Show simple item record

dc.contributor.authorBraunheim, Benjamin Blue
dc.date.accessioned2018-07-12T18:56:51Z
dc.date.available2018-07-12T18:56:51Z
dc.date.issued1999
dc.identifier.citationSource: Dissertation Abstracts International, Volume: 61-02, Section: B, page: 8340.;Advisors: Steven D. Schwartz.
dc.identifier.urihttps://yulib002.mc.yu.edu/login?url=http://gateway.proquest.com/openurl?url_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:dissertation&res_dat=xri:pqm&rft_dat=xri:pqdiss:9961301
dc.identifier.urihttps://hdl.handle.net/20.500.12202/3833
dc.description.abstractBiological molecular recognition is central to all attempts to control biological processes. I have been studying biological recognition in the context of small molecules (inhibitors, substrates) that interact with enzymes or receptors. The binding energy of an inhibitor with the active site is a complicated function determined by the quantum mechanical features of both the enzyme and the inhibitor. This function could never be solved explicitly so we have devised methods that exploit artificial neural network learning algorithms to approximate the function defined by the relationship between the enzyme and inhibitor. The relationship between inhibitors and the enzyme active site is an inherently complicated one because enzymes have evolved to be highly structurally specific with regards to the molecules they bind. Enzymes exercise their selectivity through subtle conformational variations that can change the geometric and electronic character of the active site. Modeling this dynamic process is very difficult. The central hypothesis of my work is that if we assume that similar molecules will interact with the enzyme by making similar contacts and that the degree to which the contacts are identical is governed by some function of the inhibitors' structure then we could hope to model this system based solely on comparisons between the inhibitors.;Our first attempt to model the relationship between inhibitors and an enzyme active site was based on the principle that similar molecules will have similar bioactivity. We rigorously developed the concept of "similarity" for molecules involved in biological processes. This approach was largely successful, but we found that there were many cases where similar molecules had quite different bioactivities, and vice versa. This finding led us to the development of an artificial neural network learning method that was able to learn what quantum mechanical features were important in binding. A finding of even greater importance was that the neural network was able to understand that features of the molecules worked together to determine binding and that certain combinations could have drastic effects. The neural network was able to predict the bioactivity of molecules more accurately than any previous method.;My latest work focuses on the idea that a trained neural network knows all the rules concerning molecules binding to the enzyme. With this assumption we have developed a technique that places this trained neural network in an active role in the design of new compounds that could be more effective than any known compound. These methodologies are important for the basic study of molecular recognition, and for practical use in drug design.
dc.publisherProQuest Dissertations & Theses
dc.subjectBiochemistry.
dc.subjectBiophysics.
dc.titleTheoretical studies of inhibitor recognition
dc.typeDissertation


Files in this item

FilesSizeFormatView

There are no files associated with this item.

This item appears in the following Collection(s)

Show simple item record