PAULYGroup-Contribution Approach (GCA)
A group-contribution approach — often labelled GCA (Quantitative Structure-Property Relationship) — predicts a property by breaking a molecule down into known structural fragments, such as functional groups, and summing each fragment’s individually known contribution to the property.
This makes group-contribution models simpler and more transparent than a neural network: every prediction can be traced back to exactly which fragments drove it. The trade-off is coverage — the approach only works reliably when a molecule can be decomposed into fragments the model has already seen and calibrated against, so it tends to be less flexible for structurally unusual chemicals than a graph-based model.