Fig. 2.

Turnover rates of 8mers in a long, neutrally evolving simulated primate sequence. In (A), the motif-based approach finds many more gain events than losses. In (B), we show the actual numbers of turnover events that occurred globally in this neutrally evolving sequence, in which gains and losses are balanced as every 8mer lost from an ancestor is also a gain of a new 8mer in the descendant. In (C), the maximum likelihood approach to ancestral reconstruction finds similar gain events to loss events, but slightly underestimates overall gains and losses relative to actual events. Because DNAML creates a trifurcation at the root, some turnover events from the common ancestor of Human, Chimp, Gorilla, and Orangutan are misattributed to Gibbon.