A computational investigation led by Nefel Tellioglu, David J. Price, and colleagues at the University of Melbourne has introduced an individual-based modeling framework designed to track transmission, colonisation, and clinical outcomes across individual serotypes of Streptococcus pneumoniae. The bacterium represents a multi-serotype pathogen responsible for a substantial global burden of pneumonia and invasive bacterial disease such as meningitis and bacteraemia. While pneumococcal conjugate vaccines have been implemented worldwide to target specific high-burden capsular serotypes, long-term effectiveness in many settings has been complicated by serotype replacement, where non-vaccine types expand into newly cleared ecological niches. Higher-valency vaccines have been formulated to address this challenge, but often demonstrate varying immunogenicity across individual serotypes. The authors constructed an integrated simulation system capable of capturing serotype-specific antibody responses and competitive colonization dynamics without relying on aggregate serotype groupings.
The biological complexity of Streptococcus pneumoniae presents unique challenges for public health evaluation because the bacterium possesses more than one hundred distinct serotypes distinguished by their polysaccharide capsules. These capsular antigens largely determine bacterial pathogenicity, invasiveness, and the concentration of serotype-specific immunoglobulin G required to confer protection. Previous conjugate vaccines successfully induced immune memory and interrupted nasopharyngeal transmission by conjugating capsular polysaccharides to carrier proteins, but earlier mathematical models predominantly aggregated serotypes into coarse categories, such as vaccine types versus non-vaccine types. Such aggregate formulations typically assumed that an individual could not experience co-infection with multiple non-vaccine strains simultaneously, an assumption that risks distorting transmission dynamics once vaccination suppresses vaccine-targeted strains. Furthermore, newer vaccine formulations, including the 21-valent conjugate candidate V116, are not strictly nested supersets of earlier formulations, making legacy hierarchical groupings mathematically obsolete.
To overcome these limitations, the research team designed a modular simulation architecture comprising three coupled components: an immunity model, a transmission model, and a clinical model. The immunity component computes individual antibody levels based on vaccine product history, administered dose count, and elapsed time since vaccination, allowing antibody titers to decay over time according to empirical distributions. When an individual encounters an infected contact within the transmission module, their circulating antibody concentration alters their probability of acquiring carriage according to a parameterized logistic curve. If colonization is successfully established, the individual antibody concentration subsequently determines the probability of progression from asymptomatic carriage to symptomatic infection, including invasive pneumococcal disease or community-acquired pneumonia, using a separate logistic response function.
The transmission component operates as an individual-based susceptible-infectious-susceptible demographic framework simulating carriage across age cohorts. Demography within the model accounts for annual birth rates, aging, empirical age-specific death rates, and immigration dynamics derived from national population statistics. Mixing patterns between and within specific age groups are governed by synthetic and empirical contact matrices. Within this structure, individuals can simultaneously carry up to two distinct pneumococcal serotypes, explicitly incorporating biological within-host competition by reducing the susceptibility of an already colonised individual to acquiring a secondary serotype by a calibrated factor. External introduction of serotypes is maintained through simulated migration and uniform sampling across potential serotypes, allowing novel or rare variants to emerge and compete under ongoing vaccine-induced immune selection.
Within the immunity module, serotype-specific initial antibody responses were informed by a rapid review of published clinical studies evaluating geometric mean concentrations of immunoglobulin G. The investigators categorized serotypes into low, moderate, and high immunogenicity tiers across various vaccine products and dosing schedules, explicitly capturing documented variations such as the consistently lower post-vaccination antibody titers observed against Serotype 3. The framework adopts the assumption that vaccine-elicited antibodies are specific to the serotypes included in each administered product, reflecting limited and inconsistent clinical evidence regarding cross-reactive protection against non-vaccine types. Additionally, because robust empirical data linking natural infection clearance to long-term protective antibody boosts remain lacking, individuals in the model return to a fully susceptible state upon clearing colonization, while age-dependent reductions in carriage duration indirectly capture acquired biological resistance across the lifespan.
The clinical component links individuals harboring active infections to downstream health outcomes based on age, vaccination history, and the specific serotypes carried. The probability of developing disease considers baseline age-specific rates modulated by serotype-specific invasive disease potential and remaining circulating antibody concentrations. The model differentiates between invasive pneumococcal disease and community-acquired pneumonia, reflecting empirical distributions observed in clinical surveillance data. In accordance with established immunological findings, the authors assumed that 23-valent pneumococcal polysaccharide vaccines do not confer protective immunity against mucosal community-acquired pneumonia, restricting protection against non-invasive pneumonia exclusively to conjugate vaccine platforms that elicit mucosal and humoral immune memory.
