Figure 4.

Left: Information versus sequence type for consensus sequence, profiles, and IB without priors. Right: I(Y ; X) - I(Y ; C) as a function of w for different groups of priors. The information loss for 52 categories without priors is 0.359, for 10, 0.474.