SynerGPCR

Bridging in vitro GPCR functional evidence and clinical approval
through synergistic assay-chain analysis and AI-predicted activity.

Try: Aripiprazole  ·  DRD2  ·  Parkinsonism  ·  Browse GPCR targets

SynerGPCR organises GPCR assay data into a four-layer functional chain, each layer classified by its readout in the signalling cascade:

L1 — Receptor Binding L2 — G-Protein Coupling +AI L3 — β-Arrestin Recruitment +AI L4 — Reporter Gene Activity

About  — assay layer definitions · Bliss Synergy Index · data sources    Help  — how to search · reading assay chain diagrams · AI confidence scores

Assay Chain Depth and Clinical Approval Synergy
Single-Layer Clinical Approval Rates

Approval rates remain below 1% for compounds assessed at any single signalling layer.

Single-layer approval rates across four signalling layers
Clinical translation requires multi-layer assay evidence.
Bliss Synergy Index by Assay Chain Pattern Interactive

Multi-layer assay activity co-occurs with clinical approval synergistically — quantified using the Bliss Synergy Index (SI).

Full chain: L1 (Receptor Binding) + L2 (G-Protein Coupling) + L3 (β-Arrestin Recruitment) + L4 (Reporter Gene Activity).

AI-Based Chain Completion

GPCRact (an E(n)-equivariant graph neural network) predicts missing G-protein coupling and β-arrestin activities de novo, recovering approximately 79% of the experimental synergy benchmark at both signalling layers.

Assay Chain Completion via AI Prediction
GPCRact AI completion recovers ~79% of experimental synergy ceiling
AI completion recovers 79.7% (L2) and 79.6% (L3) of the experimental synergy ceiling.
GPCRact Assay MoA Prediction Performance

(a) G-Protein Coupling (L2)

GPCRact L2 confusion matrix — BACC = 0.875

(b) β-Arrestin Recruitment (L3)

GPCRact L3 confusion matrix — BACC = 0.834

GPCRact mode-of-action prediction: Balanced Accuracy = 0.875 for G-Protein Coupling (L2) and 0.834 for β-Arrestin Recruitment (L3), evaluated on held-out approved drugs.

[AI-predicted, not experimentally validated] GPCRact predictions are computationally generated and have not been independently validated in the laboratory.
Learn more about GPCRact ↗ Paper ↗ GitHub ↗