John Reinitz, PhD, Professor of Statistics, Ecology and Evolution, Molecular Genetics & Cell Biology at the University of Chicago, died on January 23, 2025, after a battle with metastatic cancer. He was 66 years old.
In the early 1990s as a “struggling scientist,” Reinitz pioneered an approach to developmental genetics that foreshadowed broader developments in biology and beyond. This approach, which he honed in collaboration with David Sharp, a nuclear physicist at the Los Alamos National Laboratories, and Eric Mjolsness, a computer scientist at Yale, was characterized by two approaches.
Genetic research in developmental biology at the time consisted almost entirely of qualitative, hypothesis-driven experiments. Reinitz’s approach relied instead on constructing an atlas—a comprehensive, spatially and temporally resolved, quantitative dataset—with no explicit hypothesis guiding the effort. Building spatiotemporal atlases has now become routine in developmental biology, a practice foreshadowed by Reinitz’s groundbreaking innovations.
Theoretical biology, especially in development, also tended to rely on qualitative phenomenological models. Reinitz’s new approach was based on data-driven modeling instead—using the atlas data to train predictive models of gene regulation. Countless large quantitative datasets, such as the Cancer Genome Atlas and the Tabula Sapiens atlas, are now being built by huge global consortia. What Reinitz pioneered at a small scale thirty years ago is now practiced across disciplines in an industrial manner.
“John was a fiercely independent thinker. He went where the science brought him—without caring about the opinions of others—and attempted to crack problems that seemed insurmountable. You could always rely on him for a dissenting voice, a quirky remark, or a different perspective,” said Stefano Allesina, Professor and Chair of the Department of Ecology and Evolution at UChicago.
An independent beginning
Reinitz began his research career as an independent PhD student at Yale University, loosely supervised by Juozas Vaišnys. While there, he devised a mathematical model of viral gene regulation that incorporated all the mechanisms known from experimental evidence at the time. This model revealed that the known mechanisms were insufficient to account for all the virus’s behaviors, which dismayed the experimentalists who had heretofore regarded it as a solved problem.
This was a time when mathematical modeling was not only outside the mainstream but not considered relevant to biological discovery. This did not deter Reinitz, however; Sticking to his guns to contradict the conventional wisdom of a field was a recurring theme in John’s career.
After a postdoctoral stint in Mike Levine’s lab at Columbia University to learn Drosophila developmental genetics, Reinitz initiated the project that would evolve into the data-driven modeling approach that defined his career. Together with Sharp and Mjolsness, he devised the connectionist model of development that, while being biophysically motivated, was mathematically equivalent to a Hopfield neural network. The model could thus be used to learn or infer the structure of a gene network by fitting to data.
Reinitz’s approach required many technologies that were yet to be invented. He raised a panel of antibodies to the segmentation proteins that he distributed to the Drosophila community for decades and is still in use today. He also developed bespoke optimization algorithms for fitting the models, devised protocols for quantitative fluorescence microscopy, and invented approaches for automated image segmentation that enabled the construction of the atlas at scale. He led this massive multidisciplinary effort largely in house by recruiting a diverse group of biologists, physicists, and computer scientists, most notable among them a group of scientists in St. Petersburg, Russia, whose work Reinitz funded for two decades via his National Institutes of Health (NIH) grant.
His efforts bore fruit after he moved to Stony Brook University in 2000. The spatiotemporal atlas revealed that gene expression domains were not static as assumed but shifted along the axis of the embryo as development progressed. This challenged the static French-Flag model of patterning prevalent at the time. With the aid of the connectionist model, Reinitz and his lab showed that this motion was driven by asymmetric gene regulation, proving that embryos do not passively receive information from the mother, but actively self-organize to determine their body plan.
While these developments were underway, Reinitz started working on the next major model. He leveraged advances in bioinformatics to develop a model for predicting gene expression from DNA sequences that has been highly successful in explaining gene regulation, including that of evolutionarily diverged sequences. This framework was accepted much more readily by the community and is now used by many erstwhile purely experimental labs.
While there wasn’t a name for such work when Reinitz started, data-driven gene network inference and modeling came to be known as “systems biology” by the early 2000s, and the NIH started investing in the field by sponsoring centers for systems biology research.
Merging statistics with biology
Reinitz moved to the University of Chicago as part of a systems biology initiative in 2011, with appointments in both the Biological and Physical Sciences Divisions.
“I particularly admired that John, although not a classically trained statistician, made himself at home in our statistics department,” said Department Chair Matthew Stephens, Ralph W. Gerard Professor in Statistics and Human Genetics. “He embraced his move here as an opportunity to learn something new (as well as to teach us a thing or two!).”
“John personified a major tenet of our department, that statistics flourishes when engaged with serious scientific problems and vice versa,” said Stephen Stigler, Ernest DeWitt Burton Distinguished Service Professor Emeritus of Statistics. “John brought statistical modeling into his laboratory work, and he enthusiastically brought his insights from the DNA laboratory into his teaching of statistics.”
Having established the modeling approaches in developmental biology, Reinitz applied them to other areas, particularly evolutionary biology (in collaboration with Marty Kreitman and Misha Ludwig), stochastic gene regulation, and machine learning. He remained active until the very end; his latest efforts involved using cutting-edge, single-molecule techniques to understand the role and control of stochastic fluctuations in gene regulation.
Reinitz was a demanding boss who taught his trainees by walking the walk. He personally administered the lab’s Linux workstations, trained students in microscopy, and would happily troubleshoot code with them. When he was late for lab meetings at Stony Brook due to New York traffic, he would assess himself a fine. He prized straight talk and an open exchange of ideas, and it was hard for his trainees to not be infected by his independent and rebellious bulldog streak. Intellectual honesty was paramount. One of his favorite stories was how he was grateful to a reviewer for saving the lab’s reputation by rejecting a manuscript with a technical error that had sneaked past him.
A long, strange trip
Reinitz acknowledged that, given the outsider status of modeling in biology, his trainees faced an uphill battle for acceptance and went above and beyond in his support. Many of his trainees learned later that his recommendation letters were unusually long—10 pages in one case—and, while avoiding the exaggerated praise common in such missives, laid out a detailed argument of why it was in the institution’s best interest to hire the trainee.
Reinitz had a truly engaging and colorful personality with a razor-sharp wit; a comment in his code read, “I could explain how this works, but then I’d have to kill you.” A life-long Grateful Dead and Phish fan, he dedicated the source code of the connectionist model to Jerry Garcia after the musician’s death.
Science fiction and fantasy, especially the works of J.R.R. Tolkien, were another passion. Blessed with a rapacious memory, it was rare to have a conversation in which he did not quote verbatim either song lyrics or some snippet from sci-fi literature. He loved to travel, especially internationally, and developed collaborations across the globe, including Russia, France, and Brazil.
Reinitz was a consummate contrarian and would always have unconventional takes on issues gleaned by obstinately following his nose and backed up by inimitable reasoning. Interacting with him was never boring, always fun, and often enlightening.
John Bertram Reinitz is survived by his wife, Dr. Ilene Reinitz, and daughter, Julia Reinitz.