To demonstrate practical utility for public health evaluation, the authors calibrated the modeling platform to historical epidemiological surveillance from Australia spanning 2002 to 2018. During this historical window, Australia implemented a funded infant immunization schedule using 7-valent pneumococcal conjugate vaccine in 2005 at two, four, and six months of age, alongside a 23-valent polysaccharide program for adults aged 65 and older. In 2011, Australia transitioned the pediatric program to 13-valent conjugate vaccine with an accompanying catch-up initiative for children under four years old. Model parameters were adjusted using a two-stage heuristic calibration strategy designed to match pre-vaccination carriage prevalence data from young children while capturing subsequent observed shifts in age-stratified invasive disease presentations and serotype distributions documented by the National Centre for Immunisation Research and Surveillance.
The calibrated simulation accurately reproduced observed historical trajectories in the non-Indigenous Australian population, matching annual age-stratified invasive pneumococcal disease incidence and the shifting proportions of circulating vaccine serotypes over time. Following the introduction of the 7-valent vaccine in 2005, modelled carriage of the seven target serotypes dropped sharply while non-vaccine strains expanded. When the 13-valent vaccine was deployed in 2011, the additional targeted serotypes similarly declined, accompanied by continued expansion of non-vaccine serotypes and those covered only by the adult 23-valent polysaccharide vaccine. For community-acquired pneumonia, modelled incidence aligned well with observed clinical trends in younger pediatric cohorts, though alignment was less precise in older adult populations where historical empirical data relied on diagnostic approximations pooling specific and unspecified pneumonia diagnostic codes.
To evaluate the specific health impact of the 2011 policy transition, the authors conducted a counterfactual analysis comparing the actual rollout of the 13-valent vaccine against a hypothetical scenario in which Australia maintained the 7-valent vaccine continuously through 2018. Across 10,000 paired simulation iterations, the 13-valent vaccine program resulted in a lower average annual burden after 2011, yielding an annual incidence of 6.99 cases of invasive pneumococcal disease per 100,000 people (95% confidence interval: 6.38 to 7.64) compared to 7.40 cases per 100,000 (95% confidence interval: 6.91 to 7.90) under the continued 7-valent scenario. Similarly, annual community-acquired pneumonia incidence with the 13-valent vaccine was 79.79 per 100,000 (95% confidence interval: 77.53 to 81.85) versus 83.96 per 100,000 (95% confidence interval: 81.90 to 85.80) without it. These model estimates indicate that adopting the 13-valent pediatric schedule reduced overall population burdens of invasive disease and pneumonia by 5.5% and 5.0%, respectively.
Sensitivity analyses demonstrated that the qualitative benefits of the 13-valent switch remained robust when modifying key structural parameters, including doubling the assumed antibody half-life or altering susceptibility limits governing dual colonization. Doubling antibody durability increased the simulated rate of serotype replacement, underscoring the influence of antibody waning assumptions on long-term ecological competition. The investigators highlighted several inherent study limitations, noting that Australian carriage data were largely restricted to the pre-vaccine era, precluding separate empirical identification of protection against carriage acquisition versus protection against disease progression given infection. Consequently, calibrated parameters represent an operationally valid configuration reproducing historical disease patterns rather than unique biological constants. The authors concluded that while individual-level multi-serotype models require substantial data inputs, their architectural flexibility offers a necessary foundation for evaluating emerging non-nested conjugate formulations in diverse global populations.
Subject of Research: Agent-based computational modelling of multi-serotype Streptococcus pneumoniae carriage and vaccine impact evaluation
Article Title: A flexible agent-based modelling framework of multi-serotype pneumococcal carriage to evaluate vaccine strategies in large populations
Article References: Tellioglu, N., Price, D. J., Chen, X., Spirkoska, V., Wang, C. Y., Moss, R., Carvalho, N., Carville, K. S., Campbell, P. T., & McVernon, J. (2026). A flexible agent-based modelling framework of multi-serotype pneumococcal carriage to evaluate vaccine strategies in large populations. PLOS Computational Biology, 22(11), e1014946. https://doi.org/10.1371/journal.pcbi.1014946
Image Credits: AI Generated
DOI: 10.1371/journal.pcbi.1014946
Keywords: Streptococcus pneumoniae, pneumococcal conjugate vaccines, agent-based modelling, serotype replacement, computational biology, carriage transmission, invasive pneumococcal disease, community-acquired pneumonia, 7vPCV, 13vPCV, 23vPPV, antibody kinetics, transmission dynamics, infectious disease modelling, epidemiology, Australian immunization
News Source: Kristina Jarvis. (November 16, 2026). Agent-Based Model Tracks Multi-Serotype Pneumococcal Carriage and Vaccine Strategy Impacts. Scienmag.



