The scientific community has witnessed an increased interest in quantitative methods for discovering and developing new treatments for personalized medicine, defined by the U.S. Food and Drug Administration (FDA) as "the tailoring of medical treatment to the individual characteristics, needs and preferences of each patient." This is partly driven by rapid advances in genomics, computational biology, medical imaging, and regenerative medicine. Now scientists have powerful new tools for developing targeted therapeutics and to predict who will respond to a medical therapy or who will suffer its ill effects. In the last few years, several such treatments have been approved by the FDA, including ivacaftor for patients with cystic fibrosis and a specific genetic mutation, and cancer drugs crizotinib, vemurafinib, dabrafenib, and tremetinib, for patients whose tumors have specific genetic characteristics that are identified by a diagnostic test.
Statisticians from academia and industry have always been heavily involved in drug development through analysis of data from clinical trials and other experiments. With personalized medicine, however, the goal is not to show that a new drug is better on average than an existing drug or placebo. The new drug may well be no better in effectiveness when averaged over the entire patient population. But it may be superior within a subgroup of the population. This is the problem of subgroup identification: to identify (in terms of patient characteristics) the subgroup of the population for which the drug produces an enhanced effect. The task is very challenging, because clinical trial data are expensive and the number of participants is usually modest, yet the number of patient characteristics (the dimension of the search space) can reach to the hundreds or thousands (particularly with genetic variables).
The program will bring together international experts from academia and the pharmaceutical industry to share their knowledge and discuss research ideas. There is already a U.S. industry working group called "Quantitative Sciences in the Pharmaceutical Industry" dedicated to sharing information on exploratory and confirmatory techniques for subgroup identification and analysis. The proposed program will further this goal as well as provide the opportunity for academic and industrial statistics professionals to learn from each other. It will also be of interest to health and medical professionals in the pharmaceutical and biotechnology community in Singapore, where more than thirty of the world's leading biomedical companies have offices.
The workshop will consist of presentations by experts from academia, government and industry. It will have two parts, each lasting five days. The theme of the first part is design of experiments for personalized medicine. The theme of the second part is analysis of data from the experiments, with emphasis on identification of patient subgroups with differential treatment effects. A two-week tutorial on classification and regression tree techniques for personalized medicine will precede the workshop. The tutorial will be targeted towards statistics graduate students and professionals interested in a machine-learning approach to personalized medicine.