Pooja Sameet Agarwal (2027) | Molecular and Cell Biology, Computer Science
For her SURF project, Pooja used deep learning tools to better understand genetic variants connected to Autism Spectrum Disorder (ASD). By studying how these variants may affect important gene-regulating proteins, her research helps shed light on mutations that are still not well understood. This work could support a better understanding and diagnosis of ASD in the future.
How did your research question or process change over the course of the summer?
My project pivoted to become an independent model interpretability project investigating the reliability of model results. In order to be able to use the deep learning models to interpret the effect of ASD variants, we needed to understand whether the models were reliable in the first place and how they were making their predictions. This was what started my summer project, where I explored how we could use model interpretation and cross-validation with previous experimental studies to figure out why the models predicted what they did and whether these predictions were accurate at all.
What challenges did you face, and how did you overcome them?
There was definitely a steep learning curve in being able to apply tools to interpret models, since this wasn’t an area I had worked in before. I worked in the Staller Lab last summer with Professor Staller and Sanjana, a junior specialist who helped mentor me. My lab mentors were very supportive in helping me ideate and design my project. Because I found the work so exciting, I would spend extra hours and weeks reading through the original papers and debugging my code, and it was incredibly fulfilling to see the final output of my work.
What was one of the most exciting aspects of your research project that you were able to pursue?
I had long been interested in model interpretability, and I was excited that I could naturally explore this area as part of my work. It was also really fun to present my work at the end of the summer and again at a conference in the fall.
Did your research entail any travel? Fieldwork? Use of new technology, novel equipment or methodologies?
My research involved learning and using model interpretability tools and methods that were new to me. In the fall, I applied and was invited to present my poster at the Cell Symposium Precision Genomics conference. My work related to variance of uncertain significance, an important issue in genomics and precision genomics, especially in applying machine learning to interpret these VUS. It was a really great opportunity to meet more researchers and learn about more research in the area, and I got to explain my work to a lot of different people who came from different research backgrounds but found value in seeing how these models could be interpreted and understanding that they weren’t always reliable. I recently found out that I’ll also be presenting this work at the Schmidt Center Symposium on Biomedical Science and AI this coming April, which I’m excited about!
What skills did you develop that you didn’t have before?
I worked more independently over the summer than I had before, and this allowed me to learn a lot more about how to design a research question, how to find and verify methods to solve it, and how to iterate based on my results. It made me much more detail-oriented in my research and taught me to continuously ask questions based on the data I get.
Have you had any opportunities emerge for you as a result of your summer research?
I’m still working with the Staller Lab, and I’m now working on a project collaborating with the Nayak Lab. Researchers in the Nayak Lab identified a protein that controls how genes are turned on and off in archaea (micro-organisms). I’m working to use deep learning tools as well as traditional tools to track down this same protein in other species, helping us map out how these tiny organisms function. This has been a really cool project because I’m working on it completely independently and it’s very different from most of the work the Staller Lab does, so it’s been a great opportunity to learn more and build my skills.
Are you thinking about or currently applying to graduate school or scholarships to support future research?
I plan on applying to grad school this fall.
How did participating in the SURF (specifically the programming—the research cluster meetings and workshops) support your research process or feelings of preparedness?
SURF enabled me to feel more confident as an independent researcher. It guided my thought process and gave me knowledge I wouldn’t have gotten had I not been able to do research over the summer.
What was it like working with your faculty mentor/research advisor?
My faculty mentor, Professor Staller, is an incredibly supportive PI. I met with him every week over the summer, and he would give me detailed guidance on where to take my project and overall advice on how to think about problems. He gives me both independence and support in my research, allowing me to learn new skills very quickly.
How did being part of this program support or shape your research experience?
This program has reaffirmed my passion for research and confirmed that I want to pursue a career in research. I’m still in the lab I worked in during SURF, and I’ve been able to build on my work and start new projects that I wouldn’t have been able to do without SURF. I also really appreciated the ability to work in my lab over the summer and build on the work I had been doing during the school year, giving me uninterrupted time to work on really cool projects.
What resources or opportunities provided by the program were most valuable to you?
The workshops were a really cool way to get advice and learn new topics. I went to a workshop on autoencoders, which was something I didn’t previously have experience with but am now using in a research project I’m working on. Additionally, meeting the SURF cohort and learning about their research was a great way to meet new people who are also passionate about research and who I normally might not have had the opportunity to interact with.