Our latest manuscript at Roy Soc Open Science is inspired by a central question in viral ecology: who infects whom in phage-bacteria communities?
Quantifying who infects whom in a community is essential to understand how the individual-based traits affect ecosystem functions in complex environments. Cross-infection between phage and bacteria has traditionally been determined using culture-based methods, e.g., the plaque assay. Culture-based methods do not capture the entirety of the community from environmental samples because most of phage and bacteria are not available in culture. Additionally, these methods do not scale up to larger communities because they require pairwise experiments to resolve interactions.
The problem of identifying interaction networks is not unique to phage-bacteria systems. Recent advances in the estimation of competitive microbial networks show an alternative way of inferring ecological networks by exploiting information embedded in time-series when linked to nonlinear dynamic models (Monier et al. Appl Environ Microb, 2008; Stein et al. PLoS Comput Biol, 2013).
Here, we extend the method used by Stein et al. (2013) to phage-bacteria systems. This alternative perspective has the important advantage that it can be used to estimate the complete infection network without the need of pairwise experiments. The key insight is that the time series are judged by their compatibility with a model, which includes an infection network to be estimated.
We use in silico experiments to show that the approach can be used to successfully reconstruct realistic infection networks. We improve the original method by combining measurements from multiple experiments, resulting in better reconstruction of the interactions while mitigating against complications induced by evolution amongst the interacting strains.
We are hopeful that the current method can help enable in situ tests to help determine functionally relevant papers in the environment.