#15 · Systems biology

Contact and herd-level data test whether mechanistic transmission models are adequate

The dominant thread is whether empirical contact and transmission data can validate the structure of mechanistic models before their outputs are interpreted. An agent-based model with random-mixing contacts matches survey contacts in households, schools and workplaces but misses age-structured contacts in outside settings, a structural discrepancy relevant to model selection. A multi-type branching likelihood for contact trees is validated on simulations and COVID-19 data from Karnataka, while a particle-MCMC analysis of Salmonella Dublin in Öland cattle attributes roughly half of the force of infection to local spread and half to within-herd transmission, with Rt approximately 1. The items address distinct problem classes, so this thread is not uniform across them. What is new is the use of contact-level or herd-level validation to test model adequacy prior to inference on outputs.