THE 24 chromosomes that make up our genome weigh in at 3 billion base pairs
of DNA. That鈥檚 25 times larger than any other genome that has been sequenced and
eight times the total size of all the other genomes published so far.
Then there鈥檚 all the additional data being generated as the genome is
鈥渁nnotated鈥�, or analysed. It takes a fair amount of computer memory just to
store all this information鈥攁nd the rate at which we鈥檙e accumulating
genetic information quadruples each year.
But the storage problem pales in comparison with the resources we need to
make sense of all this data. When it comes to assembling genomes, finding genes
and variations between them, and comparing whole genomes, the demands are
immense. 鈥淐omputational biology will become the premier skill people will need,鈥�
says Craig Venter of Celera.
Advertisement
This means that bioinformatics is becoming one of the driving forces behind
the development of high-performance computers. Compaq, the US government and
Celera plan to build the world鈥檚 largest supercomputer to analyse the genome. It
will have 1500 gigaflops of computing power鈥攎ost PCs would struggle to
achieve just 1 gigaflop.
But it鈥檚 not just about building faster processors. Clever tricks are being
used to maximise the performance of existing computers, says Phil Butcher, head
of IT at the Sanger Centre in Cambridge.
One key tool is a piece of software called a load-sharing facility, or LSF,
developed by Platform Computing of Toronto. It distributes tasks among clusters
of computers linked by high-speed fibre-optic cables, so there鈥檚 no need to
build supercomputers. 鈥淚t鈥檚 a very much cheaper, more affordable technology,鈥�
says Butcher.
The Sanger Centre is planning to link its network with part of that of the
neighbouring European Bioinformatics Institute. And computer clusters at
different centres around the world could soon share the task of processing
genome information via a high-speed network called the Grid. 鈥淭hat will happen
within the next couple of years, I鈥檓 sure,鈥� says Butcher.