[BiO BB] Blue Gene/L

[Boris Steipe]

> Spencer Covington wrote:
> 
> My school ( Arizona State University ) has recently acquired a supercomputer (
> an IBM Blue Gene/L ).  I am preparing to begin a M.S. in Computational
> Bioscience ( Genomics / Bioinformatics ) in the fall.  Would like to design a
> project to run on this computer.  What kinds of projects should I consider ?

I find this quite amusing. This almost sounds like a troll: multi-million $$$
solution, deprecating at a significant perecentage each month, waiting for a
problem. I've heard similar machines play wicked chess though.

Seriously: think about a project to improve MM-forcefields for partially folded
proteins with an aim to improve the radius of convergence in folding
simulations. 

This is a problem that has 
- an objective function that is hard to compute ( radius of convergence is
essentially the largest RMSD which you may be off from the native structure and
still find it through minimization, i.e. not combinatorial search)
- yet parallelizes well across processors
- raises interesting issues in combinatorial optimization
- has important applications
- and needs to keep a lot of stuff in memory at the same time (explicit solvent,
atoms etc.) and thus is probably not well suited for distributed applications.

Best regards,



Boris

---
Prof. Boris Steipe
University of Toronto
Program in Proteomics & Bioinformatics
Departments of Biochemistry & Molecular and Medical Genetics
http://biochemistry.utoronto.ca/steipe