How Big Blue created the Blue Gene and is now delving into the Blue Brain

How did a chemistry PhD become one of the 50 most powerful women in business by hooking up the world's largest computer company with the life sciences? Kate McDonald spoke to IBM's Carol Kovac, who admits to being rather excited about the world of biology.

When Carol Kovac joined IBM's research department in California in 1983 with a doctorate in chemistry and very high hopes, her position was titled 'research staff member'. As one of her bosses said, at IBM "we start you at the top".

Not that her position was menial in any way. Kovac, who was in Australia recently to celebrate the 50th anniversary of the SILLIAC computer at the University of Sydney, says the position was actually very independent, a little like a university professorship in the scope it provided her. As a commercial entity, however, IBM was necessarily results based. Fortunately, that's just the way she likes it.

Kovac is now general manager of IBM Pharmaceutical & Life Sciences, a business unit within Big Blue that was only established in 2000 but that now employs more than 1500 people and brings in $US1 billion in revenue every year.

She was named in 2004 by Fortune magazine as one of America's 50 most powerful businesswomen, but what she gets excited about is the potential of the life sciences to change the world, all with the help of her company's computer know-how, of course.

"IBM Research is in many ways a very applied research environment, and because it is so problem-oriented, problem-driven, it is really interdisciplinary research," Kovac says of her early years down the ladder.

"I joined IBM when it was really at the heyday of the microprocessor world, when the personal computer really started to take off. IBM was ... starting to look at smaller machines with the acceptance of the idea of the personal computer, looking at new materials and new ways to put components together ... so my chemistry background was useful from a materials science perspective.

"That's one of the things that I have always loved about it: as a chemist I was able to work with solid-state physicists and engineers of all sorts, metallurgists, at that time we hired psychologists and economists when we got into e-business, mathematicians, and later on life scientists. It was a wonderful place to spend time."

As a vice-president of research at IBM for many years, Kovac was one of the people former CEO Lou Gerstner turned to when he decided to set up a series of emerging business units in the late '90s. IBM had a group working on computational biology but the realisation that life sciences and information technology was on the point of convergence led to the development of a separate life sciences section.

IBM saw great potential in the application of high-performance computing and powerful super-computers to manage the data that was coming out of developments like the Human Genome Project, she says.

"Through that [computational biology] group, which was part of my management responsibility at the Watson Laboratory [IBM Research's HQ in New Jersey] we came to understand that there was not just an interesting research attraction but there was a tremendous business opportunity for information technology in the life sciences.

"Lou Gerstner wanted to form a small number of independent business units that were intended to be quite entrepreneurial. We decided to create one around the life sciences and he asked me to lead that one."

When it began in August 2000, IBM Life Sciences had a grand total of two employees: Kovac and director of emerging technologies Jeff Augen, a biochemist.

Kovac admits it came as a bit of a shock at first, but she and Augen set their minds to developing the unit and bringing in some healthy revenue.

"We did grow and over the course of the next three years we identified some really key areas such as genomic processing, computer management for pharmaceutical discovery and then later something we call information-based medicine, which led us into a much bigger world of healthcare IT.

"We started with two people and negligible revenues and grew to over a thousand people, a worldwide organisation and about $1 billion in new revenue for IBM in 2003. It's an example of how if you bring a lot of the right pieces together and create the right business partnerships you can make something pretty profound happen."

Information-based medicine

While 'personalised medicine' is the mantra for many in the life sciences, Kovac likes to concentrate on what she calls information-based medicine and how information technology is bringing it about.

"Although they are certainly related, the only way to really achieve personalised medicine is to be able to create an information base that lets you say what is it about an individual that I need to know in order to define what the right treatment is.

"The way that we are going to get to that vision of personalised medicine is by creating pretty large databases of patient data - not just the genomic, although a lot of emphasis is placed on that - it's the environment that we live in, it's the food that we eat, the way our genes are controlled, what turns them on and what turns them off in different points in our lives - there's a lot of mystery in that.

"So information-based medicine is really about the collection of all that information and then being able to use computational methods to mine it, analyse it and correlate it and say, 'with this genomic profile and this set of environmental factors I can make some predictions say with breast cancer and I can then correlate that with this set of categories and I can say this is the best treatment'. This is already happening and is something that is very exciting for diseases like cancer and heart disease."

Although IBM is mainly considered a computer hardware company, much of its revenue comes from service provision and what it calls solutions. Kovac compares IBM in the life science area to a general contractor, one that integrates and pulls together the pieces that are necessary to create a full-blown solution.

"A solution, for example in the area of information-based medicine, is if you were a university and an academic medical centre or a large research hospital, you may say, 'we treat thousands of patients every year, we should be collecting data, gaining the patient's consent, collecting their data, de-identify it, put it into a database, tracking patients and keeping records of how they do on certain treatments electronically, and creating a basis for this world of information-based medicine. If we do that then the next patient who walks in the door, we can apply that knowledge. Today that knowledge is lost.

"I think this is where the power of these biobanks - information, samples - they are incredibly powerful research tools that are also a translational tool that will help get better therapies to patients quicker. You can tell I'm pretty excited about this."

The company has a pretty good history in the area of biobanks, having been a partner in Kari Steffanson's then-revolutionary ideas for building a biobank in his native Iceland. IBM assisted Steffanson's company DeCode, in association with the good people of Iceland, to genotype half of the adult population. IBM worked on some of the project's information systems and information analysis.

Since then, of course, biobanks and the like have sprung up all of over the world, with the UK Biobank planning to collect genetic information from half a million people.

