A Quantitative Approach to the Estimation of Chemical Space from a Given Geometry by the Combination of Atomic Species

In order to estimate the size of chemical spaces, a quantitative approach was employed consisting of summing all possible combinations of atomic species generated computationally for a given three-dimensional geometry. The atomic species were comprised of heavy atoms (N, C, O, S and Cl) and various numbers of H atoms (e.g., CH3 and CH2 groups). The assignment of atomic species to a given geometry maintained consistent bond orders between adjacent atoms. Between 108 and 1019 compounds were generated from individual geometries. A slightly smaller number of compounds were obtained by applying a nonleadlikeness filter that considered the properties of the substituents. Principal component analysis using descriptors representing the molecular structures and properties revealed that the compounds generated from only one geometry largely covered the chemical space of known compounds. Moreover, the compounds with the best scores, defined as the interaction energies between the compounds and the protein, contained original ones within the entire collection of compounds generated from the same geometry. This method can be used to improve the scaffold of compound structures during drug development.