I implemented a Profile Hidden Markov Model (PHMM) in Python, which is used for classification of protein or gene sequences to families.

The model takes a either a multiple sequence alignment as an input or a set of unaligned sequences.

• Known parameters (multiple alignment was given) The model uses the Viterbi algorithm for training.

• Unknown parameters When the parameters are unknown, the model first estimates them using the test data with the Baum-Welch algorithm. Then, it aligns the sequences to a multiple alignment.

The model can estimate the observation probability of a sequence given a multiple alignment.