More from IDG.
--
for genomics research
by George A. Chidi Jr., IDG News Service\Boston Bureau
January 19, 2001, 14:39
Compaq Computer Corp. has signed on to provide supercomputing
technology to the Sandia National Laboratory in New Mexico and
Celera Genomics Corp. in a four-year cooperative research and
development agreement announced Friday by the U.S. Department of
Energy. Scientists hope to develop software and computer hardware
specifically designed for the demands of computational biology and
applied life sciences research.
Researchers are aiming for a supercomputer capable by 2004 of
processing 100 trillion operations per second -- 80 times faster than
the computers Celera used to sequence the genome and about eight
times faster than the fastest supercomputer currently working.
Ultimately, they hope to develop a petaflop supercomputer -- that is,
1,000 trillion operations per second.
Compaq executives see supercomputing in biological research as a
very fast growing part of its business. "All of the supercomputing
technology we've developed in the last ten years is beginning to be
the very tool that our biologists need," said Bill Blake, Compaq's vice
president of technical computing.
Blake said the supercomputer deal could mean hundreds of millions of
dollars in revenue for Compaq, without being more specific. In the
long term, the research could be worth billions to the company. Blake
estimated the technical computing market to nearly double over the
next 10 years, and said that genomics supercomputing is the fastest
growing part of that increase.
Compaq could take the technologies developed in the project -- the
software, optimized hardware and perhaps consulting services -- and
sell it in a somewhat smaller package to biology research startup
companies.
The supercomputer developed with Celera and Sandia will be based on
Compaq's Alpha microprocessor, and will contain as many as 10,000
to 20,000 processors, Blake said. One key to the research for Compaq
will be for the team to develop a genomics software architecture that
will work effectively with a supercomputer of 100 processors as well
as one of 10,000 processors.
Until biologists applied supercomputer technologies to researching the
human genome, nuclear physics remained the greatest challenge for
computer science. With the human genome sequence complete,
biologists are looking to the computational expertise of nuclear
weapons researchers to build algorithms for genetics research on
computers.
"Now that the genome is sequenced, we're entering the era of holistic
biology," said J. Craig Venter, Celera's president and chief scientific
officer. Holistic biology, as he describes it, refers to a blurring and
blending of scientific disciplines in pursuit of biological research.
Venter and other scientists cite the small number of researchers
skilled in bioinformatics -- the science of applying information
technology to biological research -- as a bottleneck potentially
slowing genomic research. Bioinformatics is complex and
multidisciplinary, requiring facility with physics and arcane
mathematics as well as biology and computer science.
"What attracts people to biology is not what attracts people to
physics," said Blake. "There are too few people that live at the
intersection of biology and computer science and engineering."
One of the primary goals of researchers is to develop visualization
technologies for analyzing the massive quantities of experimental
data, simply to be able to look at the information as something more
than streaming lines of code or seemingly random strings of C's, G's,
A's and T's.
"We in the nuclear weapons community believed nothing could be
more complex than the development of nuclear weapons," said Bill
Camp, Sandia's director of computation, computers and mathematics.
He conceded that genomics is more complex, but called that
complexity itself a problem. "The science must be simplified," he said.
Writing programs for supercomputers requires an understanding of
parallel system architecture, which is different from computers with
single processors. Data entering a supercomputer must be broken up
into bite-sized chunks and distributed evenly among the many
processors in the computer. Sandia's scientists will apply the same
computer skills used to model nuclear explosions to develop algorithms
for breaking up genetic data for parallel processors as well as
visualization tools for genomics with Celera.
U.S. government researchers have been instrumental in developing
genomic algorithms, in no small part because the fastest
supercomputers have been in government hands. And both genetics
and IT companies have partnered extensively with government
agencies to make use of the information.
For example, last month IBM Corp. and NuTec Sciences Inc.
announced plans to build a 7.5 trillion calculations-per-second
supercomputer to investigate genetic expression and disease, using
the "multivariant analysis of gene expression" algorithm developed by
the National Human Genome Research Institute.
While the Los Alamos and Lawrence Livermore national laboratories
are involved with genome research, Venter said he sought to partner
with Sandia because their researchers have particular expertise in
massively parallel supercomputing. "Biology can't proceed without
high-end computing," he said. "We need the expertise in both
communities to proceed."
Camp said genomics and proteomics -- the study of the physical
structure and function of proteins in cells -- will require more powerful
computers manipulating more data than was necessary to sequence
the genome itself.
Compaq can be contacted at compaq.com. Celera
Genomics can be contacted at celera.com. |
