Features:

INTERVIEW WITH JILL MESIROV
By Tim Curns, Assistant Editor, HPCwire

HPCwire: What do you think are the largest obstacles deterring breakthroughs in biomedicine, and how is supercomputing being used to overcome these hurdles?

JILL MESIROV: I think the largest obstacle is not yet having the right models and methodologies to attack the challenging biomedical problems that face us. Biological processes are complex and noisy and we have a tremendous amount of work to do to understand them sufficiently well to build the right mathematical and computational models.

In addition, there are difficult issues around data integration and interoperability. This is well-recognized in the community and I expect there will be a lot of focus on trying to ameliorate these problems over the next three to five years.

HPC: How has your work using gene expression data to classify tumor types affected other health care concerns? Do you find that your research often transcends specificity and becomes of universal importance? How?

JM: The techniques we've been working on in cancer classification are aimed at extracting molecular signatures from a cell of known "phenotype" and using that signature to recognize the phenotype in a new sample. This is an extremely general approach and can, and has, been used for many different diseases. The method can be used for diagnosis, prognosis and outcome prediction. The signatures themselves contribute to understanding the mechanisms of disease as well as identifying possible targets for future therapies. We can also use this approach to identify signatures of known perturbations of a cell (e.g., RNAi, exposure to a chemical compound, etc). If we identify, annotate and catalog these signatures we can use them to gain insight when we recognize them in another setting.

HPC: How do you see the future relationship of bioinformatics and supercomputing evolving? What kinds of advances do you think we'll see in the next ten years or so? How will this relationship benefit our society in general?

JM: There is no doubt that biomedical research is becoming more reliant on computation as new technologies increase our ability to generate large amounts of laboratory data. Methodologies and resources to manage, store, analyze and mine that data will all require high performance computing.

HPC: You once said, "The way computer people think and the way lab biologists think tend to be quite different." Comment on the mixing of specialists in your area, namely biologists and computer science experts. Is communication difficult between these two expertises? How are these difficulties alleviated?

JM: I am very excited about how laboratory and computational scientists work together to attack biomedical problems. We all bring our special expertise and as we work together we learn from each other and bring that knowledge to the next question we try to answer. There is much power to be gained by this kind of multidisciplinary approach.