Model-based estimation of superinfection prevalence from limited datasets

Area of Science:

• Infectious Diseases
• Epidemiology
• Computational Biology

Background:

• Viral superinfection, reinfection with different strains of the same virus, challenges immunologic memory.
• Accurate prevalence estimation is vital for vaccine development and understanding disease dynamics.
• Existing methods may misrepresent superinfection prevalence due to sampling limitations and strain detection difficulties.

Purpose of the Study:

• To develop a novel ecological model for inferring true viral superinfection prevalence from limited clinical data.
• To improve the accuracy and precision of superinfection prevalence estimates compared to standard methods.
• To provide insights into optimal study designs for superinfection research.

Main Methods:

• Developed an ecological model defining infected individuals' richness (number of strains) and evenness (relative strain abundances).
• Employed an Expectation-Maximization (EM) methodology to infer superinfection prevalence from sparse datasets.
• Conducted simulation studies to compare the EM method against a naive calculation under various conditions (richness, evenness, sampling).

Main Results:

• The EM method significantly outperformed the naive calculation in estimating superinfection prevalence across all tested scenarios.
• The EM method showed particular advantages when sampling was low, and strain richness or unevenness was high.
• Simulation results indicated that increasing participants or samples per participant equally improved prevalence estimates.

Conclusions:

• The developed EM-based ecological model provides a more accurate approach to estimating viral superinfection prevalence.
• This method is particularly valuable for analyzing limited or sparse clinical datasets.
• Findings suggest that optimal study designs involve a balance between participant numbers and sampling depth.