MS in Applied Data Science Capstone Research Published In Cancers

For many graduate students, a capstone project marks the culmination of months of learning. For one team of students in the University of Chicago’s MS in Applied Data Science (MS-ADS) program, it became something even more: peer-reviewed published research in Cancers.

The team’s manuscript, Benchmarking Open-Source Pathology Foundation Models for Breast Cancer, grew out of a research capstone project completed under the mentorship of Associate Clinical Professor Utku Pamuksuz at the Data Science Institute and in collaboration with researchers at UChicago Medicine. 

For Jiayou Liang, one of the student researchers, the project offered an opportunity to apply advanced machine learning techniques to a real-world healthcare challenge while working alongside experts from multiple disciplines.

Choosing a Research Problem with Real Impact

Students in the MS-ADS research capstone work closely with faculty mentors and research partners on open-ended problems drawn from academia and industry. When Liang’s team heard presentations from several potential collaborators, one project immediately stood out.

The project centered on pathology foundation models, AI systems trained to analyze digital pathology images. Liang and his teammates evaluated two leading open-source models, TITAN and CHIEF, to better understand how they performed in predicting breast cancer biomarkers. Building on those findings, the team developed new techniques to improve the models’ performance, research that ultimately became the foundation of their published manuscript.

Bringing Data Science and Medicine Together

One of the defining aspects of the project was the collaboration between data scientists and medical researchers. 

Liang explained, “The medical researchers weren’t always familiar with the latest AI techniques, and we were completely new to the medical side. The medical researchers gave us detailed explanations of the pathology images and datasets, while we shared our expertise in machine learning. By the end, everything came together.”

Faculty guidance also played an important role as the team explored different technical approaches before arriving at the methodology that ultimately formed the foundation of the published work.

Research That Could Help Accelerate Patient Care

Breast cancer biomarkers help physicians determine the most effective treatment for individual patients. Traditional laboratory testing can take weeks, delaying clinical decision-making.

One of the most common tools pathologists use to examine cancer tissue is an H&E slide, which uses special stains to make cells and tissue structures easier to see under a microscope. Once these slides are digitized, AI models can analyze the images for patterns that may not be readily visible to the human eye.

Liang believes AI has the potential to significantly shorten the timeline for identifying important biomarkers.

“Breast cancer biomarkers are crucial for helping doctors decide on treatment, but today that information can take weeks to come back,” Liang said. “We’re adapting pathology foundation models to read these signals directly from the H&E slide. Once the slide is digitized, the prediction takes seconds, the remaining bottleneck is tissue processing, not the analysis. If the accuracy holds up in prospective validation, that could compress a multi-week wait down to about a day.”

From Student Project to Published Research

As the capstone progressed, the team’s findings proved significant enough to develop into a manuscript submitted for peer review.

When asked how he felt after learning the paper had been accepted for publication, Liang reflected on how the project continues to evolve.

He says, “I was very surprised when the paper was accepted. But in many ways, this is just the beginning of our project. We wanted to understand how foundation models perform on these classification problems, so we evaluated both the TITAN and CHIEF models, analyzed the results, and published our findings. We’re also continuing to build on the machine learning pipeline we developed during our capstone, and I hope that work will lead to another paper in the future.” 

Preparing for What’s Next

The experience also reinforced Liang’s long-term goal of pursuing a PhD while teaching him how to lead a collaborative research effort across disciplines.

“I’ve always wanted to pursue a PhD, and this project really helped prepare me for that,” Liang said. “It taught me how to lead a research project and communicate with people from different disciplines. We worked with experts on the technical side and the medical side, so learning how to bring everyone together and collaborate was incredibly valuable. It also taught me how to develop an idea, test it, evaluate the results, refine the approach, and ultimately turn it into something meaningful. I think that experience has been really helpful.”

His advice for future MS-ADS students interested in pursuing the research capstone is simple:

“Be patient,” Liang said. “Research takes time, and finding the right direction isn’t always easy. You have to understand the data, evaluate different methodologies, and figure out the mathematics behind the problem. If you’re interested in the research capstone, be patient with the process. Talk with your faculty advisor, your teammates, and your research collaborators. By working together and staying persistent, you’ll get there in the end.”

The capstone project marks the end of the MS in Applied Data Science graduate program. For Liang and his team, the capstone project became the foundation for research that will now reach the broader scientific community.