The accuracy of event prediction strategies when inferring the support of events in MPRs from sets of NPRs.

This table shows the average symmetric distance () between predicted and true event sets when computing event supports from their frequency in optimal and near-optimal reconciliations, and filtering the events with (Column 3), where T is the filtering threshold. The degree of non-optimality in reconciliations is indirectly measured by the noise level Δ introduced in the event elementary costs. Column 4 reports the average number of predicted , , and events, depending on both the filtering threshold and noise level. On average, the true evolutionary history of a gene family contains 13.7 such events.