL;DR
This paper explores how human genetic diversity, specifically in HLA class-I alleles, influences the spread and severity of H1N1 influenza epidemics. By modeling individual immune responses based on genetic profiles, they demonstrate that populations with a wider range of susceptibilities, rather than a uniform response, are better protected against widespread outbreaks. This work highlights the crucial role of genetic heterogeneity in shaping infectious disease dynamics.
What Problem Does This Paper Address?
Epidemiologists often model infectious diseases using simplified assumptions about host populations, treating individuals as largely identical in their susceptibility and immune response. However, it's clear that real-world populations exhibit significant genetic variation, which can profoundly impact how diseases spread. The 2009 H1N1 pandemic, with its varied impact across different regions and populations, underscored the need to account for these individual differences. This paper tackles the challenge of incorporating population-level genetic heterogeneity, specifically focusing on human leukocyte antigen (HLA) class-I alleles, into epidemiological models to better understand and predict epidemic severity.
Methods
The authors employ a multi-compartment SIR (Susceptible-Infected-Recovered) model. The core innovation lies in how they define the "susceptible" compartment. Instead of a single group, they divide the population into multiple sub-populations based on predicted immune response. This prediction is driven by:
- HLA Genotype Generation: Using publicly available HLA allele frequencies for various ethnicities, they generate plausible HLA class-I genotypes for individuals.
- Epitope Prediction: For each genotype and a given H1N1 viral strain, they use algorithms (from the IEDB resource) to predict the viral epitopes that can be presented by the individual's HLA molecules.
- Susceptibility Sub-population Formation: Individuals are clustered into sub-populations based on the number of high-affinity viral epitopes they can present. The hypothesis is that a higher number of presented epitopes leads to a lower individual susceptibility to infection.
- SIR Model Integration: These susceptibility sub-populations are then integrated into a standard SIR model, with each sub-population having its own infection rate (βᵢ) that depends on the average number of presented epitopes (eᵢ) and the individual's susceptibility (sᵢ).
Datasets
- HLA Allele Frequency Data: Obtained from the Allele Frequency Net Database (AFND) for 61 ethnicities.
- H1N1 Viral Strain Genome Sequences: Sourced from UniProt for 81 strains isolated in 2009 and 85 strains from other years.
Key Insights and Ideas
- Genetic Diversity as a Shield: A population with a broad distribution of immune responses (i.e., many sub-populations with widely varying susceptibilities) is more resilient to influenza spread than a population with a more uniform response, even if the average susceptibility is the same.
- Skewed Susceptibility is Protective: Populations with a small number of highly susceptible individuals coexisting with a large number of less susceptible individuals tend to have smaller outbreaks. This "positive skewness" in the susceptibility distribution acts as a protective barrier.
- Beyond Average Susceptibility: While the average susceptibility (β) and basic reproduction number (R₀) are important, they don't fully predict epidemic size. The detailed distribution of susceptibilities (the Susceptibility Profile Vector, SPV) and its characteristics (like skewness and coefficient of variation) are crucial.
- Allele-Specific vs. General Risk: An HLA allele associated with high risk for one influenza strain doesn't necessarily confer high risk for all strains. The specific interaction between HLA alleles and viral epitopes is key.
Key Results
- The model qualitatively reproduces observed trends in H1N1 influenza spread worldwide.
- Populations with heterogeneous HLA profiles showed smaller epidemic sizes compared to homogeneous populations with similar average susceptibility.
- Positive skewness in the susceptibility distribution (Sk(SPV) > 0) was strongly correlated with smaller epidemic sizes (FI∞).
- Higher genetic diversity, as measured by the coefficient of variation (CV(SPV)) and standard deviation (σ(SPV)) of susceptibility, also correlated with smaller epidemic sizes.
- The model's predictions for epidemic sizes in different countries (China, Russia, Japan) generally aligned with observed trends, suggesting the approach captures essential dynamics.
Limitations and Future Work
- Excluded Factors: The model does not include age, prior infection history, vaccination status, or spatial heterogeneity. These are acknowledged limitations, and incorporating them would require more complex data and modeling.
- Simplified Immune Response: The model simplifies the immune response by pooling all non-HLA-related factors into a single proportionality constant. Future work could explore more detailed immunological pathways.
- Epitope Prediction Accuracy: The accuracy of epitope prediction algorithms can be lower for rare alleles. Exploring improved prediction methods is suggested.
- Quantitative vs. Qualitative Predictions: Due to the limitations, the model is best suited for qualitative comparisons between epidemic pairs rather than absolute quantitative predictions.
- Extensions: Future work could explore alternative mathematical forms for the susceptibility-epitope relationship, incorporate spatial heterogeneity (e.g., family/household structures), or use agent-based models for more granular simulations.
Summary updated