In Australia, IBM is working with Melbourne Health to build a similar type of repository, the main idea being to build up a cancer 'grid'. "DeCode is not alone in this field anymore and there has been amazing progress," Kovac says.

Blue Gene

While IBM posits itself as a solutions company rather than a hardware company, Big Blue certainly knows how to build big computers. It was the driving force behind the design of Blue Gene, a massive supercomputer designed with a scientific purpose in mind.

"This was just before we formed the life sciences division, but we had a bunch of people ... thinking about a novel way to build a scientific supercomputer. They wanted to use an architecture that some people in the field of supercomputing refer to as 'embarrassingly parallel' - big clusters of machines. At the time there were 100 clusters and that was considered a lot.

"The Blue Gene architecture was actually intended to take thousands of racks ... gang them all together and put the right systems management in place. There's a crib, a roadmap for how supercomputing advances, and this machine was intended to beat the supercomputing roadmap by about three years.

"It was very exciting stuff but very risky science - no one really knew that we could do it in 1999 when we launched the project and then of course many of us said this is a great thing but we should think before we build this thing about its purpose.

"We put together this idea of the specialised architecture and the life sciences ... we said this is the grand challenge that we should aim this machine at, so that was the birth of Blue Gene."

There are five Blue Gene projects at the moment, with the first and fastest a collaboration between IBM, the US Department of Energy and the Lawrence Livermore National Laboratory at the University of California. "[Blue Gene/L] has about 360 trillion floating point operations per second (or 360 teraflops as the geeks like to call it) and eventually they will build a petaflop machine - a thousand teraflops," Kovac says. While the expense of such a large supercomputer seems beyond the pitiful resources of Australia, Kovac says the cost of such a machine is not as scary as you would imagine.

"It's quite affordable and it's also very scalable, so you can have a humungous Blue Gene machine like Lawrence Livermore does or you can have one rack. You can have 100 teraflops or you can have 10, and a 10 teraflop Blue Gene is about the size of a large refrigerator, although formerly it would have taken up half a basketball court."

And while a thousand teraflops sounds unimaginable, just compare it to the number of neurons in the human brain. "It's not right to make a direct comparison because they work very differently but we are approaching some really massive computational power that is already enabling us to conquer problems that we didn't really imagine five years ago."

Blue Brain

A comparison with the brain is not so out of the question, especially since IBM is using its mighty Blue Gene to assist in an extraordinary project currently taking place in Switzerland. This is the Blue Brain project, being run by the Brain Mind Institute at the Ecole Polytechnique Federale de Lausanne (EPFL) under the guidance of Henry Markram, principal researcher at the laboratory of neural microcircuitry.

Blue Brain is using a 22.8 teraflop Blue Gene to model the behaviour of 10,000 neurons in the neocortical columns of rats.

The project's Blue Gene is an 8096-processor computer that will model one to 10 neurons per processor. This will eventually allow simulations of as many as 100 million neurons, about half the number of neurons in a rat brain. Blue Gene sits on top of a room holding computer cables and cooling equipment fed from the icy waters of Lake Geneva, but the actual computer itself is only about the size of four refrigerators.

The first objective of Blue Brain is to build an accurate replica or template of a neocortical column (NCC) within two to three years, Markram says. That first template will be then be modified for different brain regions and species.

Charles Peck, head of the Blue Brain project for IBM Research, says the real value is that researchers will have access to data for every single neuron.

"Once we have modeled the neocortex, we will have to include models of other brain regions such as the basal ganglia, hippocampus, cerebellum and so on," Peck says.

Having an accurate computer-based model of the brain would allow researchers to create simulations taking seconds in silico, an unimaginable resource for neuroscience.

IBM is also building a special purpose processor for a US government agency called Cyclops, a very powerful "massively parallel" system, according to Dr Joseph Jasinski, program director for the healthcare and lifesciences institute at IBM Research. While he can't say much about the application of the new supercomputer, it is expected to be the most powerful computer in the world when it is up and running.

"We have two agreements with our client, one to design and build the chip and one to design and build the structure," Jasinski says. "The chip design is nearing completion and should be released to manufacturing in late November. The system design is complete and we are preparing to assemble the first system." Each chip has 160 processing units and the full system will have thousands of chips, he says.

Immediate future

For Carol Kovac, the next two to three years will be very interesting to watch indeed.

"We will start to see the very rapid development of biobanks and we will start to see more opportunistic computational work that's going to say we can start to create an evidence base for better medicine," she says.

"We may not know all the underlying mechanisms, we may not be able to go down to the molecular level of what is happening in the cell, but we can create correlations ... that are an order of magnitude in terms of how we treat patients.

"I think it's just very exciting. In the longer term there is the whole field of systems biology, which is enabling us to get much more targeted about identifying new genes and also interactions between networks of genes." She is particularly enthused about work being done such as that at the Institute for Molecular Biosciences at the University of Queensland, for instance. "They are doing some fantastic work on the epigenetics of what goes on in the DNA that doesn't code for proteins. We used to talk about junk DNA but people like John Mattick at the IMB are showing that it is part of a very elegant and sophisticated gene control system.

"Over the next five to 10 years we are going to start to see fundamental knowledge and new therapies, such as stem cells and siRNAs. New methods for gene therapies are also in the realm of probability. We couldn't be alive in a more exciting time. For IBM, it's exciting for us because at the core of it all is that the only way to make this happen is through the convergence of information technology and science."

Additional reporting by Matthew Hamblen, Computerworld.