| The objective of this thesis was to develop and apply four-dimensional quantitative structure-activity methodologies to model ligand-receptor systems without knowledge of the receptor geometry and use the models as a virtual high throughput screen to identify new lead chemistry. Virtual high throughput screening is a powerful tool for identifying and predicting the activities of compounds in a virtual combinatorial library. The use of VHTS has allowed chemists to save time and money by using computers to help define the geometries, reactivities, and activity profiles for compounds, which help guide the selection process. Even when the detailed structural information about the target receptor is not available it is possible to derive an abstract model indicating the key features of a series of active molecules by suggesting compounds that would be expected to interact favorably with the receptor, or compounds that contain the required functional groups of the pharmacophore.;It has been shown that there is more than one way to use the 4D-QSAR modeling as a VHTS. The Additive Enhancement model may be indicative of high binding specificity, while a Consensus model paradigm may be useful for molecules with multiple binding sites. Although, it is important to remember that the training set and the corresponding range of biological activities bias the model or models produced by 4D-QSAR methodologies. However, as the new leads are synthesized and their biological activity determined, these molecules can be added to the training set and used to update the 4D-QSAR model. In the absence of one consistent model developed in the 4D-QSAR paradigm, it was shown that a Consensus Selection approach can also give useful insight to the properties and/or characteristics of molecules which increase or decrease activity. This information can then be used to further explore the properties determined useful by the model(s). |
