Hidden Markov Model --- search for remote relatives Unlike Blast, where a single query sequence is used to search a database for related sequences, hidden Markov Model (HMM) search uses a family of related sequences as query. A multiple sequence alignment is first built from the input sequences, and then a mathematical model (HMM) or profile of the multiple alignment is built. This model or profile will be used to query a sequence database to find additional family members who are more distant relatives of the original family. HMM is behind the protein domain database Pfam. SAM HMM for remote protein homolog search and secondary structure prediction is available at UC Santa Cruz. A stand-only version of HMM program HMMER 2.2 can be down loaded from http://hmmer.wustl.edu/.