Python Programming for Biology: Bioinformatics and Beyond

Figure 21.12.  A schematic of a simple two-state HMM for buried and exposed

protein states. For a model of two hidden states (‘b’ is buried and ‘e’ is exposed) we have

a transition matrix corresponding to four possible state transitions and an emission matrix

specifying the likelihood of each of the 20 amino acids being found in each of the two

states.

---

Figure 21.13.  A schematic of a 40-state HMM for buried and exposed protein states.

In this model there is a separate hidden state for each amino acid and buried/exposed

combination (i.e. 2×20 states). This gives rise to a large, 1600-element transition matrix

between all the 40 possibilities. The emission matrix in this case, however, is simple,

given that an observed amino acid can only come from the two hidden states which

involve that specific amino acid, and the non-matching emission probabilities will be zero.