Use this URL to cite or link to this record in EThOS:
Title: Studies on molecular similarity
Author: Bender, A.
Awarding Body: University of Cambridge
Current Institution: University of Cambridge
Date of Award: 2007
Availability of Full Text:
Full text unavailable from EThOS.
Please contact the current institution’s library for further details.
In this thesis, two conceptually different approaches are implemented and explored. The first of the approaches is based on the molecular connectivity table, combining circular fingerprints (similar to ‘Augmented Atoms’), information-gain based feature selection and the Naïve Bayesian Classifier for model generation. On two standard datasets, this method outperformed many other established similarity searching methods. While both similar descriptors and Bayesian-like fragment weighting schemes have been explored previously, both the evaluation on standard datasets and the emphasis of the importance of feature selection in combination with the Bayes Classifier are novel to the work presented here. The second of the approaches presented is based on force-field (GRID) derived energetic information about the surface of the molecule, locally capturing interaction profiles of possible ligand-receptor interactions, such as hydrophobic, charge and hydrogen bond interactions. This descriptor uses points relative to the molecular coordinates, thus it is translationally and rotationally invariant. Due to the local nature of the descriptor, conformational variations are seen to cause only minor changes in the descriptor. Compared to other approaches, it performs well with respect to the retrieval of active structures and selected features are shown to correspond to binding patterns observed crystallographically. In addition to employing force-field probes the application of quantum-mechanically derived surface descriptors (defined via the COSMO continuum salvation model) was explored. The screening charges calculated here also define properties relevant to ligand-target binding such as hydrogen bond donors and acceptors, positive and negative charges and lipophilic moieties from first principles. Encoding of properties is performed by three-point pharmacophores (3PP) which were found to outperform other approaches.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available