5-HT2B Pharmacophore
Pre-Screen Validation
EBP Wedge Detector v3.1.0
Validation overview
The deterministic 5-HT2B pharmacophore pre-screen was evaluated within a panel of 110 compounds spanning 13 validation categories. The documented binary confusion matrix covers 101 compounds. Nine additional compounds were handled by the structural pre-screen, but their binary reference labels and classifications have not yet been reconciled into that matrix.
Reporting scope: Youden’s J = 1.000, sensitivity = 100%, and specificity = 100% apply to the 101 documented binary classifications, not to the entire 110-compound panel.
Confusion matrix
| Actual class | Predicted positive | Predicted negative | Total |
|---|---|---|---|
| Positive | 17 TP | 0 FN | 17 |
| Negative | 0 FP | 84 TN | 84 |
| Total | 17 | 84 | 101 |
Validation panel
The panel includes known 5-HT2B agonists, borderline and partial agonists, structural near-misses, tryptamines, triptans, and pharmacological controls. Detailed molecular decision criteria remain proprietary.
Category distribution
| Original validation · toxic | 9 |
| Original validation · safe | 18 |
| Additional known agonists | 7 |
| Borderline / partial agonists | 6 |
| Phenethylamine near-misses | 11 |
| Tryptamines / psychedelics | 4 |
| Triptans | 6 |
| 5-HT3 / 5-HT4 drugs | 5 |
| Chemotherapy agents | 6 |
| Opioids | 6 |
| Stimulants | 1 |
| Cardiovascular | 7 |
| Diverse safe drugs | 24 |
| Total | 110 |
Selected compounds
Amphetamine-related
Fenfluramine · Norfenfluramine · Dexfenfluramine · MDMA · MDA
Aminorex-related
Aminorex · 4-Methylaminorex
Ergot-related
Ergotamine · Pergolide · Methysergide · Bromocriptine · Cabergoline · Dihydroergotamine · Dihydroergocristine · Lisuride
Other listed positives
Ropinirole · mCPP
Structural challenges
Ten highlighted structural challenges—fluoxetine, fluphenazine, morphine, paclitaxel, trazodone, amphetamine, sumatriptan, DMT, carvedilol, and palonosetron—were reported as correctly classified by the platform’s deterministic molecular architecture, rather than by the structural pre-screen alone.
These compounds are examples from the existing validation panel and are not added again to the total.
Reference-label quality
Vincristine excluded from v3
The supplied report states that vincristine was removed because published 5-HT2B Ki data could not be located in ChEMBL or primary literature to support its inherited “safe” label. It was replaced with paclitaxel, reported as having confirmed 5-HT2B inactivity (AC50 ≥ 8,800 nM; ChEMBL CHEMBL428647).
Role in the screening workflow
This deterministic structural pre-screen serves as Layer 1 of the 5-HT2B cardiac risk screening workflow, evaluating molecular features before full three-dimensional docking. The reported method uses no machine-learning training data or conventional force-field simulation at this stage.
Results describe the reported validation panel. They are not a measure of prospective clinical performance or an independently replicated external validation.