HMM2 plugin: A visual solution for using HMM profiles

Back to list of plugins

HMM2 plugin

HMM2 plugin is a toolkit based on Sean Eddy's HMMER 2 package.

While working on this plugin we were guided by the following principles:

Make HMM tools accessible to a wider user audience by providing graphical interface for all supported utilities for most of the platforms.
Be compatible with the original HMMER 2 package.
Create the high-performance solution utilizing modern multicore processors and SIMD instructions.

Current version of Unipro UGENE provides user interface for three HMM tools: hmmbuild, hmmcalibrate and hmmsearch.

To access these tools select "Tools->HMM2" submenu in the main program menu.

We highly recommend reading the original HMMER 2 documentation to learn how to use utilities provided by the plugin.

Building HMM model (HMM build):

'HMM build' tool is used to build new profile HMM from a multiple alignment.
You can use any alignment file formats supported by UGENE.

The output HMM profile format is compatible with HMMER 2 package.

Note: 'HMM build' tool does not automatically calibrate a profile. Use 'HMM calibrate' dialog to calibrate the profile.

Calibrating HMM model (HMM calibrate):

'HMM calibrate' tool reads a HMM profile file, scores a large number of synthesized random sequences with it, fits an extreme value distribution (EVD) to the histogram of those scores, and re-saves hmmfile now including the EVD parameters.

To avoid modification of the original HMM file you can select a new location for the calibrated profile.

Searching a sequence using HMM profile (HMM search):

'HMM search' tool reads a HMM profile from file and searches sequence for significantly similar sequence matches.

The sequence must be selected in 'Project' view or there must be active sequence view window opened.

If the sequence selected is nucleic and profile HMM is built for amino alignment, the sequence is automatically translated and searched in all possible frames (6 total).

If a profile HMM is built for nucleic alignment, the search is performed for both direct and complement strands.

The search results are stored as sequence annotations in Genbank file format.

Limitations: all of the UGENE tools work only with files that contain a single HMM model. The complete support for search with the PFAM databases will be available in one of the future Unipro UGENE releases.

Integrated optimizations:

Users of IBM's Cell Broadband Engine can increase HMM2 speed by about 100 times when using parallelized version of hmmsearch utility.
The optimization works out of the box in our prebuilt package for Cell Broadband Engine.
UGENE port by Ivan Efremov. Novosibirk State University.
SSE2 algorithm implementation by Leonid Konyaev, Novosibirsk State University.
Use of the SSE2 optimized version of hmmsearch algorithm with quad-core CPU gives >30x performance boost when compared with the original single-threaded algorithm (single sequence mode)