The hidden Markov model. The hidden Markov model. Combined graph for subcellular location classes (upper layer) and functional classes (lower layer). Joint predictions of these two entities means finding (multiple) paths through the graph from begin to end. The nodes in the two layers encode the values of the hidden random variables location class and functional class, see also Figure 2. The arrows between the nodes encode "transition" probabilities which are estimated by frequency counts on a training set. For highlighting the main structure of this graph, only transition probabilities with p > 0.1 are shown. Width and color of the arrows encode these probabilities: > 0.8 yellow, > 0.6 blue, > 0.4 green, > 0.2 red. For instance, the yellow arrow between the nodes "745" and "07" means that more than 80% of the proteins with subcellular localization transport vesicles belong to the functional class cellular transport & transport mechanism.