We live in the information age, as we are constantly reminded, but in many areas our ability to collect data has outstripped our ability to make sense of it. Brendan Frey is working to change that, with a particular focus on the complex worlds of biology and genomics.
Dr. Frey specializes in developing mathematical frameworks and computer algorithms that can tease out important patterns in masses of data, research that he will take to the next level during his 2009 NSERC E.W.R. Steacie Memorial Fellowship.
The human brain does a great job of this kind of analysis. For example, our eyes may scan the scene in front of us and instantly pick out relevant information and ignore the rest. While cycling, a rock on the road or an object moving our way is assessed on the fly, allowing us to react.
Dr. Frey is currently studying ways in which the relatively small number of human genes can control complex processes, such as the formation of intricate neural structures in the human brain. While there are only 22,000 human genes, a process called alternative splicing expands the number of possible gene instructions to an astonishing one million. Working closely with other biologists and medical researchers, his goal is to train a computer to probe information from the genome and from normal and diseased tissue samples, in order to build a model of how the cell works and identify patterns that may indicate medically significant genetic mutations and interactions.
There’s more to it than mechanically running through all the possibilities and spitting out an answer. Humans use tricks to analyze problems and Dr. Frey's computer algorithms make use of similar tricks. It’s also not just any pattern that will yield useful information, just like it is less important for the brain to pay attention to the uniformity of a pathway than to the rock that could lead to a twisted ankle.
“What you really want to find are the unusual patterns—the patterns that are striking,” he observes. “There might be a tiny pattern in the DNA, maybe one mutation in this sea of data, which could be responsible for a gene not functioning properly. That is part of the problem. The algorithms have to be really, really good to find all the tiny patterns that could be important.”
Finding a pattern still doesn’t mean the job is done, since there is always some uncertainty about its significance. “Science is supposed to be predictive,” says Dr. Frey. “But with biology, because the system is so complex, it is very difficult to come up with predictive methods. However, that is my goal.”
Since Dr. Frey’s algorithms can’t say with absolute certainty when they have spotted a meaningful pattern, he turns to probability—basically calculating the odds that a particular pattern is associated with the specific question he is trying to answer. In the end, his model provides a list of interesting patterns, along with the likelihood of each one being the culprit in a certain condition. Then it’s back to the laboratory to test how each possibility pans out in real life.
The method first bore fruit in 2005, when an analysis conducted by Dr. Frey’s team led to the identification of a genetic mutation that causes Barraquer-Simons syndrome, a rare disease that leaves teenage children unable to retain beneficial fats.
As the science of genetics develops, analytical tools like the ones Dr. Frey has developed could eventually also be used to predict the effect of deliberately changing genetic information in an attempt to solve a problem.
Along the way, he’ll continue to work to improve the algorithms that are the tools of his trade. He is world-renowned for his contributions to techniques for analyzing probability models that deal with millions of variables. The benefits of his discoveries extend well beyond genomics–they can help in any situation that involves huge volumes of data. The list of possibilities runs the gamut from analyzing flight patterns and digital communications to documents. Dr. Frey would also like to know how the methods he studies could be used to explain how the human visual system works.
Dr. Frey is interested in both engineering solutions to problems and answering scientific questions, and that leads to interesting interactions in his research. As he puts it, “It is delightfully ironic that we are engineering biologically-inspired algorithms that enable us to answer fundamental questions about biology.”