Wendy Warr interviews Dr. Han van de Waterbeemd

WW: Has anyone been your inspiration, personally?
HvdW: Probably two names stand out. I met Hugo Kubinyi during my PhD and we still are good friends. Then I did my post doc with Bernard Testa in Lausanne and worked in his lab for 8 years. We still work together on book editing projects e.g. ADME-Tox Approaches in Comprehensive Medicinal Chemistry, the bible in the field! I also thank John Dearden for running a number of miles together.

WW: The van de Waterbeemd data set was first published in 1987. Is it still used? Are disjoint principal properties (DPPs) still used as descriptors?
HvdW: I forgot about this! You probably refer to the substituent values we collected and analyzed. We showed that lipophilicity could be seen as composed of a size (volume) and a polarity (H-bonding) term. This is a very useful insight for the medicinal chemist. DPP was a concept developed by Sergio Clementi, but it never took off.

WW: A recent paper of yours recommends using more current data for making predictions and suggests a need for auto-updating QSAR models.
HvdW: There is very little literature on the behavior of QSAR models over time. At AstraZeneca we are doing some systematic investigations on this (Rodgers, S. L.; Davis, A. M.; van de Waterbeemd, H. Time-Series QSAR Analysis of Human Plasma Protein Binding Data QSAR & Combinatorial Science 2007, 26(4), 511-521). Over time, projects move away in chemical space from the original training set and so the models age and predictions get worse. We therefore need ways of rebuilding the models automatically to make maximum use of more recent experimental measurements. We, and others are working on this.

WW: I see that you are an IUPAC Fellow. I assume, therefore, that you have made significant contributions to IUPAC.
HvdW: I am a co-author of the glossary of terms used in computational drug design . I also contributed to the glossary of terms in medicinal chemistry, in a team led by Camille Wermuth.