Notice- HMMer 3 is now in beta testing!

Sean Eddy has announced that HMMer 3 is now in beta testing. Log-likelihood scores will be used which may yield BETTER results, not just faster! I am thinking of doing a port to windows and/or Mac. Any preferences? Let me know at

Whether or not this winds up hurting accelerator sales depends on whether or not researchers accept the new format. One thing is sure, this will increase the use of HMMs in Biotech and pharmaceutical companies worldwide, not to mention the plant protein labs. The HMMer website is at and the discussion board is at

Also, the name HMMer is now limited to the open source code, by copyright law. This will not affect variations such as DeCypherHMM or SPSPfam, but will require a name change for some other programs. Let me know what you think of this policy, by writing to me at

GPU-HMMer is out now, and it looks good. So why aren't the HMM accelerator vendors worried? Because they have only accelerated the hmmsearch code, not hmmpfam. Since most big searches use hmmpfam, you still need an accelerator.

In other news, CLCbio is getting great speedups with their SIMD version of HMMer. The JCVI has bought into it for their pipelines, and report faster results than what they saw with LDHMMer.

Biomatters has no SIMD accelerated software in their GENEious package, but rather formed a partnership with TimeLogic to provide high-end services for those who need them.

HMMs in secondary structure prediction-

If you want to predict 3D structure, it is best to start with a secondary structure prediction. HMMs can help! Check out this paper by Martin et. al.

HMM introduction-

Have you ever wondered what is meant by the term hidden Markov model or 'HMM'? Perhaps you have heard that HMM searches are in some way 'better' or more sensitive or accurate that the usual BLAST search. 

Unfortunately, finding information about why you would want to run an HMM search is not trivial. The standard bioinformatics books do not get into enough detail to understand how they really work. The older textbooks on HMMs were too mathematically oriented for the average biologist or biochemist. The overwhelming number of equations and lack of specific information about applications meant that these books left a lot of people dissatified, and the textbooks became expensive paperweights.

The Handbook of Hidden Markov Models in Bioinformatics was written to inform the life science researcher or student about what HMMs are, how to use them, and which implementation to use.

Demonstrating that many useful resources, such as databases, can benefit most bioinformatics projects, the Handbook of Hidden Markov Models in Bioinformatics focuses on how to choose and use various methods and programs available for hidden Markov models (HMMs).

The book begins with discussions on key HMM and related profile methods, including the HMMER package, the sequence analysis method (SAM), and the PSI-BLAST algorithm. It then provides detailed information about various types of publicly available HMM databases, such as Pfam, PANTHER, COG, and metaSHARK. After outlining ways to develop and use an automated bioinformatics workflow, the author describes how to make custom HMM databases using HMMER, SAM, and PSI-BLAST. He also helps you select the right program to speed up searches. The final chapter explores several applications of HMM methods, including predictions of subcellular localization, posttranslational modification, and binding site.

By learning how to effectively use the databases and methods presented in this handbook, you will be able to efficiently identify features of biological interest in your data.

Buy it today at

Martin Gollery is a Bioinformaticist, businessman and an award-winning trainer. Would you like help with training, presentations, grants, presentations or next generation sequence analysis? Write to me at marty.gollery at

For more information about what I can do for your project, contact me or see my website at

SuperFamily Notice!

SuperFamily release 1.73 is now available.


-the coiled coil class of protein domains has been added to SUPERFAMILY.

-the model library has been updated to the latest SCOP version 1.73.

-there are now over 1000 genomes; old genomes have been updated and new ones added.

-among the new features, there is a browseable species tree of all 1000 organisms.


-the website has been re-organised and we hope you find the interface improved.

-we now have a twitter feed. More information is published in Wilson, D. et al. (2009) SUPERFAMILY

-sophisticated comparative genomics, data mining, visualization and phylogeny. Nucl. Acids Res. 37, D380-D386.

For more information