Geometrically centered region: A "wet" model of protein binding hot spots not excluding water molecules

Publication Type:
Journal Article
Proteins: Structure, Function and Bioinformatics, 2010, 78 (16), pp. 3304 - 3316
Issue Date:
Full metadata record
A protein interface can be as "wet" as a protein surface in terms of the number of immobilized water molecules. This important water information has not been explicitly taken by computational methods to model and identify protein binding hot spots, overlooking the water role in forming interface hydrogen bonds and in filing cavities. Hot spot residues are usually clustered at the core of the protein binding interfaces. However, traditional machine learning methods often identify the hot spot residues individually, breaking the cooperativity of the energetic contribution. Our idea in this work is to explore the role of immobilized water and meanwhile to capture two essential properties of hot spots: the compactness in contact and the far distance from bulk solvent. Our model is named geometrically centered region (GCR). The detection of GCRs is based on novel tripartite graphs, and atom burial levels which are a concept more intuitive than SASA. Applying to a data set containing 355 mutations, we achieved an F measure of 0.6414 when δδG ≥ 1.0 kcal/mol was used to define hot spots. This performance is better than Robetta, a benchmark method in the field. We found that all but only one of the GCRs contain water to a certain degree, and most of the outstanding hot spot residues have water-mediated contacts. If the water is excluded, the burial level values are poorly related to the δδG, and the model loses its performance remarkably. We also presented a definition for the O-ring of a GCR as the set of immediate neighbors of the residues in the GCR. Comparative analysis between the O-rings and GCRs reveals that the newly defined O-ring is indeed energetically less important than the GCR hot spot, confirming a long-standing hypothesis. Proteins 2010. © 2010 Wiley-Liss, Inc.
Please use this identifier to cite or link to this item: