Back to Projects

Nobel Data Intelligence

Graph Neural Networks for molecular property prediction. 10K+ compounds. 85%+ accuracy.

View on GitHub
PyTorch GeometricRDKitGNN

From Molecules to Graphs

MoleculeTransformGraph
CCO (Ethanol)
CCO
Molecular Structure
Atoms and bonds

Pipeline Flow

SMILESInput
GraphRDKit
GNNPyTorch Geometric
PredictionProperties

How GNNs Learn

Message Passing

Each node aggregates information from its neighbors. After several layers, atoms "know" about their molecular environment — capturing the chemical context that determines properties.

GCNGATGraphSAGE
Layer123
01234
Node 0 aggregating from neighbors
hi = σ(Σj∈N(i) W·hj)

The Stack

PyTorch Geometric

GNN layers, batching, GPU acceleration

RDKit

Molecular parsing, feature extraction

ProDy

Protein structure analysis

BioPython

Sequence handling, alignments

10K+
Compounds Processed
85%+
Prediction Accuracy
<2hr
Batch Inference
3
GNN Architectures