AI-Driven Protein Design Breakthrough: Zhejiang University Publishes De Novo GPCR Exoframe Modulators in Nature

Unveiling De Novo GPCR Exoframe Modulators: A New Era in Protein Engineering

G-protein-coupled receptors (GPCRs), the largest family of cell surface receptors in humans, play pivotal roles in physiology and are targets for over 30% of approved drugs. Encoded by nearly 800 genes, these seven-transmembrane proteins detect extracellular signals like hormones and neurotransmitters, transducing them into intracellular responses via conformational changes. Traditional small-molecule drugs bind orthosteric sites, mimicking natural ligands, but allosteric modulation—binding distant sites to fine-tune activity—offers greater selectivity and fewer side effects. Enter GPCR exoframe modulators (GEMs), de novo designed proteins from researchers at Zhejiang University's Shanghai Institute for Advanced Study, published in Nature on February 16, 2026. These AI-crafted proteins bind the extracellular face of GPCR transmembrane domains, inspired by natural regulators like RAMPs, enabling precise allosteric control.

This breakthrough underscores Zhejiang University's prowess in computational biology, positioning it as a leader in China's burgeoning AI-driven biotech landscape. Lead author Shizhuo Cheng and colleagues, including Jia Guo and Yun-Li Zhou, leveraged deep learning to hallucinate novel protein scaffolds, marking a shift from sequence-based to function-oriented design.

What Are GEMs and Why Do They Matter?

GEMs represent proteins designed from scratch (de novo) to encircle and stabilize GPCR transmembrane helices from the extracellular side, akin to an 'exoframe' scaffold. Unlike antibodies targeting extracellular loops or nanobodies stabilizing inactive states, GEMs engage the helical bundle directly, inducing specific conformations. In their study, four GEMs modulated the dopamine D1 receptor (D1R), a class A GPCR implicated in Parkinson's, schizophrenia, and addiction.

• Ago-PAM GEM enhanced agonist potency and efficacy, boosting cAMP signaling.
• NAM GEM suppressed activity, reducing off-target effects.
• Biased GEM selectively activated G protein over β-arrestin pathways.
• Another variant fine-tuned signaling bias.

This functional diversity stems from tailored structural prompts in AI design, promising tools for dissecting GPCR pharmacologies and treating receptoropathies.

The AI-Driven Design Pipeline: Step-by-Step Innovation

Zhejiang researchers employed a hallucination-like strategy using diffusion models, likely RFdiffusion or similar, trained on protein structures. Here's the process:

1. Motif Scaffolding: Fixed GPCR TM extracellular motifs as anchors.
2. Prompt Engineering: Three prompts ensured helical grip, shape complementarity, and dynamic stability—e.g., symmetric arms wrapping helices.
3. Generative Sampling: AI generated thousands of candidates; ranked by predicted affinity via AlphaFold-Multimer.
4. Filtering: Biophysical simulations discarded unstable designs.
5. Expression & Testing: Yeast display and mammalian cells validated binders.

This yielded high-affinity GEMs (Kd ~10-100 nM), validated by cryo-EM showing buried interfaces >2000 Ų. Such methods build on David Baker's lab tools, now powering global de novo design.

Structural and Functional Validation: Cryo-EM Insights

Cryo-electron microscopy resolved GEM-D1R complexes at 3.2 Å, revealing how exoframes rigidify TM helices, stabilizing active-like states. Functional assays in HEK293 cells confirmed modulation: ago-PAM shifted dose-response curves leftward, amplifying dopamine effects 10-fold. Critically, it rescued 5/6 D1R loss-of-function mutants (e.g., TM6 mutations), restoring 50-80% wild-type activity—vital for genetic disorders.

Biophysical metrics like BRET and SPR corroborated specificity, with no off-target binding to 20+ GPCRs tested.

Therapeutic Horizons: From Bench to Clinic

GEMs unlock GPCR drugging beyond orthosteric pockets, ideal for intractable targets like class B/C GPCRs. For D1R mutants, ago-PAMs could treat rare neurological conditions; biased modulators mitigate schizophrenia side effects. Encodable genetically, GEMs suit gene therapy, CAR-T synergies, or synthetic circuits. Challenges include delivery (e.g., AAV vectors) and immunogenicity, but protein engineering mitigates these.

Link to full study: De novo design of GPCR exoframe modulators (Nature).

Zhejiang University's Rising Star in AI and Structural Biology

Zhejiang University, a C9 League member, invests heavily in interdisciplinary hubs like SIAS-ZJU, fostering talents like Cheng under mentors Yan Zhang. China now rivals the US in Nature-indexed biotech papers, with ZJU ranking top-5 domestically.Explore opportunities at top Chinese universities. This publication exemplifies national pushes like 'Made in China 2025' for AI-biotech fusion.

China's Leadership in Computational Protein Design

China produces 25% of global AI papers; in protein design, labs at Tsinghua, Peking U complement ZJU. Stats: 2025 saw 15% rise in de novo designs from Chinese teams. Implications for higher ed: surging PhD programs in comp bio, attracting international talent. Research jobs in this field abound.

• Benefits: Accelerated drug discovery (months vs. years).
• Risks: Ethical AI use, dual-use concerns.
• Comparisons: GEMs vs. nanobodies—deeper TM access.

Challenges, Solutions, and Future Directions

Scalability to 800+ GPCRs demands broader AI training; solutions include multi-objective diffusion models. Regulatory paths for protein therapeutics mirror monoclonals. Outlook: GEM libraries for high-throughput screening, partnerships with pharma like Novartis.

For aspiring researchers, craft a winning academic CV to join this revolution. Research assistant roles at ZJU-like institutions are booming.

Career Insights: Thriving in AI-Protein Design

This field demands skills in PyTorch, AlphaFold, cryo-EM. In China, postdocs earn ~¥400k/year; faculty at ZJU >¥1M. Platforms like Rate My Professor offer insider views. Postdoc positions bridge to independence.

Stakeholders—from patients to policymakers—hail GEMs as a paradigm shift, blending computation with biology for healthier futures.

Photo by yonatan huaman castro on Unsplash