HIV-1 Protease Inhibitor Discovery

Research Demo

SELFIES-Transformer based de novo molecular generation and candidate evaluation.

Top Lead

Ligand_0150

Generated Molecules

5,000

Validity

100%

Uniqueness

69.88%

Novelty

96.77%

Best Docking Score

-10.923 kcal/mol

Successfully Docked

198

Study Snapshot

Representation: SELFIES

String-based molecular representation

Architecture: Decoder-only Transformer

GPT-style molecular generator

Pretraining: ZINC chemical library

General molecular grammar learning

Fine-tuning: HIV-1 protease inhibitors

Target-focused inhibitor generation

Docking: AutoDock Vina

Predicted receptor-ligand binding affinity

Viewer: 3Dmol.js

Browser-based molecular structure renderer

RANK #1

Top-Performing Ligand

Binding Affinity

-10.923kcal/mol

QED

0.60

4.15

Weight

477.6Da

logP

4.36