At Baylor College of Medicine, scientists have moved beyond the long-standing paradigm of silencing disease proteins and toward erasing them entirely — a distinction that may prove as consequential as the difference between muting a fire alarm and extinguishing the fire. By weaving artificial intelligence, structural modeling, and high-throughput proteomics into a single discovery workflow, the team identified molecular glues capable of directing VAV1, a protein entangled in blood cancers and autoimmune disease, toward the cell's own disposal machinery. The work, published in Nature Communicat
AI-Guided Proteomics Uncovers Molecular Glues to Degrade VAV1 in Blood Cancers
Complete protein removal might produce a more thorough therapeutic effect
Mark
So the basic idea is that instead of turning a protein off, you're trying to delete it entirely. Why does that matter?
Mimi
Because a protein that's blocked can sometimes find workarounds. It might activate a different pathway, or the cell might make more of it to compensate. If you remove the protein altogether, those escape routes don't exist.
Luke
But we should be clear—these are preclinical compounds. We don't know yet if they'll be safe in humans or if they'll actually work in a patient with blood cancer.
Mark
Right. So what's the actual innovation here? The molecular glue concept isn't new, is it?
Mimi
The innovation is twofold. First, they found a new degron on VAV1—a degradation signal that's structurally different from what cereblon usually recognizes. Second, they used AI to predict how the three-part complex would assemble without needing an experimental structure first.
Luke
That's the GluePlex tool. But I want to flag something: the off-target degradation of LIMD1. They found it through proteomics, which is good, but it shows that even with careful design, these glues can hit unintended targets.
Mark
Is that a deal-breaker?
Mimi
Not necessarily. LIMD1 degradation might not be harmful, or it might even be therapeutic. But you're right that it needs to be studied carefully.
Luke
And the T-cell experiments—those are in isolated cells, not in living organisms. We don't know how NGT-201-18 would behave in a whole immune system.
Mark
So what's the real takeaway here?
Mimi
That you can use AI and proteomics together to discover and optimize molecular glues faster than traditional methods. And that VAV1 might be a tractable target for blood cancers and autoimmune disease.
Luke
Assuming the compounds survive the next phase of testing. Which is a big assumption.
O Pulso
- VAV1 has long resisted conventional drug strategies because blocking its activity leaves the protein intact and able to recover influence — complete elimination is the more radical and potentially more durable solution.
- The team's AI-guided platform, GluePlex, predicted how a three-part molecular complex — drug, target protein, and cellular disposal machinery — would assemble, bypassing the usual bottleneck of waiting for laboratory-derived crystal structures.
- Iterative chemical refinement transformed an early candidate, NGT-201-12, into the more potent NGT-201-18, which successfully reduced VAV1 levels and suppressed T-cell activation in primary human cells.
- Proteome-wide profiling uncovered that NGT-201-18 also degrades a second protein, LIMD1, signaling that molecular glues can engage multiple distinct degradation signals — a finding that demands careful mapping of unintended effects.
- The compounds are still preclinical, with pharmacology, safety, and disease-model validation all ahead, but the integrated discovery strategy itself is immediately transferable to other hard-to-drug targets.
At Baylor College of Medicine, scientists have moved beyond the long-standing paradigm of silencing disease proteins and toward erasing them entirely — a distinction that may prove as consequential as the difference between muting a fire alarm and extinguishing the fire. By weaving artificial intelligence, structural modeling, and high-throughput proteomics into a single discovery workflow, the team identified molecular glues capable of directing VAV1, a protein entangled in blood cancers and autoimmune disease, toward the cell's own disposal machinery. The work, published in Nature Communications, offers not only a preclinical foothold against VAV1-driven illness but a replicable blueprint for how computation and experiment might together accelerate the search for medicines that eliminate, rather than merely restrain, the proteins that make us sick.
Researchers at Baylor College of Medicine have identified compounds that eliminate a key immune-signaling protein rather than merely suppressing it — a distinction with potentially significant therapeutic consequences. Published in Nature Communications, the work targets VAV1, a protein implicated in blood cancers and autoimmune diseases, using a class of agents known as molecular glues.
Unlike conventional drugs that block a protein's function while leaving it in place, molecular glues act as cellular matchmakers with a lethal purpose: they simultaneously bind the disease protein and the cell's own disposal machinery, forcing the target into the trash entirely. Jin Wang and his team, including first author Hanfeng Lin, reasoned that complete removal might produce a more thorough effect than partial inhibition.
The team screened a molecular library using high-throughput proteomics — a method capable of measuring thousands of proteins at once — and identified NGT-201-12 as a promising candidate. To understand and improve upon it, they built a computational tool called GluePlex, which combined AI-based protein-structure prediction with physics-based modeling to simulate how VAV1, the glue, and a disposal-directing protein called cereblon would assemble into a three-part complex. The model pinpointed a specific surface loop on VAV1 — a degradation signal called a degron — that proved structurally novel compared to signals typically recognized by cereblon-targeting compounds.
