#17 · Systems biology

Bayesian inference of dynamic models compared across structures and data regimes

The common thread is model-based Bayesian inference of dynamic processes from limited or indirect observations, with emphasis on comparison across model structures and data regimes rather than fitting a single model. OM-NPE extends masked amortised neural posterior estimation to heterogeneous epidemiological surveillance data, applied to Zika SEIR and Xylella models. A preprint proposes an ordering-matrix method to compare inferred evolutionary accumulation dynamics across algorithms, datasets, and covariates. A published numerical scheme computes survival probabilities for first-order birth-death processes on arbitrary directed graphs, addressing clonal expansion dynamics. What is new is the explicit framing of comparison across model structures and data regimes. The items span distinct domains, so this reflects the dominant pattern rather than a method shared by every item.