New UChicago platform rapidly generates custom protein binders to target disease

In the human body, the boundary between health and severe illness can be microscopic. For decades, molecular scientists have grappled with a frustrating biological reality: the proteins driving devastating diseases often look nearly identical to the ones keeping us alive.

"The difference between a healthy protein and an unhealthy protein can come down to just a few amino acids or building blocks," explains Joshua Pixley, who recently earned his undergraduate degree from the University of Chicago after completing a rare triple major in molecular engineering, chemistry, and biochemistry.

This extreme similarity makes designing targeted therapeutics remarkably difficult. "There are a lot of diseases, cancers and neurodegenerative diseases, that come down to ultimately the dysregulation and misfunction of specific proteins, many of which we can't directly target because we don't have ways to either measure them, purify them, or interact with them," Pixley notes.

Consider the RAS family of proteins, which are dysregulated in more than 90% of pancreatic cancers. Two members of this family, HRAS and KRAS, share 94% of the same sequence, making selectively drugging one RAS member, but not interfering with the normal functioning of the other, a challenging biophysical problem.

Now, a breakthrough platform developed in Prof. Bryan Dickinson’s lab at UChicago offers a powerful way forward. In a new paper published in PNAS, co-first authors Pixley and postdoctoral scholar Matthew Styles detail the creation of "PANCS-spec-Binders", a technology capable of rapidly engineering highly specific protein-interacting partners, or "binders," that can zero in on a single target while entirely avoiding its near-identical twins.

Accelerating the Discovery Timeline

Traditionally, finding a molecular binder to distinguish between two proteins that share a 94% sequence identity requires months of intensive, exhaustive laboratory labor. Within the RAS protein family, that minor 6% difference carries vastly different clinical profiles: KRAS is famously implicated in pancreatic cancer, whereas HRAS drives complex interactions in Bladder and Head and Neck cancer.

The co-authors' breakthrough built directly upon a foundational paper published in eLife by former Dickinson lab graduate student Victoria Cochran Xie, who utilized Phage-Assisted Continuous Evolution (PACE) to engineer protein specificity. Likewise, inspired by Kevan Shokat’s pioneering RAS-targeting work at UCSF, the UChicago team sought to push those boundaries further.

They engineered PANCS-spec-Binders, a platform powered by a massive synthetic library containing over 10 billion potential binding partners. To screen this vast pool rapidly, the system utilizes active bacteriophage replication to select functional variants within days. Isolated binders are then validated using engineered strains of bacteria that physically glow during a successful molecular interaction, completely bypassing conventional protein purification bottlenecks.

"You can go from an idea for a protein that you'd like to find an interacting partner for all the way through something that is completely workable and ready to employ within just a couple of weeks," says Pixley.

The lab proved exactly how nimble this ecosystem is in a concurrent study published in the Journal of the American Chemical Society. Collaborating with a UChicago cancer laboratory, researchers went from receiving uncharacterized protein IDs to delivering fully functional binders optimized for mini-protein degraders in just 26 days, a workflow that would traditionally drag on for six to twelve months using standard techniques.

Uncovering an AI Blindspot

The road to this platform was defined by nearly a year of intense trial and error. During development, the research team encountered a massive biological surprise when trying to decipher why their newly discovered binder displayed such flawless, unprecedented selectivity for HRAS over KRAS. Viewed through traditional structural lenses, the mechanism remained completely hidden.

The breakthrough revealed that the secret lay in a highly flexible, frequently overlooked region of the target protein. As Pixley notes:

"A single amino acid, just a couple atoms in an otherwise not very frequently studied part of the protein, which is very flexible, was the key driver of the selectivity which we were seeing. And this was really a surprising finding on our part."

This finding exposed a critical blind spot in modern computational models employed by the group. Remarkably, AlphaFold—the industry-leading artificial intelligence protein-prediction tool—failed to predict this crucial interaction. Even live tests with the newer AlphaFold 3 revealed lingering gaps in predicting these complex, flexible protein faces.

To definitively validate their discovery, the team collaborated with Argonne National Laboratory, utilizing advanced synchrotron crystallographic data to capture a clear visual confirmation of their original non-specific binder locked onto its HRAS target. Ultimately, the discovery underscores a vital truth in modern structural biology: even the most sophisticated predictive AI models still require rigorous, real-world experimental validation.

Mapping New Hotspots

Beyond targeting known cancer proteins, the platform uses a capability called "inference" to map entirely new drug targets. By subtracting binders that shield already-known interaction sites, researchers can uncover uncharacterized structural vulnerabilities, or "hotspots". The team successfully demonstrated this with the protein LC3B, a strategy that could eventually help target subtle post-translational modifications like phosphorylation or glycosylation.

This step outside their own walls offered the team a profound moment of scientific clarity. "Inside of a lab, it can be a little bit of an echo chamber sometimes... But that first meeting with the [synchrotron] team and hearing that they were also excited about helping us solve this problem to me was really validating that what we were working on was impactful," Pixley recalls. The high-resolution structural data is now headed to the open-source Protein Data Bank, soon making their discovery freely available to cancer researchers worldwide.

A Culture of Innovation

Behind the platform’s technical success lies a deeper story of institutional trust and intentional mentorship. Over three and a half years in the lab, Pixley navigated extensive experimental dead-ends under the close guidance of Styles, who provided the safety net necessary to take bold risks.

"I can't emphasize enough how important Bryan and Matt were in making this research possible... them letting me try things and take risks and also giving me the support when things failed," Pixley reflects.

This collaborative environment also empowered Pixley to step up as a leader within the university ecosystem. Alongside his triple major, he served as president of UChicago’s iGEM (International Genetically Engineered Machine) research team, passing his laboratory knowledge down to a new cohort of undergraduate students.

The impact of this high-trust culture continues to expand. The underlying biosensor technology is currently moving through the university's patent pipeline, and Styles is transitioning into a faculty role at South Dakota State University. Meanwhile, Pixley has moved on to the Christina Woo lab at Harvard University to pursue a Ph.D. in chemistry and chemical biology, where he plans to merge his understanding of physical protein interactions with computational AI models.

Looking back on his journey, Pixley views the breakthrough as a direct testament to the unique culture of the UChicago research community: "While I feel very proud of the research I was able to do, it's definitely a product of the environment I was in."

Citation: J.A. Pixley, M.J. Styles, K. Aphicho, M. Endres, P. Gade, K. Michalska, A. Joachimiak, & B.C. Dickinson, PANCS-spec-Binders: A system for rapidly discovering isoform- or epitope-specific binders, Proc. Natl. Acad. Sci. U.S.A. 123 (23) e2531008123, https://doi.org/10.1073/pnas.2531008123 (2026).

Funding: This research was supported by the National Institutes of Health, including the National Institute of General Medical Sciences and the National Cancer Institute. Additional funding was provided by the Camille and Henry Dreyfus Foundation Teacher Scholar Award, the Barry M. Goldwater Scholarship and Excellence in Education Foundation, and the Advanced Research Projects Agency for Health (ARPA-H) via Argonne National Laboratory. B.C.D. is a Biohub Investigator. The use of the Structural Biology Center resources at the Advanced Photon Source was supported by the U.S. Department of Energy Office of Science and operated by Argonne National Laboratory.