Submit Job All Jobs Debug
Pipeline ready Database not configured — check /debug Email not configured — check /debug
0
Jobs Submitted
0
Jobs Completed
0
Jobs Failed

Submit a Job

Protein Name
Upload CSV
Auto-fetches bioactivity data (ChEMBL), crystal structures (PDB), and decoys (DUD-E).
For structure-based pharmacophore. Leave empty to auto-find.
More compounds = better model but longer runtime.
Immediate confirmation + results-ready email with download link. Leave empty if you don't want emails.

What you get

1. QSAR Model

RandomForest vs PLS comparison, cross-validated, with applicability domain and Y-randomization.

2. Ligand Pharmacophore

Consensus features from active compounds (3D) with labeled feature images.

3. Structure Pharmacophore

Features from crystal structure binding site (if PDB available).

4. 3D Hypothesis Ranking

Multiple hypotheses ranked by enrichment (AUC-ROC, EF, BEDROC).

5. Library Screening

All compounds screened against the best hypothesis using 3D Kabsch alignment.

6. Interactive 3D Viewer

View the pharmacophore in your browser with labeled feature spheres.

How to read the metrics

AUC-ROC

0.5 = random, >0.7 = good, >0.9 = excellent discrimination of actives from decoys.

EF(1%)

Enrichment Factor at 1% — how many more actives in the top 1% vs random (EF=5 means 5x better).

Q² (CV)

Cross-validated R² of the QSAR model (>0.5 acceptable, >0.7 good).

Pharm Score

3D alignment score (0-1) — fraction of features matched × alignment quality.