Matched Molecular Pairs Analysis (MMPA) is a technique that links differences in molecular structure with changes in properties and so allows useful information to be extracted and shared without disclosing full structures. By encoding the output as SMIRKS, it can then be used to suggest new molecules. It also provides the probability of each property changing in the desired direction. Like picking the winning horse by studying its previous results and current race conditions, the decision can be logical and data driven but success cannot be guaranteed.<br/><br/>MMPA has recently been applied to the ADMET databases of Roche, AstraZeneca and Genentech and the output combined to create what we believe is the world’s largest repository of medicinal chemistry knowledge, akin to a textbook. A by-product of this merging is that a large set of MMPA data are available that can answer fundamental questions about the technique: i) how many pairs are needed in order to be confident that a particular change in structure actually causes an increase (or decrease) in property? ii) for small sets of molecules, how large must the average change in property be in order to be confident that a particular change in structure actually causes an increase (or decrease) in property? iii) is there an upper limit to the number of pairs needed? iv) do chemically specific changes need less data than general changes? Each of these questions will be addressed.<br/><br/>A starting point for the analysis is that each pair can be viewed as a coin-flip experiment where an increase in a property corresponds to “heads” and a decrease to “tails”. This reduces the challenge to detecting biased coins. The coin flip analogy reduces the likelihood of misassigning the direction of change, but greatly reduces the number of structural transformations that are identified as having a significant effect upon properties. The large database available has permitted us to probe the link between the coin flip approach and the mean changes in property. The encoding of the chemical environment has further permitted us to analyse its impact which is to reduce the number of pairs required as the chemical environment becomes more specific.

Quantifying the effect that chemical environment exerts upon changes in property in matched molecular pairs analysis

251st National Meeting (2016)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.