Understanding the true relation between the chemical structure and biological properties of a molecule has always been a great challenge for a medicinal chemist. Is the binding of a ligand solely based on structural fit, or are the physical and chemical property profiles more important? Which descriptors are most relevant?

Interpretable models built using principal components analysis (PCA) and projections to latent structures (PLS) allow the causes for binding to be identified. A major benefit of PLS is the ability to handle multiple Y responses allowing optimisation of primary activity simultaneously with secondary or undesired ADMET properties.

For more detailed information, see the application notes “Statistical Molecular Design”, “Chemical Maps using SIMCA-P”, and “A multivariate QSAR model for CyP450”.