Human Genetic Variation Alters Anthrax Toxin Sensitivity
Martchenko, Candille et al., PNAS
Researchers at Stanford University School of Medicine show that genetic variation affecting capillary morphogenesis gene 2, or CMG2, dramatically alters toxin sensitivity in humans. In its analysis, the team reports on "a CMG2 single-nucleotide polymorphism occurring frequently in African and European populations [that they found] independently altered toxin uptake." The group goes on to suggest "testing of genomically characterized human cell populations may offer a broadly useful strategy for elucidating effects of genetic variation on infectious disease susceptibility."
NVIDIA Launches Community Website Aimed at Biotech
January 14, 2010
Today NVIDIA is singing from the rooftops about its new "Tesla Bio Workbench," a community website aimed at spreading the GPU gospel to the life sciences community. Included in the site are links to resources for GPU computing, discussion forums, and a slew of freely available, CUDA-enabled GPU software. Currently, the site offers the following MD and quantum Chemistry applications for use of NIVIDIA cards:
AMBER
GROMACS
HOOMD
LAMMPS
NAMD
TeraChem (QC)
VMD
Coming soon are GROMOS, GPU-HMMER, MUMmerGPU, and CUDA-SmithWaterman. Of note here is NVIDIA's attempt to show off its stuff with standard bioinformatics algorithms for sequence analysis, which are not usually thought of as benefiting from being ported to a graphics processor.
In a recent conversation I had with Sumit Gupta, a senior product manager with the NVIDIA Tesla GPU Computing Group, it was clear that NVIDIA's CUDA software development kit was the go-to choice for porting these life sciences applications to GPUs over OpenCL, the open standards programming platform for multicore and GPU chips. Not surprisingly, Sumit says that most of the developers who are responsible for the applications on the new website have played with OpenCL, but decided to stick with CUDA, which has a two-year jump in development on OpenCL. "A bunch of them have explored the use of OpenCL but have chosen to stay with CUDA C [which] is a very mature technology," says Gupta. "OpenCL is kind of a basic standard, it tries to be everything for everyone...if you write a program for OpenCL you have to optimize it for NVIDIA's GPUs, then optimize it for ATI's GPUs, and then optimize it for a multicore CPU, so you have to optimize it for every platform-OpenCL is not the panacea that everyone thought it would be."