Protein Surface Search in DNA-binding Protein Prediction by Delaunay Triangulation Modeling

Abstract

Various protein researches have been published, such as transcription factor prediction, DNA-binding protein prediction, homology modeling, and protein-protein interaction. Structure alignment techniques have played an important role in those researches. Many protein alignment methods have been brought up, yet only few have concentrated on protein surfaces. Among this research, we have developed a DNA-binding protein prediction system based on protein surface search, using Voronoi diagram and Delaunay triangulation to model molecular surface. Also, we designed an iterative method to construct a consecutive and closed surface of protein. Finally, a system integrating surface structure information is applied to search common surface of proteins, and to predict protein-DNA interaction. While comparing to our previous method, there’s a significant increase of inner triangles identified and removed. Furthermore, experimental data showed that our method has better performance in identifying surface residues than the widely used method using an RASA cutoff. Besides, as we search the entire Protein Data Bank (PDB) for similar surface to Estrogen Receptor α (ERα), we discovered that the majority of the family of ERα in Pfam has been found with high scores. Finally, we compared our data with results of chemical experiment that suggest several proteins tending to bind to Estrogen Response Elements (ERE), finding an interesting result that the 14-3-3β protein with high score in our prediction is proven to bind to ERE in the further chemical experiment.