Year: 2009
Type: dissertation
Abstract: A drug is typically a small molecule that interacts with the binding site of some \ntarget protein. Drug design involves the optimization of this interaction so that the \ndrug effectively binds with the target protein while not binding with other proteins \n(an event that could produce dangerous side effects). Computational drug design \ninvolves the geometric modeling of drug molecules, with the goal of generating \nsimilar molecules that will be more effective drug candidates. It is necessary that \nalgorithms incorporate strategies to measure molecular similarity by comparing \nmolecular descriptors that may involve dozens to hundreds of attributes. We use \nkernel-based methods to define these measures of similarity. Kernels are general \nfunctions that can be used to formulate similarity comparisons. \n \nThe overall goal of this thesis is to develop effective and efficient computational \nmethods that are reliant on transparent mathematical descriptors of molecules with \napplications to affinity prediction, detection of multiple binding modes, and generation \nof new drug leads. While in this thesis we derive computational strategies for \nthe discovery of new drug leads, our approach differs from the traditional ligandbased \napproach. We have developed novel procedures to calculate inverse mappings \nand subsequently recover the structure of a potential drug lead. The contributions \nof this thesis are the following: \n \n1. We propose a vector space model molecular descriptor (VSMMD) based on \na vector space model that is suitable for kernel studies in QSAR modeling. \nOur experiments have provided convincing comparative empirical evidence \nthat our descriptor formulation in conjunction with kernel based regression \nalgorithms can provide sufficient discrimination to predict various biological \nactivities of a molecule with reasonable accuracy. \n \n2. We present a new component selection algorithm KACS (Kernel Alignment \nComponent Selection) based on kernel alignment for a QSAR study. Kernel \nalignment has been developed as a measure of similarity between two kernel \nfunctions. In our algorithm, we refine kernel alignment as an evaluation tool, \nusing recursive component elimination to eventually select the most important \ncomponents for classification. We have demonstrated empirically and proven \ntheoretically that our algorithm works well for finding the most important \ncomponents in different QSAR data sets. \n \n3. We extend the VSMMD in conjunction with a kernel based clustering algorithm \nto the prediction of multiple binding modes, a challenging area of \nresearch that has been previously studied by means of time consuming docking \nsimulations. The results reported in this study provide strong empirical \nevidence that our strategy has enough resolving power to distinguish multiple \nbinding modes through the use of a standard k-means algorithm. \n \n4. We develop a set of reverse engineering strategies for QSAR modeling based \non our VSMMD. These strategies include: \n \n(a) The use of a kernel feature space algorithm to design or modify descriptor \nimage points in a feature space. \n(b) The deployment of a pre-image algorithm to map the newly defined \ndescriptor image points in the feature space back to the input space of \nthe descriptors. \n(c) The design of a probabilistic strategy to convert new descriptors to meaningful \nchemical graph templates. \n \nThe most important aspect of these contributions is the presentation of strategies that actually generate the structure of a new drug candidate. While the training \nset is still used to generate a new image point in the feature space, the reverse engineering \nstrategies just described allows us to develop a new drug candidate that is \nindependent of issues related to probability distribution constraints placed on test \nset molecules.
Source: UWSpace (University of Waterloo)
Author: William Wai Lun Wong
Language: English
Subfield: Drug Discovery
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W2529291290