Figure 2

HMM structures. States are depicted by circles and transitions by arrows. The thickness of the arrows indicates the probability of the transition. The initial transition probabilities used for starting the estimation algorithm are 0.995 for thick arrows, the remaining 0.005 being equally divided between the thin arrows. Three different HMM structures were used. (a) 3 coding states, 1 non-coding state (b) 3 coding states, 2 non-coding states (intronic and non-intronic) (c) 3 coding states, 3 non-coding states. In (b) a path in the HMM starts and ends in the T state ; in (a) and (c) a path can begin and end in any state.