Towards identifying lateral gene transfer events.

This paper is concerned with evaluating the performance of the model and algorithm in 5 for detecting lateral gene transfers events. Using a Poisson process to describe arrival times of transfer events, a simulation is used to generate "synthetic" gene and species trees. An implementation of an efficient algorithm in 5 is used to estimate the minimum number of transfers necessary to explain disagreements between the generated gene and species trees. Our first result suggests that the algorithm can solve realistic size instances of the problem. Our second result suggests that the mean error and variance are low when saturation does not occur. Additionally, certain plausible evolutionary events allowed by our model of evolution used to generate gene and species trees but not detectable by the algorithm occur rarely implying the framework should work well in practice. Our third, surprising result suggests that the number of optimal scenarios is on average low for realistic input sizes.