Figure 3

Model. The model underlying USF, illustrated by a toy example. The example uses an alignment and a query sequence of length 4. The query sequence is composed of the nucleotide sequence GTAA. The top row and bottom row of states each constitute a pHMM K, the middle one pHMM U. The top pHMM K models the situation of pHMM U not having been visited yet, the bottom one that of pHMM U having been visited already. Above and below, respectively, the states, their emission probabilities are given, with the nucleotide in the query sequence being marked red for the states in the Viterbi path. To the very left, resp. the very right, the initial, resp. the final state are situated. The short-dashed arrows represent transitions with probability p₂, the long-dashed ones transitions with probability p₁. The dotted arrows constitute transitions from and to special states (initial and final state). The Viterbi path is colored in red, with the first two positions and the last position of the query sequence being classified as 'known' and the third position as 'unknown'. Notice that the first state of the bottom pHMM K is missing since this pHMM can only be entered if pHMM U has been visited before.