CONVERSATION · CHRIS SANDER
February 10, 2003 | Computational biology and bioinformatics have already proven to be invaluable research and development tools in biomedical research labs. But what if they could be applied to the clinical world? Chris Sander intends to do just that in his newly appointed role as head of the Computational Biology Center at Memorial Sloan-Kettering Cancer Center (MSKCC) in New York City. Sander sees profound benefits in using the technologies to enable clinicians to make better diagnoses and prognoses, and to even help hone therapeutic plans. Sander was recruited by MSKCC President (and former NIH Director) Harold Varmus, who says Sander's expertise in in silico biology "will help us deliver on the promise of the Human Genome Project."
Sander brings a wealth of experience in bioinformatics and computational biology to this new post. He has a long track record in developing software to analyze and predict the properties of proteins, worked at the European Bioinformatics Institute, and was chief information officer at Millennium Pharmaceuticals Inc.
Sander was one of the driving forces in the founding of the department of biocomputing at the European Molecular Biology Laboratory in Heidelberg, Germany. And he helped found EMBnet, a network for molecular biologists in Europe. Sander serves as the editor of the respected computational biology peer-review journal Bioinformatics, and as an advisor to IBM Corp.'s Deep Computing Initiative.
Bio·IT World Senior IT Editor Salvatore Salamone spoke with Sander recently in his office at Memorial Sloan-Kettering.
Q: What made you come to Memorial Sloan-Kettering?
A: I thought very hard about taking on this challenge of building a significant center for computational biology. At Memorial Sloan-Kettering there is a unique opportunity to bring together an interdisciplinary set of people who are going to work together to achieve a common goal, which is simulating biology in the computer.
Q: What are the benefits of having a computational biology center in a clinical environment?
A: There's an opportunity to build bridges by applying [computational biology] basic research to the clinical space. If you look at the history of molecular biology with the sequencing of the human genome and the development of a number of high-throughput technologies like miniaturization, robotics, and bringing computers in to analyze experiments, we now have an enormous amount of detail about biology. But most of it is for yeast and animals. We do not put humans in a lab to do experiments on. So I see molecular biology, genomics, and technology being applied now to human biology in a way that is unprecedented.
Q: How will that be done?
A: We will seek out opportunities where [computational biology] methods can be applied to diagnosis, prognosis, and therapy. The goal is the overall improvement of the quality of life for people who have cancer. This work will also have implications for other clinical areas.
We have a unique opportunity here because of our location. Our campus in Manhattan is in the middle of a constellation of institutions, which, by the way, are not devoid of history in this field. The Computational Biology Center's efforts will be part of a tri-institute initiative where we'll be trying to develop a collaborative research program in a number of areas between Memorial Sloan- Kettering Cancer Center, Rockefeller University, and the biomedical college of Cornell University.
Q: What's the game plan?
A: We will build a center that plans to have seven research faculty, of which I'm the first. The center will provide a bridge between basic research and clinical research. There will be about 70 or 80 people in the center and a support unit called the bioinformatics core that will build up the infrastructure that you need in high-performance biological computing to prepare the simulations.
Q: When will other faculty members be added?
A: The search is open. Over the next few months we'll identify the first one or two faculty. We're aiming for high-quality people with interdisciplinary backgrounds. For example, we're looking for someone who would do computational genetics. That would be someone who has a professional background in genetics but will [incorporate] the methods that allow us to look at genetic variation across the human genome. This would include using the rapidly growing database of SNPs (single nucleotide polymorphisms) and using whatever comes out of the National Institutes of Health's HapMap project, which is what you really need to have half a chance to connect the genotypic variation in a human being to phenotypic consequences of the relevant disease.
Q: What other areas are of interest?
A: Systems analysis or systems biology. We're particularly interested in the effect of external influences from a systems point of view on cells and collections of cells. From a systems biology point of view, the perturbation of a biological system at the cell level and the organ level, the perturbation by an external agent — in particular, a small compound — will be a prime tool to develop a certain flavor of systems biology. That is what we will aim for.
Another area is computational physiology, including, for example, the modeling of the heart, of cancer physiology, and computational modeling of the development of organs. We'll be looking for someone to come on board to study computational physiology with the particular advantage of connecting the physiological phenomena at the level of organs to the microscopic biology at the level of genes, genetic variation, and molecular properties within cells.
The long-range goal, which is very complicated and very hard, is to connect genotype to phenotype — microscopic biology to macroscopic biology. That's a tall order. We know that when we take small steps in that direction there is a huge impact. In particular, we know it's already possible to connect certain specific genetic variations to specific disease consequences.
Q: What computational facilities will the center have?
A: We're going to build up a small, high-quality facility for biological computing. We will probably start with a Linux cluster that will have on the order of a couple of hundred processors. And we'll expand that as needed. At the same time, we will very actively seek out opportunities to connect to Cornell [University] and the San Diego Supercomputer Center. Cornell is part of the tri-institutional research program [that includes Memorial Sloan-Kettering]. And San Diego has a very good biology orientation. We are not going to primarily have all the hardware in the building. We want to exploit distributed computing, especially with those centers. And, opportunistically, we will collaborate with people and vendors in grid computing. When I was at the European Bioinformatics Institute, we had excellent collaboration with [high-performance computer] vendors like SGI and DEC [Digital Equipment Corp.].
But keep in mind our prime goal is not the hardware or the software, but solving biological problems using computational means, which is the mathematics, the algorithms. So collaborations with other communities are important.
Q: What kinds of partnerships do you anticipate?
A: We're interested in partnering with software developers. There's a very interesting area of information processing combined with natural language processing. This is an aspect of computational analysis and information processing I would like to link into the scientific space. I've done projects like this when I was at Millennium, collaborating with computational linguists. Some of the knowledge extraction and adding value to collected information that can be derived in this area is very important in a biomedical setting.
Q: What are some of the advantages computational biology will bring to clinical workers?
A: It goes back to the question of how do you use the human genome and other results of modern molecular biology, and how do you bring that to bear in human biology? One obvious place to start is to take molecular profiling [DNA micro-array analysis] and apply that to cancer. You want to be able to use the power of molecular profiling on tumor samples, on body fluids, and on blood to do early diagnosis, to have a more accurate diagnosis of a tumor, to have a prognosis. You want to know what's going to happen to the patient, what's the survival probability, what's the chance of going to metastasis, what would be the metastatic pattern, what is the survival time.
Q: How would this be applied to the clinical work here?
A: Clinicians tell me there's considerable uncertainty in terms of predicting what will happen to a patient. This is important for the individual, and it's important when considering what type of observations you bring to bear and what kind of therapy you choose for a particular set of patients.
So it is extremely important to be able to make an accurate prognosis. And you want to be able to choose one therapy treatment over another based on the very detailed molecular picture of the body fluids and of the genetic background of the patient. The current status is you choose therapy according to a set of fixed criteria. The clinical staffers would like to refine this and make it much more personalized, much more appropriate, much more dynamic. In other words, give them a chance to adjust the therapy as the patient is observed.
|We want to develop the decision-support systems for clinicians based on understanding the pathways and making very detailed molecular measurements. The vision is to use integrated molecular profiling and bring it to bear to make better and more refined decisions in the clinical setting.
One way to do this is to look at the molecular composition of tissues and fluids. This will take some time to develop. The promise is there; the opportunity is there. Imagine you have different pathways that are perturbed in a cancer and you are able to look not just at the messenger RNA levels but also at the protein, at the covalent modification of the protein, and do all that knowing the genetic background of the patient. You can imagine that in principle that gives you a very powerful combination. We want to develop and make available the decision-support systems for clinicians based on understanding the pathways and making very detailed molecular measurements. This will take many years, but in time I think this will have enormous impact potentially on the quality of health care, particularly in cancer.
Q: How will computational biology get into the clinical environment?
A: While we are basic scientists, and we are going to develop the basic science of computational biology that will allow us to simulate human biology, that's our goal. We definitely want to apply this knowledge as soon as we can and in a very directed fashion to clinical research.
We already have an excellent genomics facility that does microarrays, and a strong proteomics facility. We'd like to add genotyping and develop something that I call 'integrated molecular profiling,' where we combine information about messenger RNA, proteins, protein modifications, excess functional RNAs, and metabolic profiling, plus take into account the background of the patient. This will take time, but the vision is to use integrated molecular profiling and bring it to bear to make better and more refined decisions in the clinical setting.
PHOTO CREDIT DENNIS GALANTE