On the use of information theory for assessing molecular diversity

In a recent article published in Molecules, Lin presented a novel approach for assessing molecular diversity based on Shannon's information theory. In this method, a set of compounds is viewed as a static collection of microstates which can register information about their environment at some-predetermined capacity. Diversity is directly related to the information conveyed by the population, as quantified by Shannon's classical entropy equation. Despite its intellectual appeal, this method is characterized by a strong tendency to oversample remote areas of the feature space and produce unbalanced designs. This paper demonstrates this limitation with some simple examples and provides a rationale for the failure of the method to produce results that are consistent with other traditional methodologies.