Quantum molecular similarity: Use of atoms in molecules derived quantities as QSAR variables

We demonstrate how quantum mechanically derived data can be used in QSAR. Quantum Molecular Similarity Measures have utilised the total electron density with which to compare molecules. We have been able to use the properties of Bond Critical Points (BCPs) as defined by the topological theory of Atoms in Molecules (AIM) to accurately summarise and characterise any molecular electron density in a compact manner. Expanding earlier ideas we use 9 properties of the electron density, as measured at BCPs, as QSAR variables. These variables are determined for a series of congeners for which some physical property or activity has been measured. Using the Partial Least Squares (PLS) methodology the variables are fitted to this activity and a QSAR with predictive power is produced. Additionally, PLS highlights the variables which are most important in predicting the activity. To simplify interpretation of the QSAR, we reduce each variable set which is centred on the same BCP to its first Principal Component (PC). These provide us with a quantum chemical fingerprint of a region of molecular electron density which can be directly linked to the chemist's intuitive view of a bond. We apply this methodology to various groups of molecules and cover a range of properties. Physical properties are predicted and interpreted as well as biological and environmentally relevant toxicities and biodegradabilities.