PAULYPredicting Chemical Properties Before Testing
PAULY builds predictive models for physico-chemical properties, environmental fate, and toxicity — so data exists before a single test tube is filled.
How it works
From molecule to model
From Molecule to Code
Every prediction begins with the chemical structure of a substance. This is represented as SMILES code—a compact text notation that uniquely describes a molecule’s atoms, bonds, and rings.
Structure Becomes a Graph
To perform calculations on the molecule, Pauly translates the structure into a mathematical representation: atoms become nodes, and bonds become edges. This creates a graph that a model can process.
Three Approaches to Prediction
Depending on the endpoint, Pauly utilizes GNNs to analyze molecular graphs, GCA models based on specific molecular fragments, or kNN algorithms that compare structural similarity against established reference data.
From Molecule to Predicted Value
The model translates the molecular representation into concrete predictions of physico-chemical properties. The resulting metrics can be flexibly integrated into individual use cases and toxicological assessments.