Comments on: HMMER, reloaded

By: Martin Gollery

Martin Gollery — Mon, 05 Jan 2009 15:30:53 +0000

Will the input and output formats be identical to v2.3.2, or will we have to rewrite the parsers?

By: Ryan Richt

Ryan Richt — Sun, 04 Jan 2009 03:32:20 +0000

Woohoo! Obviously alluded to in the last super-awesome Eddy paper about forward scores, but the insane performance optimizations are quite a surprise!

Will HMMER3 use multi-core parallelization, OpenCL/CUDA parallelization, the PVM network parallelization or some combination?

For next-gen sequencing applications we are still at the point where even WU-BLAST (now AB-BLAST http://www.advbiocomp.com/) is orders of magnitude too slow for aligning the 12 giga-bases of sequence from a single paired-end Illumina or SOLiD run.

Are there any applications that would benefit from aligning each read separately using HMMER3 instead of some consensus version? Maybe ChIP-seq where binding motifs are what we are searching for? Or allele specific expression of RNA-seq where a consensus masks important information in diploid and polyploid genomes? Could HMMER ever be fast enough to align 200 million 100bp reads (one Illumina PE run a few months from now) in a reasonable amount of time on a modest number of processors to Pfam for “meta-genomic community as a bag-of-genes” type work? HMMER4?

By: Sean Eddy

Sean Eddy — Fri, 19 Dec 2008 16:30:19 +0000

It’s going to show alignment uncertainty on all the alignments it produces. That’s deeply intrinsic to how it’s doing the inference now. Implementation wise, the first useful applications will only be for protein (DNA is going to require more work; hopefully DNA applications will appear by late 2009), either for profile alignments or pairwise alignments. True de novo multiple alignment of initially unaligned sequences (as opposed to aligning a bunch of sequences to a given profile) may or may not come; we will be studying the excellent mutliple alignment programs that already exist, to see if we can outperform them, but HMMER has generally taken the view that de novo multiple alignment is a separable problem.

By: Andreas Wallberg

Andreas Wallberg — Fri, 19 Dec 2008 15:50:36 +0000

Looking forward to give it a spin. Will it be possible to produce alignments or sets of alignments that describe alignment uncertainty from unaligned (nucleotide or aa ) sequences?

By: Khader Shameer

Khader Shameer — Fri, 19 Dec 2008 11:02:13 +0000

Dear Eddy,

Happy to see your blog.
Looking forward for the Reloaded-HMMER.
Hope my web apps that use HMMER-programs will run faster than ever before Great !

K. Shameer

By: Kyle Ellrott

Kyle Ellrott — Thu, 18 Dec 2008 03:21:54 +0000

Keep up the good work. I’m looking forward to seeing that alpha as well.

By: Stinus Lindgreen

Stinus Lindgreen — Tue, 16 Dec 2008 08:58:22 +0000

This sounds very cool! Can’t wait to try it out when the alpha is released.

Best,
Stinus

By: Linkfest - Dec 15, 2008 : business|bytes|genes|molecules

Linkfest - Dec 15, 2008 : business|bytes|genes|molecules — Tue, 16 Dec 2008 03:48:05 +0000

[...] Let’s start with the news that Sean Eddy has started blogging. Cryptogenomicon is a suitably geeky name for a blog I am looking forward to reading (Sean’s obviously a Neal Stephenson fan). He brings the blog to life by posting about the program he is best known for, HMMER. Apparently HMMER3 has been written from the ground up, and has a clear goal in mind; to compete with BLAST. Read all about what to expect in HMMER3. [...]

By: Blogging Professors: Big Boffins with Blogs « O’Really?

Blogging Professors: Big Boffins with Blogs « O’Really? — Mon, 15 Dec 2008 18:42:09 +0000

[...] Title to be announced [...]