ICAR-IASRI, New Delhi

Natural Product Classifier AI

Deep learning–powered hierarchical classification of natural products using ChemBERTa transformer architecture. Classify any molecule from its SMILES representation into Pathway, Superclass, and Class.

3
Classification Levels
ChemBERTa
Transformer Model
Real-time
Inference Speed

Why NPC-BERT AI?

Leveraging state-of-the-art deep learning to bridge cheminformatics and natural product research.

Hierarchical Classification

Three-level taxonomy — Pathway, Superclass, and Class — capturing the full biological and chemical ontology of natural products.

ChemBERTa Backbone

Built on the ChemBERTa transformer, pre-trained on millions of molecules for robust chemical language understanding.

Real-Time Predictions

Sub-second inference for individual molecules and efficient batch processing for large-scale screening campaigns.

Confidence Scores

Every prediction comes with probability scores and top-N alternatives, enabling informed decision-making.

REST API

Fully documented API endpoints for programmatic access, enabling integration into existing computational pipelines.

2D Structure Rendering

Instant 2D molecular structure visualization from SMILES input using RDKit-based depiction.

How It Works

01

Input SMILES

Provide the SMILES (Simplified Molecular Input Line Entry System) notation of your compound.

02

Tokenization

The SMILES string is tokenized using the ChemBERTa tokenizer into sub-word tokens optimized for chemical notation.

03

Transformer Encoding

ChemBERTa encodes the tokens into a rich molecular representation capturing structural and chemical features.

04

Hierarchical Classification

Three specialized heads predict Pathway → Superclass → Class in a hierarchical manner with confidence scores.

Ready to Classify Your Molecules?

Try our AI-powered classifier now — just paste a SMILES string and get instant results.

Launch Predictor