Programs

There are currently nine programs included in the package:

hmmt

Build an HMM from initially unaligned training sequences. Various options allow you to provide a starting hint at the correct alignment, or to fix part of the alignment (say, if you already know a structure-based alignment of some of your sequences).

hmmb

Build an HMM from a multiple alignment. hmmb includes options for building various sorts of models. These include ``maximum discrimination'' or weighted models, which compensate for overrepresentation of some sequences in the data set.

hmma

Calculate a multiple sequence alignment using an existing HMM.

hmmls

Search a sequence database for local matches to an HMM. The full HMM must match part of the sequence.

hmms

Search a sequence database for global matches to an HMM. The full HMM must match the full sequence.

hmmsw

Smith/Waterman searching of a sequence database with an HMM, looking for local matches of part of the sequence against part of the HMM. This is perhaps the most useful and sensitive program for protein database searching.

hmmfs

Search for multiple fragments in long sequences (repeats or motifs); an optimal non-overlapping set of local matches between the HMM and the sequence are reported. This program was specially designed for finding and annotating dispersed repetitive DNA sequences such as the human Alu family. It is also useful for parsing the domain structure of proteins with tandemly repeated domains (such as the immunoglobulin superfamily).

hmme

Generate likely sequences stochastically from an HMM. This is rarely used, but sometimes useful for seeing what an HMM thinks it is modeling.

hmm-convert

Convert HMM files to/from ASCII and binary formats. Normally, HMM files are saved in a compact but unreadable (by you, anyway) format. This program lets you convert HMM files to a readable ASCII format, and back again. You might use this to look at HMMs yourself, or to use your own programs to construct HMMs and use the hmms/hmmsw/hmmls/hmmfs search programs.