It sounds like some sort of Frankenstein sequel: Far away in the United Kingdom, there are ancient lungs dating back to the 1700s that are suspended for eternity in jars of ethanol. These lungs are part of the collection of medical museums in places such as Glasgow, Scotland, and a few years ago — around the time AI came onto the scene — Daniel Blanco-Melo, PhD, visited the museums, hobnobbed with the historians and took tiny samples to bring back to his virology and immunology lab at Fred Hutch Cancer Center.
Erin Barnett is a PhD student in Blanco-Melo’s lab. Her thesis project focuses on ancient RNA. She’s working with those human lung samples from centuries past to isolate and identify pathogens, which can have RNA or DNA genomes, so she can better understand what infectious diseases circulated and contributed to respiratory disease hundreds of years ago.
Barnett has been able to identify both bacterial and RNA viral pathogens in those lungs and is particularly focused on a human rhinovirus from a lung dating to sometime between 1740 and 1783, the span during which it was preserved.
“What's really cool is that now we have a 250ish-year-old viral genome in the computer that we have reconstructed, and we can compare it to modern rhinoviruses that are circulating because we have a unique snapshot from the past to compare it to,” said Barnett.
Here’s where AI enters the picture: One viral genome is not enough. Barnett bemoans the "diversity of data missing between the one genome I isolated and the ones we isolated from COVID-19 samples.” So she’s leveraging AlphaFold3, an AI-based structural prediction tool that looks at amino acid sequences and predicts protein structure. Protein structure is useful because these viruses mutate slower on a protein level, preserving their function. This concept, broadly known as structural phylogenetics, may help shed light upon viral evolution when considering historical viral sequences that were lost in time such as this 250-year-old rhinovirus.
“Ancient RNA is highly degraded, so AI can help build these structures to generate enough data to perform this analysis,” she said. “AI basically simplifies the evolutionary analysis.”
As AI becomes more entrenched in society, scientists are increasingly exploring ways that the technology can enhance their work and make the way they approach both research and patient care more efficient. At the Translational Data Science Integrated Research Center’s annual retreat, a panel of experts discussed how AI can impact cancer diagnosis and treatment by helping aggregate raw data and streamline data science initiatives.
Those efforts are already bearing fruit. In its first year, for example, the Cancer AI Alliance (CAIA) created a multi-cloud, multi-institution platform that launched eight pilot-use projects. One of those projects is focused on predicting skeletal impacts across cancer types for patients with bone metastases.
But the highlight of the daylong retreat was a rapid-fire session that felt like a cross between speed dating and “Shark Tank”, the reality show where aspiring entrepreneurs pitch their products in hopes of securing start-up funding. Fast Pitch AI featured graduate students and postdocs presenting a three-minute pitch about their project and how applying AI would enhance the outcomes. Drug codependency mapping using AI to identify cancer drugs with mitochondrial off-targets? Harnessing AI to better understand pathways of DNA repair? AI-driven personalized near-term breast cancer risk prediction using longitudinal MRI? Yes, yes and yes.
Barnett joined David Glass, Sarah Huang and Lucas Liu as one of four winners who earned $10,000 each from Fred Hutch to put toward advancing their project. Six additional winners were awarded $5,000 each. The competition is intended to get trainees comfortable showcasing their research and proposing a project.
“We want to give postdocs and grad students an opportunity to share out their research in a succinct and engaging way,” said Melissa Alvendia, director of programs and strategy for the Translational Data Science Integrated Research Center. “Pitching it is very different from doing a research presentation. Developing that skill is really the purpose of the Fast Pitch approach.”
Students were encouraged to work with the Fred Hutch philanthropy team on their pitch to help them home in on their message in just one slide and three minutes.
Glass’ pitch focused on vaccine response. Working with the HIV Vaccine Trials Network, Glass is trying to use AI to understand which specific B cells deliver a durable immune response.
Scientists have found that about a year after the HIV vaccine is administered, some cells have a specific “memory" of the vaccine; that is, they remember their encounter with the pathogen and they limit reinfection. Exactly what are the cells that appear right after the vaccine is administered and how do they persist over time?
“That’s the big biological question we’re trying to ask in this study,” said Glass, a postdoctoral fellow in Evan Newell’s lab.
There are two types of cells that have this long-lived memory: antibody-secreting cells and memory B cells. Memory B cells keep the antibody on their cell surface, which makes them easy to identify. But the antibody-secreting cells live up to their name; they secrete the antibody, so Glass has proposed a method to pinpoint the antigen-specific HIV antibody-secreting cells.
You will receive a monthly newsletter from us about our innovative research and compassionate care as well as updates on events and other helpful tools and tips.
