A third of approved drugs act through GPCRs. GPCRclaw makes nanobody campaigns inspectable.
GPCRs are one of medicine's largest target classes. GPCRclaw turns an ECL2-focused receptor brief into a GPU-run VHH design campaign with visible structures, model outputs, rankings, and research-use boundaries.
- Structure-native nanobody design
- Cloud GPU model execution
- Evidence-rich candidate dossiers
36%of approved drugs target GPCRs
Active path
cloudLPAR1 ECL2 Campaign
- Target
- LPAR1
- Template
- 7TD0
- ECL2
- 188-211
- Candidates
- 10
A100 GPU VMrunning
RFantibodydesign
Boltz-2evaluate
Recorded demo
Open on YouTube Watch the GPCRclaw campaign walkthrough
The demo shows the current campaign flow from receptor context through GPU model execution and ranked research-support evidence.
Live cloud control
Run the full campaign loop locally, then launch real GPU jobs when needed
GPU VM
runningA100 worker VM running
Google Batch GPU runtime
Drug design modelRFantibody / RFdiffusiongenerating LPAR1 ECL2 VHH candidates
runningEvaluation modelBoltz-2scoring ipSAE, epitope contacts, ipTM, pTM, and VHH developability
runningProjectnot configured
RFantibody buildunknownrfantibody-worker:latest
Boltz-2 imagemissingboltz2-worker:latest
Runs0local .gpcrclaw records
Campaign loops0none started
Published images
No latest model images visible yet.
Run records
No recent Batch jobs returned.
Real campaign loops
No RFantibody to Boltz-2 loop has been started.
Live workflow
The live path now follows one GPU VM from launch through drug design and model-based evaluation.
Boot GPU VM
Start the A100 worker VM, attach the LPAR1 campaign inputs, and keep the run record tied to the cloud job.
Run drug design model
Execute RFantibody/RFdiffusion on the GPU VM to generate ECL2-focused VHH candidates and output artifacts.
Run evaluation model
Filter and gate generated candidates with Boltz-2 ipSAE, epitope contacts, ipTM, pTM, VHH developability checks, and structure artifacts before ranking returned results.
Built for GPCR drug discovery teams
Nanobodies are useful because they can stabilize specific GPCR conformations. GPCRclaw makes that campaign logic visible from receptor context to ranked evidence.
Structure-native
The campaign starts from receptor structure, loop context, and hotspot constraints rather than a text-only target brief.
GPU-native
Design and evaluation jobs run as explicit model workers with artifact paths, model names, and retryable run state.
Evidence-first
Every returned candidate is framed as a research-support dossier: structures, metrics, provenance, and limitations.