Conference addresses quantitative challenges of using complex genomics data in medical research

HSPH Prof. Xihong Lin

December 13, 2011

With the completion of the 13-year Human Genome Project in 2003, many scientists have been hopeful that the wealth of new genetic information would help predict disease risk for individuals, help doctors ascertain which drugs would be most likely to help individual patients, and help manufacturers make safer drugs. Genomics, it was thought, would usher in a new era of personalized medicine.

Experts speaking at the fifth annual Program in Quantitative Genomics Conference, hosted by the Harvard School of Public Health (HSPH) Department of Biostatistics and the Department of Biostatistics and Computational Biology at the Dana-Farber Cancer Institute on November 17-18, said that the benefits of genomic medicine are promising but there’s still a long way to go. The event was held at Harvard Medical School’s Joseph B. Martin Conference Center.

In opening remarks, Xihong Lin, professor of biostatistics at HSPH, noted that the conference would focus on three key areas of research— personalized medicine, reproducibility and validation of genomic signatures in translational medicine, and genetic risk prediction.

Lon Cardon, senior vice president of genetics for GlaxoSmithKline and one of the conference keynote speakers, said that finding new targets for new medicines was a key objective of the Human Genome Project and subsequent initiatives, and that some of that work has shown promise. However, he added, “Outside of oncology, the personalized medicine focus has not been successful. The translation hasn’t been there…despite a great deal of hyperbole.” Several speakers discussed the need to develop novel clinical trials using genomic signatures.

Lon Cardon of GlaxoSmithKline

Others discussed the importance of reproducibility and data provenance in translational medicine and the challenges of compiling complex genomic data and using the data to develop effective ways to predict or treat disease. Keith Baggerly, professor of bioinformatics and computational biology at the MD Anderson Cancer Center, told of how a 2010 clinical trial at Duke University—in which patients’ gene patterns were assessed to determine which drugs would best attack their particular cancers—had to be stopped after Baggerly and his colleagues discovered that the Duke researcher, Anil Potti, was using erroneous genomic data.

Following Baggerly’s talk, Robert Califf, Duke’s vice chancellor for clinical research and director of the Duke Translational Medicine Institute, described the university’s efforts to develop and implement new measures to better oversee translational research and clinical trials in the wake of the incident. Some of those measures include a stronger focus on data provenance, on adequate quantitative expertise in managing genomic data, and on quality accountability.

Califf also talked about the tremendous desire, particularly in the cancer arena, to create and test new therapies, and the conflicts of interest that can arise when lots of money is at stake. While there is no substitute for integrity, he said, it’s very important for research institutions to have adequate research controls in place and a culture that promotes adherence to rigorous research standards.

Gilbert Omenn, a keynote speaker and professor and director of the Center for Computational Medicine and Bioinformatics at the University of Michigan, stressed the importance of confirming and validating analytical findings and clearly demonstrating the clinical usefulness of new genetic tests and therapies.

Genomics research has been most successful in predicting and treating diseases that can be traced to a single defective gene—so-called “monogenic diseases”—such as Huntington’s disease or cystic fibrosis or BRCA-related breast cancer, said another keynote speaker, Cecile Janssens, associate professor of epidemiology at Erasmus University Medical Center in the Netherlands. Genomics analyses have been less helpful when it comes to predicting or treating diseases where many genes play a role, such as type 2 diabetes or lung cancer. With type 2 diabetes, she noted, predicting disease risk based on known factors—such as age, sex, and body mass index—is likely to be much more helpful than looking at genetic factors that have been discovered.

In spite of the current limitations in using genomics data, Janssen said, “It doesn’t mean we can do nothing at all.” The bottom line? Much more research is needed, she said.

–Karen Feldscher

photos: Aubrey LaMedica