TY  - THES
T1  - Novel approaches for bond order assignment and NMR shift prediction
A1  - Dehof,Anna Katharina
Y1  - 2012/07/30
N2  - Molecular modelling is one of the cornerstones of modern biological and pharmaceutical research. Accurate modelling approaches easily become computationally overwhelming and thus, different levels of approximations are typically employed. In this work, we develop such approximation approaches for problems arising in structural bioinformatics. A fundamental approximation of molecular physics is the classification of chemical bonds, usually in the form of integer bond orders. Many input data sets lack this information, but several problems render an automated bond order assignment highly challenging. For this task, we develop the BOA Constructor method which accounts for the non-uniqueness of solutions and allows simple extensibility. Testing our method on large evaluation sets, we demonstrate how it improves on the state of the art. Besides traditional applications, bond orders yield valuable input for the approximation of molecular quantities by statistical means. One such problem is the prediction of NMR chemical shifts of protein atoms. We present our pipeline NightShift for automated model generation, use it to create a new prediction model called Spinster, and demonstrate that it outperforms established, manually developed approaches. Combining Spinster and BOA Constructor, we create the Liops-model that for the first time allows to efficiently include the influence of non-protein atoms. Finally, we describe our work on manual modelling techniques, including molecular visualization and novel input paradigms.
KW  - Bioinformatik
KW  - Biomolekül
KW  - NMR-Spektrum
KW  - Chemische Bindung
KW  - Visualisierung 
CY  - Saarbrücken
PB  - Universitäts- und Landesbibliothek
SN  - 978-3-8482-0773-2
AD  - Postfach 151141, 66041 Saarbrücken
UR  - http://scidok.sulb.uni-saarland.de/volltexte/2012/4869
ER  -