Hidden Markov Model-based Algorithms

Why do we need HMM in multiple alignment? Think about it in the following way. When you are presented with a family of proteins, you know from their alignment that some parts of the protein are more conserved than the other. Regular search programs use fixed scores for match and mismatch, and cannot utilize the information available from the highly conserved and less conserved parts of the protein. Hidden Markov model based methods can take care of such factors by adjusting the scores along various parts of the reference.