Modeling Immune Variability: Uncertainty Quantification, Identifiability, and Digital Twins in Systems Immunology

Immune responses vary at every scale. Genes are expressed stochastically, cells make probabilistic fate decisions, pathogen loads fluctuate, measurements are noisy, and patients respond differently. In immunology, this variability is part of the signal. Many systems immunology workflows still depend on deterministic models and single best-fit parameter values. These can hide ambiguity, weaken reproducibility, and make it harder to apply findings in clinical or other real-world settings.This Research Topic brings together work that treats uncertainty as a core result of immune modeling. Our aim is to help researchers show what their models can support, what remains unresolved, and what new data would settle it. We also want to encourage shared, practical standards for reporting uncertainty.We welcome studies that apply:- Local and global sensitivity analysis.- Structural and practical identifiability analysis.- Forward uncertainty propagation.- Bayesian inverse uncertainty quantification, such as MCMC, Sequential Monte Carlo, and variational inference.We encourage clear reporting of priors, posterior distributions, convergence diagnostics, model discrepancy assumptions, emulator validation, and reproducible workflows.High-dimensional immune models are costly to run. We therefore particularly welcome work on surrogate models and statistical emulators, such as Gaussian processes, polynomial chaos expansions, and neural network surrogates. We are interested in how they are built, validated, and used for calibration, sensitivity analysis, and experimental design.We also invite work on digital twins of immune systems and patient-specific immune responses. These integrate mechanistic models, data assimilation, Bayesian updating, and uncertainty quantification to support prediction, personalization, and decisions under uncertainty.A central question runs through the collection: when can a model support a mechanistic claim, and when should uncertainty lead to more cautious conclusions? Submissions may cover any scale, from molecules to populations. They may use any data type, including single-cell multi-omics, imaging, and longitudinal cohorts. They may use any modeling approach, including ODE and SDE models, agent-based models, Bayesian hierarchical models, mechanistic–ML hybrids, emulator-augmented frameworks, and digital twin architectures.Themes include, but are not limited to:- Stochastic immune dynamics (SDEs, CTMCs, branching processes, agent-based models)- Noise-aware analysis of single-cell and spatial immunology data (technical versus biological variance)- Bayesian inference, probabilistic programming, likelihood-free inference, and posterior predictive checks- Structural and practical identifiability, sloppiness, and profile likelihood methods- Forward and inverse uncertainty quantification and global sensitivity analysis- Model discrepancy and model-form uncertainty- Digital twins for immune dynamics and personalized immunology- Applications to infection, vaccination, inflammation, autoimmunity, cancer immunology, and immunotherapy response Topic editors: Barbara De Melo Quintela, Ruy Freitas Reis Source: https://www.frontiersin.org/research-topics/87124/modeling-immune-variability-uncertainty-quantification-identifiability-and-digital-twins-in-systems-immunology