You may unsubscribe at any time.
“The goal is to pull those out so we can look at their cell identity,” he said. "Now, we don’t know which B cells we should be trying to induce. We want to understand which B cells' identity is optimal for a durable response so we can better tailor vaccines to select for that type of B cell."
This is Glass’ first project where he is interactively coding with AI.
“Instead of making a plot showing the composition of cells at different timepoints, I tell Claude what I want to accomplish and I edit it, make decisions about the results and decide what to do next. It is an incredibly effective autocomplete for me. I can offload more of the coding mindset to AI so I can spend more time thinking like a scientist.”
Liu, a statistical programmer and postdoctoral researcher in Ruth Etzioni’s lab and Michael Haffner’s lab, is trying to determine if it’s possible to use AI to help treat prostate cancer. Etzioni holds the Rosalie and Harold Rea Brown Endowed Chair.
“We’re at the very beginning stage of using AI to help guide cancer treatment,” he said.
Liu’s work starts by considering pathology images from tumor biopsy slides. Physicians examine pathology images of tumor specimens to determine if patients have cancer and the grade of the cancer. But to determine whether a patient might benefit from certain treatments including immunotherapy, providers currently need to send the patients’ tumor tissue for molecular testing to look for biomarkers such as MSI-high (microsatellite instability-high) alterations, which are associated with treatment response.
This process is expensive, considering only 3% of prostate cancer patients have MSI-H tumors, which struggle to repair DNA errors. Patients with MSI-high tumors have a great response rate to immunotherapy.
“It’s too expensive to do that testing for every patient for only about 3% who have MSI-high tumors.” Liu said. “Especially in low-resourced settings, no one wants to do that testing. But without it, we may miss patients who could benefit from the treatment.”
Liu has built an AI tool to predict an MSI-high profile from a pathology image for prostate cancer. Working with researchers at Fred Hutch and other institutions, he has tested the tool on multiple datasets and found its performance “promising.”
“We need to continue to collect data to do further validation and move it closer to clinical use,” he said. “The next step is to see if AI can directly predict treatment response to immunotherapy. Our goal is to use AI to make precision oncology more scalable, accessible, and equitable."
Huang, a graduate student in Manu Setty’s lab, relied on her computational background to pitch a plan that uses AI to study head and neck squamous cell carcinoma recurrence. The disease is particularly challenging because it is often diagnosed at an advanced stage. The first-line treatment is surgery, including removal of a margin of tissue around the tumor to reduce the likelihood that cancer cells remain.
Yet up to 50% of patients experience recurrence within a year. Why does the disease relapse so quickly if the tumor has been removed and no residual tumor is left? A previous study that Huang was involved in found an emergent cell population in tissue adjacent to the tumor that is not quite normal and not quite tumor.
“You can think of those cells as being in a state of limbo,” she said.
Huang hypothesizes that signaling from the tumor may have changed nearby cells, pushing them into this in-between state and possibly contributing to the cancer’s return.
Huang is using the power of AI to characterize these “limbo” cells, examining whether the way they communicate with nearby cells is linked to recurrence and whether they have a recognizable appearance under the microscope that could eventually help surgeons more precisely identify high-risk tissue.
The full list of 2026 Fast Pitch AI competition winners:
Mechanisms, Functional Genomics, and Disease Biology
Translational Data Science Integrated Research Center Attendee Favorite ($10,000): Erin Barnett
Internal Advisory Board Favorite ($10,000): David Glass
Best Science/Innovation ($5,000): Sarah Becker
Best Pitch/Scientific Communication ($5,000): Betty Yu
Best Translational Impact ($5,000): David Sokolov
AI, Computational Modeling, and Clinical Translation
Translational Data Science Integrated Research Center Attendee Favorite ($10,000): Lucas Liu
Internal Advisory Board Favorite ($10,000): Sarah Huang
Best Science/Innovation ($5,000): Isabelle DuPlessis
Best Pitch/Scientific Communication ($5,000): Weston Hanson
Best Translational Impact ($5,000): Patrick McDeed
Bonnie Rochman is a senior editor and writer at Fred Hutch Cancer Center. A former health and parenting writer for Time, she has written a popular science book about genetics, "The Gene Machine: How Genetic Technologies Are Changing the Way We Have Kids—and the Kids We Have." Reach her at brochman@fredhutch.org.
Are you interested in reprinting or republishing this story? Be our guest! We want to help connect people with the information they need. We just ask that you link back to the original article, preserve the author’s byline and refrain from making edits that alter the original context. Questions? Email us at communications@fredhutch.org
Every dollar counts. Please support lifesaving research today.