Fig. 2.

Total number of interactions detected with the association screening method (A and B) and fraction of these interactions that are consistent with model predictions (C and D). This analysis is ran for a model with two host types (A and C) or a single host type (B and D). The blue lines show the result of a linear model fit (A and B) and generalised linear model fit assuming a Poisson distribution of the outcome variable (C and D). Grey areas are prediction confidence intervals based on the standard error of the fit. These methods do not correct for multiple hypothesis testing, which could lower the number of interactions detected. In panels A and C, h = 1 and a = 0. We assume that there are N = 5000 hosts in the population.