Monday, May 31, 2010

PLASMA AND MAGMA, future keywords to keep in mind

I had the great opportunity to attend a talk by Jack Dongarra in 2001 at a seminar-workshop at Merida, Venezuela, called Second Latin-American School on Parallelism and High Performance Computing. This was the first time I heard about the top 500 list of supercomputers, and also the first time I ever heard about grid computing, which has in a way evolved into cloud computing nowadays.
I remember that Dongarra's talk was one of my favorites because of its strong link to Moore's Law, and the prediction of the computational future as a simple power law. Since then, I've learned that I must change my laptop roughly every 14 months in order not to loose my investment completely, and also to be able to work competitively in my field, that of computational chemistry.
Due to a short news article by the BBC, I remembered, once again, the name of Jack Dongarra. When I googled "top500 dongarra" I stumbled upon a youtube talk at the University of British Columbia, I higly recommend watching it.

From the talk I learned that, if you do any kind of programming, and you do some sort of linear algebra operations in your code, you will inevitably have to move to PLASMA, that is, the new algorithms which take advantage of parallel execution, which is a growing need, according to the frequency limit on computer chips, clearly and neatly explained in Dongarra's talk at U.B.C.
Additional to PLASMA, we will have to learn MAGMA, that is, if we are going to use GPU's efficiently to solve computational chemistry problems, which brings to mind two relevant but naive questions:

1) Does the newly GPU implemented Amber 11 use MAGMA?
2) Would Dongarra be interested in having the generator matrices scheme for polymer modeling adapted to benchmark supercomputers? They already have gromacs and gamess in the fastest supercumputer, JAGUAR, so, why not?

Anyhow, I highly recommend periodically checking Dongarra's talks and ideas if you want a peek into the future.

Tuesday, May 18, 2010

Weekly Review / Monday May 17, 2010

In PDB code 2KR8, one has an example of a K-turn motif. The authors of the study call it specifically a U4-Kt RNA. It seems like there are no specific structural studies of K-turn RNA not bounded to protein. The study which discloses this PDB structure says to be the first to shown an unbound Kt-RNA structure using NMR

The authors are based in Grenoble France and Heidelberg Germany and publish in Nucleic Acids Res. 2010 May 13

Turner Lab. just released a new reparametrization of the amber99 forcefield. They call it amber99chi. So, the name makes it obvious. It has been reparametrized to get the chi torsion angles right in RNA, and also it has been reparametrized to get the sugar puckering conformations of RNA right, or at least better than amber99. They compared the new force field with NMR data and get better agreement than with just amber99.
J Chem Theory Comput. 2010 May 11;6(5):1520-1531

Friday, May 14, 2010

Uracyl-Tetrad Stabilizes Guanine-Quadruplexes

Simply put, the result shown in JACS that I want to blog about is the following:

"Besides G-platforms, there are also U-platforms"

That is, RNA can make a "planar" hydrogen-bonded network with four Uracyl bases.
It has been known for a while that such type of tetrads are possible for Guanine, and when various Guanine tetrads stack in top of each other they are called Guanine-Quadruplexes. It seems like the place where this structures occur is the human telomere. The human telomeres are regions of repeated DNA sequence at the end of chromosomes which protect DNA from degradation, their shortening is linked to the process or cellular aging.

The Tokyo University group of Xu, Ishizuka, Kimura, and Komiyama, shows that the U-tetrad stabilizes a G-Quadruplex structure. The results are obtained using NMR, so I don't know if atomic coordinate files are available to play with, if there are, I will post them here for you and I to play with.

Tuesday, May 4, 2010

Ribosome Controls 2/3 of the Dogma in Bacteria

transcription speed = translation speed
speed of mRNA making = speed of protein making

Messenger RNA is synthesized by RNA polymerase (RNApol).
In the synthesis process two states can occur, either a forward synthesis state, or a backwards (i.e. backtracking) state making an equilibrium of:

Backwards = Forward

Sergey Proshkin, A. Rachid Rahmouni, Alexander Mironov and Evgeny Nudler show in Science. 2010 Apr 23;328(5977):504-8 that the ribosome shifts equilibrium to forward. They measure the rates of transcription and translation and use molecular tricks to enhance translation. Surprisingly the enhancement of translation induces higher transcription rates.

The second part of this discovery is the molecular component, which comes from Björn M. Burmann,Kristian Schweimer, Xiao Luo, Markus C. Wahl, Barbara L. Stitt, Max E. Gottesman, and Paul Rösch in Science. 2010 Apr 23;328(5977):501-4.
They characterize by NMR a transcription factor (a polymerase protein) called NusG, which, it seems, will bind to a ribosomal protein called NusE, therefore binding RNA polymerase and the ribosome.

With this in mind, it is not clear to me how to interpret the polysome micrographs for the coupled process in bacteria.

My guess is that only the "first" ribosome in the polysome couples to RNA polymerase, the other ribosomes which attach to mRNA in the polysome do not control the coupled rates of transcription-translation, and most likely don't care much about transcription fidelity. I wonder, if you were a protein, and then you were synthesized in that first ribosome, would you be different to your twin copy from the second or the third ribosome in the polysome?

I found this movie for the eukaryotic case:

NOTE: One reason why this is only possible in bacteria is that prokaryotes don't have a spliceosome, there is no splicing of pre-mRNA into mature-mRNA as far as I know.