Sequencing the human genome gave scientists a treasure trove of genetic information to mine as they search for the cause of life-altering genetic diseases. But it soon became clear that only a very few diseases could be tied to a single gene. Understanding the genetic basis of disease requires understanding the intertwined effect of many different genes
This is where Hertz Fellow Hilary Finucane comes in. During her Hertz Fellowship, Finucane developed a set of mathematical tools to more rapidly and easily uncover relationships in large sets of genetic data. After graduating with her PhD in applied mathematics from MIT, Finucane started her own research group at the Broad Institute of Harvard and MIT, a group which continues to find new ways to unlock the relationship between genes and diseases.
Though she majored in mathematics as an undergraduate at Harvard University, Finucane found herself drawn to computational biology as she finished her M.Sc. studies in Computer Science at the Weizmann Institute in Israel. “I really enjoyed the theoretical work, but I wanted to do something where I could see more clearly what the impact would be,” she says.
When Finucane started her PhD and Hertz Fellowship in 2013, then, she knew she wanted to bring her mathematical expertise to bear on open problems in biology, and statistical genetics was a field ripe for innovation.
During her Fellowship, Finucane pioneered a pair of tools, stratified LD score regression and cross-trait LD score regression, which addressed two linked problems in genetics. Many traits are tied to variations at many different places on the genome, each only weakly correlated with the trait of interest. Finding the most important variations in the genome can drive understanding and treatment of the disease – since genes affect disease, but not vice versa, Finucane describes them as “golden clues” where the causal relationship is clear – but in these so-called “polygenic” cases, the correlations are especially weak and noisy.
The LD score regression tools, however, look for connections between genetic variants, so are actually more powerful the more polygenic a trait is, says Finucane. Even better, the LD score tools are based not on individual patients’ genetic sequences – which can be difficult to obtain, often for privacy reasons – but on the “summary statistics” that other studies of these traits report. This makes for a much faster analysis, which can be performed with more widely available data, giving more rapid insight into disease.
Now a Schmidt Fellow at the Broad Institute, Finucane makes use not only of the mathematical tools she developed during her Fellowship, but also of the intellectual flexibility she learned in entering a new field as a PhD student.
“As I made that transition, I didn’t just have to learn biology, but also statistics and machine learning and how to program from scratch. But I was in a really supportive community where a lot of people helped me pick up the skills I needed,” says Finucane. This community included the tight-knit, interdisciplinary research group working with her PhD advisor, Alkes Price; her husband, Yakir Reshef, with whom she has been publishing research papers since 2011; and the collection of Hertz fellows she met during and since her PhD.
Among her collaborators is Hertz Fellow Jesse Engreitz, whose lab at the Broad Institute is also working on mapping the connections among genes and between genes and disease, but Finucane says the value of the Hertz Community goes beyond individual collaborations. “Going to the retreats is such a great way to learn about a lot of interdisciplinary work that’s going on both within and outside of academia,” she says.
Finucane applies that interdisciplinary spirit to developing her own lab, she says, as she works to teach biologists to think like mathematicians and mathematicians to think like biologists. “Since I’ve been through it I can think more about what it is the people in my lab need to do their best work,” she says.