Irene Kaplow is a Ph.D. student in the Fraser and Kundaje Labs. She has a B.S. in mathematics with a minor in biology from MIT, where she did research on analysis of RNAi experiments and use of conservation to identify exonic splicing enhancers. She now uses machine learning to study regulatory genomics.
This content has been transcribed from an interview that took place on Stanford campus Wednesday, November 4, 2015 with CEHG’s Director of Programs, Cody Montana Sam and Communications Manager, Katie M. Kanagawa.
Can you start by telling us a bit about yourself, personally and professionally?
I’ve been doing computational biology since summer after freshman year of college, and I’ve been interested in gene regulation since freshman year of high school.
I still remember being in ninth grade biology class when my teacher drew a piece of DNA on the board; she then made a box and saying, “this is a gene,” made another box to the left of the gene saying, “this is a promoter,” and labeled the rest as “junk DNA” [laughs]. And a student raised his or her hand and said, “I take it the human genome is mostly genes and promoters,” and the teacher said, “No, the majority of the human genome is junk DNA.” The thought went through my head, it’s probably not all junk, I want to figure out what it’s doing.
Although I did not know much about evolution at the time, I could not understand how it could be evolutionarily advantageous to produce billions of base pairs of DNA in every cell that do not have any function. And that curiosity stayed with me, no matter how much I tried to consider working on other questions. Hence, I ended up going into genomics with a specific interest in understanding the role of non-protein-coding regulatory regions of the genome.
That was my freshman year of high school. Then, freshman year at MIT, I had heard about something called computational biology, but I didn’t really know what it was. I emailed ten compbio professors, and a couple of them offered to let me work in their labs. I ended up working with Bonnie Berger that summer and liked the research so much that I stayed in the lab for three years. Her lab worked on a wide range of topics, from protein structure prediction to epidemiology to genomics, and I worked on genomics projects. While I thoroughly enjoyed my undergraduate research, my main projects in the Berger Lab were on protein-coding regions, leaving me still curious about the role of non-coding DNA. Therefore, when I came to graduate school, I decided to focus on aspects of transcriptional regulation involving non-coding regions.
What do you think it is about genomics that draws you in and keeps your interest over time?
I think there are many interesting transcriptional regulatory mechanisms that we don’t understand, and we are getting unprecedented amounts and types of data that give us the tools to understand them. For example, we know that transcription factor binding or DNA methylation near a gene can activate or repress the gene, but ”near” can mean next to the transcription start site or close to the TSS in three dimensions but tens of thousands of bases away. Assays that limit us to studying a few locations of interest, where locations near TSS’s are usually selected, therefore might miss some of the most important transcription factor binding or methylation events. High-throughput sequencing data enables us to study regulatory mechanisms genome-wide across multiple species, individuals, or cell types and observe how these mechanisms change in relation to when the gene is transcribed. This data is therefore allowing us to answer new questions about the roles of events that happen on non-coding DNA in transcriptional regulation.
Can you describe your current research?
Earlier in my Ph.D., I worked on a novel approach to studying DNA methylation genome-wide across many individuals, which enabled us to propose potential mechanisms about how some non-coding variants affect expression through their effects on methylation. My current research involves predicting and understanding the binding of a transcription factor called CTCF. CTCF is involved in holding DNA loops together, which can bring transcription factors to genes and also block transcription factors from accessing genes, thereby playing a key role in determining how genes are regulated. Thus, we can study some of the mechanisms behind gene expression variation across cell types by understanding how CTCF binding varies. However, our current understanding of the relationship between DNA sequence and CTCF binding is incomplete. We have therefore developed a deep neural networks approach to predicting CTCF binding, and our approach is providing unprecedentedly accurate predictions. We have also used a similar approach to predict which subset of CTCF binding sites are at the bases of DNA loops. We are now exploring what sequence patterns our models are learning.
What is next for you? What is your timeline for graduation?
My hope is to graduate in the fall and apply for postdocs in computational genomics. I would like to someday have my own lab where we could work on regulatory genomics questions from a computational perspective.
Is there one person to whom you would attribute your educational and scientific success?
There’s so many people; it’s really hard to select only one person. I owe a lot to Bonnie Berger, who was my undergrad research supervisor, because I didn’t even know computational biology existed before I got to know her, and she told me about the field and gave me the opportunity to work in her lab, which had so many talented people. She also gave me multiple choices of projects (which is unusual for an undergrad), providing me with the opportunity to think about what I was most interested in from an early stage.
I also owe a lot to my co-advisors, Hunter [Fraser] and Anshul [Kundaje], whose thoughtfulness and creative insights about a wide range of problems have helped me grow so much as a scientist.
Finally, I’ve had a number of labmates, in Bonnie’s, Daphne’s, Hunter’s, and Anshul’s labs, who have given me great suggestions that have sometimes transformed the way I think about a project. I think having wonderful labmates throughout my career, even when things were up in the air with advisers, has helped keep me excited about what I do.
Going back in time a bit, can you describe your childhood and how you first became interested in science?
My parents were always supportive of me in pursuing math and science. As a child, my dad would sometimes do science experiments with me after school, and my parents got me science kits. I would also do math problems with my dad on the weekends. In addition, my parents encouraged me to sign up for a computer programming class freshman year of high school, which I really enjoyed, even though I was the only girl in my grade in the class. I feel like I got a lot of encouragement from family going back as far as I can remember to pursue math and science, which is really rare in our society, especially for girls. And I always had teachers who really encouraged me to go after whatever I found most interesting and challenged me to think rigorously and creatively.
Would you give any specific advice to grad students, postdocs, and undergrads interested in this field?
For undergrads, I would say make sure you get a position where you’re working on a real research question and not just doing busy-work that no one else in the lab wants to do. Get your hands dirty with some of the data analysis, and do not be afraid to ask questions and propose your own ideas.
For all students, work with a professor who makes you think outside the box. It makes coming to lab every day really exciting, regardless of how well your research is working at the moment. In addition, working with such creative advisers has enabled me to work on novel, important projects that others in the field did not think to pursue and helped me learn to come up with my own new ideas.
You also want to work with faculty and labmates who give you substantial feedback on what you’re working on and regularly solicit your feedback. One of the great things about working in a collaborative lab is that you both get detailed suggestions and you get to give others ideas. When your project is going slowly, you both get help from others in turning it around and get to have fun helping others make progress in their research.