Guided by these predictions, the researchers modified NGT-201-12 by adding halogen atoms to restrict its flexibility, producing NGT-201-18, a more potent degrader. In primary human T cells, the optimized compound reduced VAV1 levels and suppressed T-cell activation. Proteome-wide profiling also revealed that NGT-201-18 degrades a second protein, LIMD1, underscoring the need to map the full cellular landscape when deploying such compounds.
The work remains preclinical — pharmacology, safety, and disease-model studies all lie ahead. But the broader significance may be methodological: by integrating AI, structural modeling, and proteomics from the earliest stages of discovery, the Baylor team has demonstrated a strategy that does not require a crystal structure in hand and could be applied to other disease targets that have so far resisted conventional approaches.
Researchers at Baylor College of Medicine have identified a new class of compounds that could attack blood cancers and autoimmune diseases by eliminating a key immune-signaling protein entirely rather than merely dampening its activity. The discovery, published in Nature Communications, combines artificial intelligence with high-throughput proteomics—a technique that measures thousands of proteins at once—to find and refine what scientists call molecular glues.
Molecular glues work like cellular matchmakers with a lethal purpose. They bind to disease-causing proteins and simultaneously attach to the cell's built-in disposal system, forcing the entire protein into the trash. This is fundamentally different from conventional drugs, which typically block a protein's function while leaving the protein itself intact. Jin Wang, director of Baylor's Center for NextGen Therapeutics, and his team, including first author Hanfeng Lin, reasoned that complete protein removal might produce a more thorough therapeutic effect than partial inhibition. The target was VAV1, a protein that regulates immune-cell signaling and has been implicated in both blood cancers and autoimmune conditions.
The researchers screened a library of candidate molecules using high-throughput proteomics, looking for compounds that would cause VAV1 to disappear while leaving most other proteins untouched. One compound, designated NGT-201-12, showed promise. Further work revealed that the degradation required two cellular components: the proteasome, which is the cell's primary protein-disposal machine, and cereblon, a protein that helps direct targets to that machinery. But the researchers wanted to understand exactly how these pieces fit together—and to improve upon their initial discovery.
They developed a computational tool called GluePlex that combines AI-based protein-structure prediction with physics-based modeling. Rather than waiting for experimental structures to emerge from the lab, GluePlex predicted how VAV1, cereblon, and the molecular glue would assemble into a three-part complex. The model pinpointed a specific region of VAV1 called the SH3-2 domain as essential for degradation. More precisely, it identified a small surface loop on VAV1 that acts as a degradation signal—what researchers call a degron. This particular degron was structurally distinct from the degradation signals typically recognized by cereblon-targeting glues, suggesting the team had found something novel.
Using a technique called Free Energy Perturbation, the researchers ranked candidate compounds based on how strongly they would form the three-part complex. They then modified NGT-201-12 by adding halogen atoms, which restricted the molecule's flexibility and improved its ability to degrade VAV1. The result was NGT-201-18, a more potent degrader that formed a stronger complex and more efficiently eliminated VAV1 from cells. When tested in primary human T cells, NGT-201-18 reduced VAV1 levels and suppressed T-cell activation—a sign that the compound was working as intended.
Proteome-wide profiling revealed that NGT-201-18 targeted VAV1 as its primary goal but also degraded a second protein, LIMD1, which carries a different type of degron. The finding underscores both the power and the complexity of the approach: a single molecular glue can engage structurally distinct degradation signals, which means researchers must carefully map the entire proteome to understand what else a compound might affect. The compounds remain preclinical, and substantial additional work lies ahead—pharmacology studies, safety assessments, selectivity testing, and validation in disease models will all be necessary before any therapeutic potential can be realized.
What the Baylor team has demonstrated, however, is a broader discovery strategy that could accelerate the development of molecular glues for other disease targets. By integrating artificial intelligence, structural modeling, and proteomics from the earliest stages of a project, they have shown a path forward that does not depend on having a crystal structure in hand or relying solely on traditional screening methods. The work offers both a starting point for therapies aimed at VAV1-driven blood cancers and autoimmune disorders and a template for how computational and experimental tools can work in concert to find entirely new ways to eliminate disease.
Citações Notáveis
Because degradation via molecular glue removes the whole protein rather than inhibiting one of its functions, the strategy could produce a more complete therapeutic effect.— Jin Wang, director of Baylor's Center for NextGen Therapeutics
This work introduces a series of VAV1-targeting molecular glues and shows how artificial intelligence, structural modeling and proteomics can work together at the earliest stage of a project.— Jin Wang
Quer a matéria completa?
Leia o original em Genetic Engineering and Biotechnology News ↗