Figure 2From: Efficient context-dependent model building based on clustering posterior distributions for non-coding sequencesStepwise likelihood-based clustering of the Ancestral Repeats data. The stepwise clustering of contexts using the likelihood-based clustering approach shows, from top to bottom, the subsequent merges of contexts for the Ancestral Repeats dataset. The starting point is a full (96-parameter) context-dependent model, shown in white. The optimal model has 10 clusters and 60 parameters.Back to article page