Quantum Bayesian Networks

March 20, 2010

China Needs Quantum Computers to do BioInformatics

Filed under: Uncategorized — rrtucci @ 3:07 am

Check out these 2 very interesting posts by Dr. Eric Drexler (nanotech guru, highly respected and famous scientist, author of the seminal book “Engines of Creation”):

Is 华大基因 doing science? (aka BGI)

Learning Bioinformatics

Excerpt:

In a decade, the global Human Genome Project sequenced 3 billion DNA base pairs. Today, a single machine (the Illumina HiSeq™ 2000) can sequence 25 billion base pairs per day, and BGI (the Shenzhen company formerly known as the Beijing Genomics Institute) has purchased 128 of them. This puts BGI “on track to surpass the entire sequencing output of the United States”.

Quantum computers could be very helpful in the field of bioinformatics, because a QC can do MCMC (Markov Chain Monte Carlo) quadratically faster than a classical computer, and MCMC is an important tool for performing some tasks in bioinformatics.

As I mentioned in my last post, I am currently writing Quibbs, a computer program for doing Gibbs sampling (a type of MCMC) with a QC.

I often get many panic attacks, filled with self-doubt, throughout the writing of a complicated piece of software. Will this code be truly useful to anyone, or is it just a dumb idea? Will someone soon invent a better algorithm or computer language, and make all my work a total waste of time? Will I be able to do all that I promised? Have I bitten this time more than I can swallow? Am I guilty of hubris? Is what I’m trying to do possible at all? Is my code failing because of a bug, or because my understanding of the theory is wrong? How to find the bug, assuming it’s a bug? Am I painting myself into a corner with this choice of code architecture? Is the code easy to understand? Do the variables, methods and classes have clear, consistent names? How to code this or that part more efficiently?

Two days ago, I was practically reduced to tears while coding a difficult part of Quibbs. Was my home-made algorithm, the one for coding Szegedy operators efficiently, failing its testing because of a bug, or because my algorithm was outright wrong? I was drowning in my own code, and in some tricky array indexing. But after a night of restful sleep and a morning of systematic sleuthing, praise the lord, I was able to fix it, and it started to work for my test examples. I think Quibbs is about 80% finished, and I believe that the 80% which is already done (and tested) includes all the hardest-to-do parts. Yippee!

Create a free website or blog at WordPress.com.

%d bloggers like this: