Similarity-based clustering of sequences using Hidden Markov Models