A new measure is introduced to assess the complexity of organic molecules in drug discovery. It is common knowledge in medicinal chemistry that molecules need a certain complexity to be successful in the drug discovery process. The presented method calculates the self similarity s and complexity c of molecular graphs. An algorithm was developed to calculate the total number of connected subgraphs m, the number of unique connected subgraphs n and the number of non overlapping identical subgraphs o. A straightforward function for the complexity of a molecule was defined c = f(m, n, o, p), with p equal to the number of bonds between heavy atoms. Five datasets with 50 molecules each were analyzed. Each molecule contained 25 up to 27 heavy (non hydrogen) atoms and between 25 to 30 bonds. The datasets were compiled from bioactive natural products, approved drugs, highly bioactive molecules, commercially available compounds for virtual screening and artificial generated molecules. Comparing the median of c for the five datasets showed a significant increase in the following order: commercially available compounds < bioactive molecules < approved drugs << natural products < artificial molecules. It is striking that the complexity of natural products is much higher than the complexity of synthesized molecules. This has been observed by medicinal chemists since decades, without having an easy measure to quantify complexity. With the introduced complexity value c a meaningful figure of merit was developed to assess automatically the complexity of single compounds and compound libraries in drug discovery. Details for the calculation of s and c, and the results of their application to the test datasets will be reported here.

Measuring complexity and self-similarity of chemical structures in drug discovery

245th National Meeting (2013)

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