| With the cost of drug development on the rise, significant effort in pharmaceutical research is being directed toward streamlining, and thus lowering the cost, of the drug development process. To that end, this study focuses on deriving and validating two novel concepts, namely the manifold model theory and universal descriptors that would aid in accelerating the drug discovery process. The first concept, manifold model theory, refers to the realization of multiple solutions, each unique in nature, for the modeling of a single molecular system resulting in a more robust solution. The second concept, universal descriptors, relates to deriving descriptors of compounds that embed three-dimensional and conformational information, yet remain independent of alignment, and thus reducing analysis time.; Various data sets of drug-like compounds were used to validate the concepts. The results of the manifold model theory validation, using two sets of human immunodeficiency virus protease inhibitors, illustrated the importance of considering models in parallel by identifying important ligand-receptor binding interactions. Five data sets, varying in their chemical and statistical properties, were employed in the validation scheme for the universal descriptors. The results showed that the statistical measures of fit and predictive power of the models generated using the alignment-independent universal descriptors were of equal or greater quality when compared to various established methods that rely on alignment dependent-descriptors. |
