Roche Bets on Molecular Glue Degraders in Discovery Deal With China's Defand

Roche has signed a research and licensing agreement with Defand Therapeutics, a Hangzhou-based biotech founded in 2023, to discover molecular glue degraders for cancer and immune-mediated diseases. The deal was announced on October 8, 2026. Defand will use its screening platform to find candidate molecules against several targets chosen with Roche. Roche will take over preclinical and clinical development and hold exclusive worldwide commercial rights.

Neither company disclosed financial terms in Defand's announcement. STAT reported that the deal could be worth close to $1 billion including milestones, citing a source familiar with the agreement. Endpoints News, as cited by European Biotechnology Magazine, reported an upfront payment of under $10 million, which means almost all of the headline value depends on programs reaching future preclinical, clinical and regulatory milestones. Defand would also receive tiered royalties on sales, and Roche has an option to add further targets.

The deal is small in upfront terms. It is notable for two reasons: it buys discovery capability rather than a clinical-stage drug, and it targets one of the more technically demanding areas of small-molecule research.

From blocking proteins to removing them

Most small-molecule drugs work by occupation. They sit in a pocket on a target protein, often an enzyme's active site, and block its activity for as long as enough drug is present. This approach has two well-known limits. Many disease-driving proteins, including transcription factors and scaffolding proteins, have no suitable pocket and have been regarded as "undruggable". And a protein that is merely blocked still exists, so it may keep acting through other surfaces, and cells can compensate by making more of it.

Targeted protein degradation takes a different route. Instead of inhibiting a protein, a degrader directs the cell's own disposal system to destroy it. In human cells, unwanted proteins are tagged with chains of a small protein called ubiquitin by enzymes known as E3 ubiquitin ligases, and tagged proteins are then broken down by the proteasome. Humans have several hundred E3 ligases, each recognizing its own set of substrates. A degrader works by persuading one of these ligases to treat a disease-related protein as one of its substrates.

This changes the pharmacology. A conventional inhibitor works only while it occupies its target, so it usually has to be dosed to keep high, continuous exposure. A degrader acts catalytically: once the target has been tagged, the degrader can let go and bring another copy of the protein to the ligase. The effect can also outlast the drug, because the cell has to make new protein before the target's activity returns. In principle this allows lower or less frequent doses, and it removes every function of the protein, including roles that do not depend on the pocket an inhibitor would block.

How molecular glues differ from PROTACs

There are two main classes of degrader.

• Proteolysis-targeting chimeras (PROTACs) are bifunctional molecules. One end binds the target, the other binds an E3 ligase, and a chemical linker joins them. They can be designed systematically, but they tend to be large molecules, which can make oral absorption, cell penetration and distribution into tissues harder.
• Molecular glue degraders are smaller, single molecules. A glue typically binds to the E3 ligase and reshapes its surface so that the ligase now recognizes a protein it would otherwise ignore, or it stabilizes a weak natural contact between the two. The target does not need a deep binding pocket of its own, because the glue works at the interface between the two proteins.

The best-known molecular glues were discovered by accident. Thalidomide and its derivatives lenalidomide and pomalidomide were used for years before researchers showed that they bind cereblon, a component of an E3 ligase complex, and redirect it to degrade transcription factors such as IKZF1 and IKZF3 that multiple myeloma cells depend on. Newer cereblon-modulating degraders, including Bristol Myers Squibb's iberdomide and mezigdomide, have since advanced into late-stage myeloma trials.

The term "molecular glue" is also used more broadly for small molecules that stick two proteins together without causing degradation. The most prominent recent example is Revolution Medicines' daraxonrasib, an inhibitor of the RAS cancer protein. In the phase 3 RASolute 302 trial in previously treated metastatic pancreatic cancer, median overall survival was 13.2 months with daraxonrasib versus 6.7 months with chemotherapy (hazard ratio 0.40). Results of that kind have drawn wider investment into protein-interface chemistry, including degraders.

Why discovery is the bottleneck

The main obstacle for molecular glues is finding them. Because a glue's effect depends on the shape of two proteins meeting, it is hard to design from first principles, and many historical glues were identified only after their biological effect was known. Companies in this field are therefore building screening systems that test large compound libraries for the specific outcome they want: the loss of a chosen target protein from cells.

According to European Biotechnology Magazine, Defand's platform can screen cytoplasmic, nuclear, transmembrane and viral proteins. The company has established screening cell lines for more than 100 clinically relevant targets and compound libraries focused on several E3 ligases. Its chief operating officer, Huanwei Ru, said the platform allows parallel screening across dozens of targets. The selected targets, the stage at which programs pass to Roche and any clinical timeline have not been disclosed.

The China discovery angle

STAT noted that the structure of the deal is itself significant. Much of the recent wave of Western licensing from Chinese biotechs has involved molecules that were already in the clinic, where China-based developers have built a reputation for running early studies quickly. Here, Roche is paying a young Chinese company to carry out early discovery research on its behalf, which suggests large drugmakers increasingly see Chinese platforms as a source of new chemistry, not only of late-stage assets. European Biotechnology Magazine reported that the deal is one of several China collaborations Roche has signed in recent weeks, alongside agreements in antibody-drug conjugates and T-cell engagers.

Risks and limitations

• Early stage. The deal covers discovery. No candidate molecules, targets or preclinical data have been disclosed, and most discovery-stage programs never reach the clinic.
• Back-loaded value. With a reported upfront payment under $10 million, the near-$1 billion figure is a ceiling that assumes many milestones are met. Neither company has confirmed the numbers.
• Selectivity. Because glues reshape an E3 ligase's surface, they can cause unintended degradation of other proteins. Off-target degradation of proteins such as GSPT1 or SALL4 has been a recognized concern with cereblon-based compounds, and it has to be screened out carefully.
• Ligase dependence. Most clinical degraders still rely on a small number of E3 ligases, chiefly cereblon. Tumors can become resistant by losing or mutating the ligase, and ligase expression varies between tissues.
• Translation. Degrading a protein in a screening cell line does not guarantee a clinical benefit. Pharmacology, safety and patient selection still have to be established program by program.

What to watch

• Disclosure of the first targets and any candidates nominated for preclinical development under the collaboration.
• Whether Roche exercises its option to extend the agreement to more targets.
• Clinical progress of other molecular glue degraders, including the cereblon-based myeloma programs, as evidence of how well the class translates.
• Efforts to use E3 ligases beyond cereblon, which could widen the range of tissues and targets degraders can reach.
• Further discovery-stage deals between Western drugmakers and Chinese platform companies.

This article is for information only and is not medical advice. Treatment decisions should be made with a qualified clinician.

Sources: STAT; European Biotechnology Magazine; Dana-Farber Cancer Institute; The ASCO Post