Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.751790
Title: The application of data processing and artificial intelligence technique to pesticide research
Author: Saggers, D. T.
Awarding Body: University of Surrey
Current Institution: University of Surrey
Date of Award: 1978
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
To date the discovery of pesticides has depended largely upon random synthesis and serendipity. However, there is now, for a variety of reasons, a strong incentive to rationalise the inventive process. To this end techniques have been investigated whereby routinely aquired biological data may be subjected to numerical analysis with a view to establishing empirical relationships between chemical structure and pesticidal activity. These procedures have been aimed at the discovery of entirely new chemical families and the rapid optimisation of structure within families. The study initially required the codification of biological observations and chemical structures for computer recording and processing; the preferred procedure for structure recording being the Wiswesser Line Notation (WLN). Use of a permuted WLN index, in conjunction with a computer summary of the biological results, proved to be a simple, but valuable, tool for providing structure-activity guides to previously defined areas of chemistry. A method of summarising biological activity, the Categorisation Analysis (CA) was developed. This used in combination with a permuted WLN index, showed the range of biological response produced by structural variants of a common substructural feature, whilst the reverse procedure illustrated the variations in structure which produced the same spectrum of biological response. Biological activity coefficients for specific responses : Structure Activity Frequency (SAF) values : were obtained automatically for WLN-derived substructural fragments, in a set of compounds of known activity, by computing the proportion of compounds in the set possessing the fragment,which were active. A highly significant relationship was established between the mean SAF (MSAF) value for the fragments of a compound and its activity. From this it was possible to predict the activity of further untested compounds using their computed MSAF values. It was also possible to utilize the results as a lead generative procedure by identifying combinations and associations of fragments with high SAF values. In structure optimisation studies, modifications of the Free-Wilson model were found to give reliable predictions, provided an appropriate transformation of the computer stored data was carried out. A procedure was developed for automatic chemical fragment generation from the WLN for use in the Free-Wilson data matrix. Throughout the studies it was apparent that adequate definition of the biological response and chemical fragments was essential for good structure-activity prediction and the relevance of these findings to further improvements is discussed.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.751790  DOI: Not available
Share: