A systems approach

SPL

Research associate Katie Hoadley works in a genetics wet lab, but she hasn't picked up a pipette for three years. On this particular morning, at the University of North Carolina at Chapel Hill, she scrolls through lines of computer code to identify patients who responded well to a breast-cancer treatment in a published clinical trial. Using algorithms based on gene-expression data from human breast-tumour cell lines, she is looking for molecular signatures that might help to predict which breast-cancer patients will be most amenable to the treatment.

An experimentalist by training, Hoadley has benefited from the increasing demand worldwide for cancer systems biologists, especially those who can handle computation. Instead of collecting data on animals or cells and funnelling them through basic statistical formulae, Hoadley has learned to merge and analyse data from others' experiments.

The rise of systems biology could not have come soon enough for cancer: recent findings point to daunting heterogeneity within individuals, and even within tumours over time. This genetic complexity plays a part in many tumours' tendency to resist treatment, and rummaging through that complexity is exactly what systems biologists do. For young cancer researchers searching for a niche, it's clearly an area with promise.

Pinpointing the start of the systems-biology approach to cancer is difficult. Researchers have been examining cellular systems for therapeutic leads for some time. But the Human Genome Project and technological improvements have brought a new level of complexity to the approach. Rather than focusing on one molecular pathway, this integrative approach blends many contexts, including DNA, RNA, proteins, signalling networks, cells, organs, whole organisms and even environmental factors.

This varied data mix requires scientists to build complex mathematical models of cancer, which in turn drive new research questions. Once they have been validated in cells, animals or human tissues, those results inform new or better models. The ultimate goal is to create a more comprehensive understanding of cancer and to forecast outcomes and therapeutic efficacy in the clinic. The increasing volumes of data reaped from comprehensive scans of molecular markers offer a fertile career ground for those willing and able to acquire computational know-how. In Hoadley's case, for example, doing so required on-the-job learning to code in the programming languages R and Perl.

Technological advances have enabled researchers to measure molecular processes such as DNA methylation, copy-number variants or single-base mutations in a single experiment. And they can use a variety of approaches, including next-generation sequencing platforms, mass spectrometry, high-end imaging and analysis tools. “This sounds terrific, except for the fact that the data analysis and visualization are very difficult,” says David Botstein, director of the Lewis-Sigler Institute for Integrative Genomics at Princeton University in New Jersey. The approaches require a level of mathematical sophistication that many biologists do not have, he adds. Biologists now often receive such training in graduate school, as part of postdoctoral positions, or they at least learn how to communicate and work closely with computational experts. Industry and academia need researchers who can manipulate and interpret reams of data, turning the information, for example, into markers of disease progression or therapeutic efficacy.

Because training in systems or integrative biology is so new, and programmes are still emerging, many established cancer systems biologists began with either theoretical or experimental backgrounds, but not both. Many graduate-level programmes provide a good grounding in systems biology, but students may have to make a concerted effort to steer their training towards cancer applications. For example, there are no specific curricula in cancer systems biology in Germany, says Ralf Herwig, group leader of bioinformatics at the Max Planck Institute for Molecular Genetics in Berlin. In the next five years, he predicts, more systems-biology curricula will appear at universities, and he expects that cancer-biology applications will appear in places affiliated with clinics, such as the Charité University Hospital in Berlin.

Even though technology is becoming more user-friendly, Joe Gray thinks applicants will need to understand the basic principles of computation.R. KALTSCHMIDT, BERKELEY LAB PUBLIC AFFAIRS

Irrespective of location, the key is often on-the-job training. Those with a foundation in biology or clinical oncology may glean statistical and computational skills from other projects. This is still a reasonable strategy, although the field is beginning to recognize the need for more interdisciplinary training, says Hoadley's supervisor Chuck Perou.

Hoadley and Perou both began as bench scientists and learnt computational skills by working with accomplished collaborators. As a postdoc, Perou listened to statisticians talk about specific methods during lab meetings. Afterwards, he would study those methods on his own.

Those interested in experimental aspects of systems biology should be comfortable with commonly used tools in data analysis and mathematical modelling, notes Steven Altschuler, a pharmacologist at the University of Texas Southwestern Medical Center's Green Comprehensive Center for Computational and Systems Biology in Dallas. They should take classes in cell and molecular biology and discuss their thoughts and questions with people in those fields. “If they come from engineering, physics or math, we always ask, 'Have you sat in on a biology class?',” he says.

To encourage interaction, Perou arranges his lab space so that at least one computational student or postdoc works in each room of the wet lab. “There's this constant interplay between statisticians, biologists and computer scientists,” Perou says. Two papers have come from algorithms generated through these interactions. “It is important that we are all in the same room talking to each other,” he adds.

In an effort to improve systems-biology training at the postdoctoral level, the US National Cancer Institute in Bethesda, Maryland, earlier this year awarded Sage Bionetworks, a non-profit medical-research organization based in Seattle, Washington, a US$6.7-million grant to help train more systems and network biologists. The programme will pair clinical biologists with physicists or mathematicians, preferably those who have already completed a postdoc in their area of expertise. The pairs will then work together on a project for two years. Sage Bionetworks president and chief executive Stephen Friend says that the programme will generate more systems and network biologists for both academia and industry.

Training in cancer systems biology will change as the tools advance. Massive efforts such as the Cancer Genome Atlas, a project funded by the US government and started in 2005 to uncover genetic mutations responsible for cancer, have led to better analysis tools. And some speculate that these tools will become more user-friendly with time. “As that happens, it will become easier for people who aren't deeply trained in computational biology to ask computational questions,” says Joe Gray, director of the life-sciences division at the Lawrence Berkeley National Laboratory in California. “But you are still going to have to understand how the algorithms work, and the nuances of how the data are analysed.”

But Peter Sorger, director of the Massachusetts Institute of Technology's Cell Decision Process Center in Boston, says he doesn't see the challenges of modelling cancer going away any time soon. “I don't think that one person can do it all,” he says. The field will demand an even greater shift in academic structure and culture towards interdisciplinary training, according to Sorger. That means that all individuals on a team should have at least a working knowledge of both modelling and experimentation. Then trainees will gain a grounding for future collaborations in academia and industry, he adds.

Academic job opportunities abound, particularly in the United States, Britain, Germany and Switzerland, and at postdoctoral and group-leader levels. These include jobs in traditional university departments and in specialized research centres such as the Institute of Systems Biology in Seattle or the European Bioinformatics Institute in Cambridge, UK. In addition, the Cancer Genome Atlas and other efforts to sequence thousands of individual tumours are upping the demand for bioinformaticians. “That's really one of the limiting things for them — finding competent people to analyse all these data,” says Botstein, who adds that the same is true for industry. According to Botstein, this means that interested scientists should be learning computation in any way possible.

In many places, however, the field remains nascent, says Duncan Odom, leader of the cancer-genomics group at Cancer Research UK's Cambridge Research Institute. Odom emigrated to Britain after completing a postdoc at the Whitehead Institute for Biomedical Research in Cambridge, Massachusetts. He says that systems biology is a strong theme at Cancer Research UK, but that the organization does not have a set programme, something that Odom is pushing to change.

Historically, industry has viewed cancer systems biology, particularly large-scale proteome and genome analysis, as requiring a large investment with few immediate returns. In the past year or so, however, this perspective has changed as firms start to realize that relatively small-scale but systematic experiments can help them to prioritize potential drug targets and uncover drug mechanisms. If this initial promise pans out, pharma and biotech will probably expand their hiring efforts, Sorger says. In Europe, that means big companies, and in the United States, it ranges from small to big firms. Some, such as Pfizer, Genentech and Microsoft Research, are already starting to hire, particularly postdocs. “I think we're in transition,” Friend says, noting that several companies have invested in projects to model networks of genes, proteins and other molecules.

Altschuler says that he and his co-lab head Lani Wu are looking for researchers with training in cancer biology, imaging and cell signalling who can examine the responses of cancer cells to drugs. “The people we hire who have strong mathematical backgrounds also have a strong interest in applying their skills to cancer biology,” he says. But with the field constantly changing, graduate education is not the only qualification. In the case of cancer systems biology, Wu says, “your background matters less than the research questions you want to answer”.