A bayesian approach for parameterizing and predicting plasmid conjugation dynamics

Population dynamic models that explain and predict the spread of conjugative plasmids are pivotal for understanding microbial evolution and engineering microbiomes. However, prediction uncertainty of these models has rarely been assessed. We adopt a Bayesian approach, employing Markov Chain Monte Carlo (MCMC), to parameterize and model plasmid conjugation dynamics. This approach treats model parameters as random variables whose probability distributions are informed by data on plasmid population dynamics. These distributions allow us to estimate credible intervals of the model's parameters and predictions. We validated this approach using synthetic population dynamic data with known parameter values and experimental population dynamic data of mini-RK2, a miniaturized counterpart of the well-characterized and widely used RK2 conjugation plasmid. Our methodology accurately estimated the parameters of synthetic data, and model predictions were robust across time scales and initial conditions. Incorporating long-term population dynamic data enhances the precision of parameter estimates related to plasmid loss and the accuracy of long-term population dynamic predictions. For experimental data, the model correctly explained and predicted most population dynamic trends, albeit with broader credible intervals. Incorporating long-term data also improves credible ranges of most parameters. However, in some cases, such as with the growth parameter of cells with the conjugative plasmid, the inclusion of long-term data can lead to stronger correlations and potential identifiability issues between key parameters. Overall, our method allows for deeper investigation of plasmid population dynamics and could potentially be generalized to study population dynamics of other mobile genetic elements.