How Statisticians Can Help Bring on Personalized Medicine

• Activity Number: 273
• Type: Invited
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Terence (Terry) Paul Speed*+
• Companies: Walter & Eliza Hall Institute of Medical Research
• Keywords: "omic" data analysis ; drug selection ; individualized treatment

Abstract

It is almost a truism these days that every cancer is different, and that the best treatment for a given cancer is likely to be something tailored to that particular tumor. The expectation is that the best treatment will be defined after subjecting tumor tissue to a battery of genomic, epigenomic, proteomic, metabolomic, collectively called "omic" assays, and then making a decision based on the results. However, there are many formidable obstacles to implementing this idea widely, and statistics plays a key role in overcoming several of them. Here are two. How do we find out which "omic" differences are clinically important, and which are just "variation". Having done that, how do discover which drugs and doses are best suited to treating tumors exhibiting particular clinically important "omic" characteristics? I will discuss some of the analysis challenges in this area.