BACKGROUND: Molecular recognition in proteins occurs due to appropriate arrangements of physical, chemical, and geometric properties of an atomic surface. Similar surface regions should create similar binding interfaces. Effective methods for comparing surface regions can be used in identifying similar regions, and to predict interactions without regard to the underlying structural scaffold that creates the surface. RESULTS: We present a new descriptor for protein functional surfaces and algorithms for using these descriptors to compare protein surface regions to identify ligand binding interfaces. Our approach uses descriptors of local regions of the surface, and assembles collections of matches to compare larger regions. Our approach uses a variety of physical, chemical, and geometric properties, adaptively weighting these properties as appropriate for different regions of the interface. Our approach builds a classifier based on a training corpus of examples of binding sites of the target ligand. The constructed classifiers can be applied to a query protein providing a probability for each position on the protein that the position is part of a binding interface. We demonstrate the effectiveness of the approach on a number of benchmarks, demonstrating performance that is comparable to the state-of-the-art, with an approach with more generality than these prior methods. CONCLUSIONS: Local functional descriptors offer a new method for protein surface comparison that is sufficiently flexible to serve in a variety of applications.

In this paper, we present an approach to creating illustrations of molecular flexibility using normal mode analysis (NMA). The output of NMA is a collection of points corresponding to the locations of atoms and associated motion vectors, where a vector for each point is known. Our approach abstracts the complex object and its motion by grouping the points, models the motion of each group as an affine velocity, and depicts the motion of each group by automatically choosing glyphs such as arrows. Affine exponentials allow the extrapolation of nonlinear effects such as near rotations and spirals from the linear velocities. Our approach automatically groups points by finding sets of neighboring points whose motions fit the motion model. The geometry and motion models for each group are used to determine glyphs that depict the motion, with various aspects of the motion mapped to each glyph. We evaluated the utility of our system in real work done by structural biologists both by utilizing it in our own structural biology work and quantitatively measuring its usefulness on a set of known protein conformation changes. Additionally, in order to allow ourselves and our collaborators to effectively use our techniques we integrated our system with commonly used tools for molecular visualization.

Dynamic properties are functionally important in many proteins, including the enzyme adenylate kinase (AK), for which the open/closed transition limits the rate of catalytic turnover. Here, we compare our previously published coarse-grained (double-well Gō) simulation of mesophilic AK from E. coli (AKmeso) to simulations of thermophilic AK from Aquifex aeolicus (AKthermo). In AKthermo, as with AKmeso, the LID domain prefers to close before the NMP domain in the presence of ligand, but LID rigid-body flexibility in the open (O) ensemble decreases significantly. Backbone foldedness in O and/or transition state (TS) ensembles increases significantly relative to AKmeso in some interdomain backbone hinges and within LID. In contact space, the TS of AKthermo has fewer contacts at the CORE-LID interface but a stronger contact network surrounding the CORE-NMP interface than the TS of AKmeso. A "heated" simulation of AKthermo at 375K slightly increases LID rigid-body flexibility in accordance with the "corresponding states" hypothesis. Furthermore, while computational mutation of 7 prolines in AKthermo to their AKmeso counterparts produces similar small perturbations, mutation of these sites, especially positions 8 and 155, to glycine is required to achieve LID rigid-body flexibility and hinge flexibilities comparable to AKmeso. Mutating the 7 sites to proline in AKmeso reduces some hinges' flexibilities, especially hinge 2, but does not reduce LID rigid-body flexibility, suggesting that these two types of motion are decoupled in AKmeso. In conclusion, our results suggest that hinge flexibility and global functional motions alike are correlated with but not exclusively determined by the hinge residues. This mutational framework can inform the rational design of functionally important flexibility and allostery in other proteins toward engineering novel biochemical pathways.

Local shape descriptors compactly characterize regions of a surface, and have been applied to tasks in visualization, shape matching, and analysis. Classically, curvature has be used as a shape descriptor; however, this differential property characterizes only an infinitesimal neighborhood. In this paper, we provide shape descriptors for surface meshes designed to be multi-scale, that is, capable of characterizing regions of varying size. These descriptors capture statistically the shape of a neighborhood around a central point by fitting a quadratic surface. They therefore mimic differential curvature, are efficient to compute, and encode anisotropy. We show how simple variants of mesh operations can be used to compute the descriptors without resorting to expensive parameterizations, and additionally provide a statistical approximation for reduced computational cost. We show how these descriptors apply to a number of uses in visualization, analysis, and matching of surfaces, particularly to tasks in protein surface analysis.

MOTIVATION: For many biotechnological purposes, it is desirable to redesign proteins to be more structurally and functionally stable at higher temperatures. For example, chemical reactions are intrinsically faster at higher temperatures, so using enzymes that are stable at higher temperatures would lead to more efficient industrial processes. We describe an innovative and computationally efficient method called Improved Configurational Entropy (ICE), which can be used to redesign a protein to be more thermally stable (i.e. stable at high temperatures). This can be accomplished by systematically modifying the amino acid sequence via local structural entropy (LSE) minimization. The minimization problem is modeled as a shortest path problem in an acyclic graph with nonnegative weights and is solved efficiently using Dijkstra's method.

MOTIVATION: One bottleneck in high-throughput protein crystallography is interpreting an electron-density map, that is, fitting a molecular model to the 3D picture crystallography produces. Previously, we developed ACMI (Automatic Crystallographic Map Interpreter), an algorithm that uses a probabilistic model to infer an accurate protein backbone layout. Here, we use a sampling method known as particle filtering to produce a set of all-atom protein models. We use the output of ACMI to guide the particle filter's sampling, producing an accurate, physically feasible set of structures. RESULTS: We test our algorithm on 10 poor-quality experimental density maps. We show that particle filtering produces accurate all-atom models, resulting in fewer chains, lower sidechain RMS error and reduced R factor, compared to simply placing the best-matching sidechains on ACMI's trace. We show that our approach produces a more accurate model than three leading methods--Textal, Resolve and ARP/WARP--in terms of main chain completeness, sidechain identification and crystallographic R factor. AVAILABILITY: Source code and experimental density maps available at http://ftp.cs.wisc.edu/machine-learning/shavlik-group/programs/acmi/

We present a novel modular approach to integrating multiple input/output (I/O) modes in a virtual environment that imitate natural, intuitive and effective human interaction behavior. The I/O modes used in this research are spatial tracking of both hands, finger gesture recognition, head/body spatial tracking, voice recognition (discrete recognition for simple commands, and continuous recognition for natural language input), immersive stereo display and synthesized speech output. Intuitive natural interaction is achieved through several stages: identifying all the tasks that need to be performed, grouping similar tasks and assigning them to a particular mode such that it imitates the physical world. This modular approach allows inclusion and removal of additional input and output modes as well as additional users. We described this multimodal interaction paradigm by applying it to a real world application: visualizing, modeling and fitting protein molecular structures in an immersive virtual environment.