hmmer1

Usage

Overview

HMMER is a suite of several programs for biological sequence alignment and database homology search. It uses probabilistic models called "profile hidden Markov models" (profile HMMs) to represent the likely evolutionary homologs of a single sequence or a multiple alignment of a sequence family. A main avenue of research is to improve the evolutionary predictive models in HMMER to be able to recognize and accurately align increasingly remote homologs, distant in time.

HMMER is also used as an organizational tool, to group the exponentially growing number of biological sequences into a vastly smaller set of well-annotated sequence families. New sequences can be annotated by comparison against curated sequence family databases of prebuilt HMMER profiles, in addition or instead of comparison to the entire sequence database. Databases such as Pfam, SMART, and TIGRfams, among others, are based on this principle.

HMMER is used in three main modes: to search a sequence database for new homologs of a sequence or a sequence family; to search a profile database (like Pfam) to find what known family a query sequence belongs to, or what domains it has; and to automatically construct large multiple alignments (i.e. with an effectively unlimited number of sequences) using a profile representative of a sequence family.

Suppose you have a multiple sequence alignment of a sequence family of interest, and you want to search a sequence database for additional homologs. The hmmbuild program builds profile(s) from multiple alignment(s). The hmmsearch program searches protein profile(s) against a protein sequence database, and nhmmer searches nucleotide profile(s) against a nucleotide sequence database.

Suppose you have a single sequence of interest, and you want to search a sequence database for additional homologs. The phmmer program searches a single protein sequence against a protein sequence database. The jackhmmer program does the same thing but iteratively -- homologs detected in a previous round are incorporated into a new profile, and the new profile is searched again. phmmer is used like BLASTP, and jackhmmer is used like a protein PSI-BLAST. The nhmmer program searches a single nucleotide sequence against a nucleotide sequence.

Suppose you have sequence(s) that you want to analyze using a HMMER-based profile HMM database like Pfam (http://pfam.sanger.ac.uk). The hmmpress program formats a profile HMM flatfile (such as the file you would download from Pfam) into a HMMER binary database. The hmmscan program searches protein sequence(s) against that database. The nhmmscan program can similarly search nucleotide sequence(s) against a pressed database of nucleotide profiles, such as from Dfam (http://dfam.janelia.org).

Suppose you want to align lots of sequences. You can construct a manageably small alignment of a representative set of sequences, build a profile with hmmbuild, and use the hmmalign program to align any number of sequences to that profile.

HMMER also includes some auxiliary tools for working with large profile databases. hmmfetch fetches one or more profiles from a database. hmmstat prints summary statistics about a profile file.

For compatibility with other profile software and previous versions of HMMER, the hmmconvert program converts profiles to a few other formats. We intend to add more support for other formats over time.

The hmmemit program generates (simulates) "homologous" sequences by sampling from a profile. It can also generate a "consensus" sequence.

The hmmsim program is a simulator used for collecting statistics about score distributions on random sequences.

Each program has its own man page.