Diseases like Alzheimer's or cancer don't happen overnight; they are the result of many years of hidden cellular changes. Yet most of what we know about human disease at the cellular level comes from studying tissue samples collected at advanced stages of illness.
University of Washington researcher Kevin Lin is working to reconstruct the hidden history of a cell, from healthy to diseased, using a single tissue snapshot. With a $2 million award from the National Institute of General Medical Sciences (NIGMS), he aims to develop a suite of AI-driven tools that can build a “computational time machine” for human biology that will help researchers identify therapeutic targets at the earliest stages of disease, long before patients experience symptoms.
“Studying the human brain or heart is tricky because they are 'longitudinally inaccessible' – you can't exactly take repeated samples from a living person's brain to watch a disease progress,” said Lin, an assistant professor of biostatistics in the School of Public Health.
By using biostatistical expertise in longitudinal modeling, along with fundamental concepts in cell biology where certain genetic changes leave clues about what happened before, Lin hopes his team’s models can infer what those cells were many years prior, an innovation he refers to as “peering backwards.”
“The analogy I like to give is akin to a detective story: we arrive at the 'crime scene' of what happened to the cells after a disease has reached a terminal stage, and we and our collaborators use our biological and statistical skills to infer how these cells arrived at their current state,” said Lin.
UW PhD and master’s students working with Lin will have the chance to gain hands-on experience in what he dubs the “frontier of AI and biology.”
“They’ll learn to adjust for the messy reality of human data, ensuring that AI discoveries are biologically grounded and not just statistical noise,” said Lin.
Lin emphasized that the project is part of a larger, interdisciplinary push to equip future biostatisticians with tools to tackle the most difficult questions in medicine.
“It is a phenomenal time to be a biostatistician,” said Lin. “We are moving beyond simply observing data to hypothesizing the underlying mechanisms driving the data over time. When we work closely with our clinical and wet-bench collaborators, we can validate our AI predictions using animal models and cell lines and turn 'crazy' computational ideas into verified biological insights.”
