Multi-Agent In Silico Program · 5 Phases · 4 Hard Gates
GLP-1 Glyco-Masking Drug Discovery
A fully in silico, 5-agent computational pipeline that designed synthesis-ready GLP-1 analogs with delayed receptor activation via glycan masking — reducing the nausea-associated early receptor hit while preserving full therapeutic efficacy post-cleavage.
Pipeline Overview
🔬
Phase 1
Masking Feasibility
50 → 10 ✓
⏱
Phase 2
Linker Timing
20 → 5 ✓
🧬
Phase 3
Receptor Activation
50 → 3 ✓
📈
Phase 4
Temporal PK
3 → 3 ✓
🏆
Phase 5
Final Selection
top 3 ✓
Top Synthesis Candidates
Rank 1GLP1-GM-F54D54
HTEGTFTSDVSAYLQGQAWKEAIAWLVKGR
- SASA suppression
- 82.5%
- Predicted efficacy
- 102.7%
- Late/Early ratio
- 5.58×
- Composite score
- 0.794
Glycan: Biantennary sialylated @ Glu9
Linker: Cathepsin B GFLG · t½ 5.72 h
Mutations: A8T · S18A · E21Q · Y25W · G28A
Rank 2GLP1-GM-942303
HTEGTFTSDVSAYLQGQAWKEFIAWLVKGR
- SASA suppression
- 82.5%
- Predicted efficacy
- 101.1%
- Late/Early ratio
- 5.58×
- Composite score
- 0.784
Glycan: Biantennary sialylated @ Glu9
Linker: Cathepsin B GFLG · t½ 5.72 h
Mutations: A8T · S18A · E21Q · Y25W
Rank 3GLP1-GM-E562AD
HTEGTFTSDVSAYLQGQAAIEFIAWLVKGR
- SASA suppression
- 82.5%
- Predicted efficacy
- 92.8%
- Late/Early ratio
- 5.58×
- Composite score
- 0.747
Glycan: Biantennary sialylated @ Glu9
Linker: Cathepsin B GFLG · t½ 5.72 h
Mutations: A8T · S18A · E21Q · L26I
5-Agent Architecture
- ✓Planner — phased execution plans + compute budgets
- ✓Generator — 50→10→5→3 candidate funnel
- ✓Evaluator — Shrake-Rupley SASA + ODE PK/PD model
- ✓Scorer — weighted composite (masking·timing·binding·stability)
- ✓Controller — hard gates with automatic pruning
Mol* Glyco-Aware Viewer
- ✓Side-by-side comparison of reference vs candidate
- ✓Glycan residue detection (NAG · MAN · GAL · SIA)
- ✓Glycosite highlighting with annotation tooltips
- ✓Sequence↔structure synchronization
- ✓Upload your own PDB/mmCIF files