Characterizing the microenvironment surrounding protein sites

Abstract

Sites are microenvironments within a biomolecular structure, distinguished by their structural or functional role. A site can be defined by a three-dimensional location and a local neighborhood around this location in which the structure or function exists. We have developed a computer system to facilitate structural analysis (both qualitative and quantitative) of biomolecular sites. Our system automatically examines the spatial distributions of biophysical and biochemical properties, and reports those regions within a site where the distribution of these properties differs significantly from control nonsites. The properties range from simple atom-based characteristics such as charge to polypeptide-based characteristics such as type of secondary structure. Our analysis of sites uses nonsites as controls, providing a baseline for the quantitative assessment of the significance of the features that are uncovered. In this paper, we use radial distributions of properties to study three well-known sites (the binding sites for calcium, the milieu of disulfide bridges, and the serine protease active site). We demonstrate that the system automatically finds many of the previously described features of these sites and augments these features with some new details. In some cases, we cannot confirm the statistical significance of previously reported features. Our results demonstrate that analysis of protein structure is sensitive to assumptions about background distributions, and that these distributions should be considered explicitly during structural analyses.Keywords: biophysical properties; calcium binding; computational biology; disulfide bridges; microenvironment; protein structure analysis; serine proteases; software Central to molecular biology is the determination of macromolecular structure and the analysis of how structural elements produce an observed function. The principles by which structure relates to function have been elucidated in a piecemeal fashion, from work on single structures or small classes of structures. Computational assistance has come primarily in the form of graphical methods for scientific visualization and from special purpose programs for analyzing individual biophysical properties (such as solvent accessibility or electrostatic fields). Unfortunately, studying structures individually entails a risk of missing important relationships that would be revealed by pooling relevant data. The expected surfeit of protein structures provides an opportunity to develop tools for automatically examining biological structures and producing useful representations of the key biophysical and biochemical features. The utility of a general purpose system for producing these representations would extend from medical/pharmaceutical applications (model-based drug design, comparing pharmacological Reprint requests to: Russ B. Altman, Section on Medical Informatics, Stanford University School of Medicine, MSOB X-215, Stanford, California 94305-5479; e-mail: altman@camis.stanford.edu. activities) to ind...

Molecular and Structural Basis for Redox Regulation of β-Actin

“…It has been pointed out that hydroxyl groups or aromatic rings through interactions with the disulfide bond confer stability and restrict accessibility to reducing agents. 20,38 Cys374 is partially solvent-accessible in monomeric and filamentous actin. 36,39 It is oriented towards the protein interior in most monomeric actin structures having intact C termini.…”

Novel function discovery through sequence and structural data mining

“…Structure-based function prediction approaches include docking and template-based structural modelling [71 ,72], detection of geometric and physicochemical similarities among functional sites [73][74][75][76], and structural alignment. These approaches have been successful in predicting a wide range of new protein functions/interactions including enzymatic functions [69,70 ,77], binding sites [78,79], protein-DNA interactions [76,80], and protein-protein complexes [68,71 ].…”

Local Pre-processing for Node Classification in Networks

“…Fixed size microenvironments have been used as a basis for k-means clustering with a view to exploring protein structure [29]. This approach has also been successful in identifying calcium binding sites [6]. The prediction of protein-protein interaction sites has been explored by using combinations of attributes and SVM classifiers [30,11,35].…”