Jingyuan
Zhang

Asthma is a familiar word, but it covers patients whose disease is driven by different biology. Two of those patterns are called th2-high and th2-low. If we can tell them apart reliably for an individual patient — their endotype — that distinction can help frame better questions about pathways and care. That's the problem I'm working on in the Atasoy Lab at Michigan Medicine.

This summer, with support from the EHAIL Student Summer Support Program and in collaboration with the University of Michigan Bioinformatics Department, I'm developing a machine learning model to analyze th2-high and th2-low asthma and predict endotypes in patients. The model is only part of the goal. I'm also looking for the key pathways behind the classification and for gene-gene interactions that might be driving it — the places where genes don't act alone.

That's why interpretability matters so much to me. A model that outputs a label without reasons isn't very useful in medicine. A patient-level model needs to be something a researcher or clinician can interrogate: which signals pushed it one way, which pathways are involved, and where it might be uncertain. Building that discipline into the work from the start — rather than bolting an explanation on at the end — is one of the biggest things I'm learning.

I'm a data science student, and I came to this lab with a strong interest in both bioinformatics and biological sciences. This project sits right at that intersection. The statistics have to be sound, but they also have to respect the biology underneath the data. I'm still learning how to hold both at once, and I'm grateful to be learning it on a real healthcare problem, with real data, alongside people who do this work every day.

In the summer of 2024, before I started at Michigan, I was a research intern in the Delgoffe Lab at the UPMC Hillman Cancer Center in Pittsburgh. My placement was facilitated through the UPMC Hillman Academy program, and it was my first time inside a working cancer research lab.

The question I worked on sounds simple: what effect does glucose have on CD8+ T cell function and vitality? The context is what makes it matter. In adoptive cell therapy, T cells are part of the treatment itself, so understanding the conditions that affect how they function and stay vital is a serious question, not a classroom exercise. I learned quickly that a simple question still takes careful, methodical work to examine well.

A lot of what I took from that summer wasn't a single technique. It was the rhythm of a lab: planning an experiment before you touch anything, keeping a clear record, handling cells carefully, paying attention to controls, and being honest about what you did and didn't see on a given day. Good lab work is patient work. You show up, you follow the protocol, and you learn to tell the difference between a result and a wish.

I was a high school researcher in a lab full of people doing this at a much higher level, and they treated the work — and the learning — seriously. That stuck with me. It's a big part of why I chose to study data science and biophysics, and why I want to keep working where biological questions and careful analysis meet.

Before Michigan, one of the most important parts of my week at Bethel Park High School was band. I played clarinet in the marching band and in Symphonic Band, and I served as the clarinet section leader. It was the first leadership role I really had to earn day after day, not just hold as a title.

A section leader's job is mostly unglamorous, and that's the lesson. You learn your part cold so you can help someone else learn theirs. You listen for balance, not just for your own line. You run the sectional, you notice who is struggling, and you show up early and prepared because the section takes its cue from whether you did. Practice isn't a motivational idea in band — it's just what you did yesterday, and what you'll do again today.

I recognize that same pattern in research now. Most progress doesn't come from one big moment. It comes from careful repetition, from checking your work, and from being part of a group where everyone is trying to get a shared thing right. Band taught me that before any lab did.

The rest of life matters too. Outside the lab I still love making time for magic, for Minecraft, and for good food — the things that have nothing to do with a model or a protocol and everything to do with staying curious and enjoying being with people. I don't think that time is separate from the work. It's part of what makes the work sustainable.