| Statistical models of the three-dimensional protein structural environments conducive to DNA-binding were created. These models were of two types: one able to discern DNA-binding alpha helices and the second able to discern DNA-binding beta strands. When combined, the models were used to search through the entire Protein Data Bank and identify structures that are likely to bind to DNA and possible locations on those structures at which DNA-binding may occur.; Statistical models were also created for discerning specific DNA-binding structural motifs. These models were used to automatically discern each DNA-binding motif, and proved to be useful tools that can be used to identify previously undocumented examples.; The measurements used to create each of the statistical models described were novel and are proposed to be general measurements of protein structure that can be used to discern proteins of a given ligand binding class (such as DNA-binding) from protein structures not in that class. Due to generality of the measurements, they can be viewed as effectively reducing the three-dimensional coordinates of a protein structure into a list of measurements in a simple table format. In such format state-of-the-art statistical methodologies can be employed to classify protein structures, including DNA-binding protein structures, as shown here. |
