Bayesian mixture analysis for metagenomic community profiling.

Sofia Morfopoulou; Vincent Plagnol

If the step is accepted, then the chain moves to the new proposed state M l . Otherwise if not accepted, the chain's current state becomes the previous state of the chain, i.e the set of species remains unchanged. metaMix outputs log-likelihood traceplots so that the user can visually inspect the mixing and the convergence of the chain. The default setting is to discard the first 20% of the iterations as burn-in. We concentrate on the rest to study the distribution over the model choices and perform model averaging. We can then summarize appropriately the posterior distribution and answer the important questions of interest. Examples of such questions include: what species have probability p or greater being included in the set of present species? what is the probability of having the n specific closely related strains in the set of present species? Depending on the biological context, one may ask numerous similar or other case-specific questions.