Traditional molecular modeling is performed at atomic resolution, which relies on X-ray and NMR experiments to provide structural information. When deal with biomolecular assemblies of millions of atoms, atomic description of molecular objects becomes very computational inefficient. We developed a method that uses map objects for molecular modeling to efficiently derive structural information from experimental maps, as well as conveniently manipulate map objects, perform conformational search directly using map objects. This development work has been implemented into CHARMM as the EMAP module. This implementation enables CHARMM to manipulate map objects, including map input, output, comparison, docking, etc. Other experiment such as transition metal ion FRET (tmFRET) is becoming a useful way to obtain protein structure information. A new focus of our research is to combine efficient simulation technique with structural information from experiment to assist high throughput protein structure determination. Structure determination from low resolution EM maps We developed a map-restrained self-guided Langevin dynamics (MapSGLD) simulation method for efficient targeted conformational search. The targeted conformational search represents simulations under restraints defined by experimental observations and/or by user specified structural requirements. Through map-restraints, this method provides an efficient way to maintain substructures and to set structure targets during conformational searching. With an enhanced conformational searching ability of self-guided Langevin dynamics, this approach is suitable for simulating large-scale conformational changes, such as the formation of macromolecular assemblies and transitions between different conformational states. A direct application of this method is to determine macromolecular structures by flexible fitting of atomic structures into density maps derived from cryo-electron microscopy. Molecular basis of Chemotaxi In chemotaxic bacteria, transmembrane chemoreceptors, CheA and CheW form the core signaling complex of the chemotaxis sensory apparatus. These complexes are organized in extended arrays in the cytoplasmic membrane that allow bacteria to respond to changes in concentration of extracellular ligands via a cooperative, allosteric response that leads to substantial amplification of the signal induced by ligand binding. Here, we have combined cryo-electron tomography studies of the 3D spatial architecture of chemoreceptor arrays in intact E. coli cells with computational modeling to develop a predictive model for the cooperativity and sensitivity of the chemotaxis response. The predictions were tested experimentally using fluorescence resonance energy transfer (FRET) microscopy. Our results demonstrate that changes in lateral packing densities of the partially ordered, spatially extended chemoreceptor arrays can modulate the bacterial chemotaxis response, and that information about the molecular organization of the arrays derived by cryo-electron tomography of intact cells can be translated into testable, predictive computational models of the chemotaxis response. EM study of maturation of the HIV-1 core The formation of the HIV-1 core is the final step in the viral maturation pathway, resulting in the formation of infectious virus. Most current models for HIV-1 core formation suggest that, upon proteolytic cleavage from the immature Gag, capsid (CA) dissociates into the viral interior before reforming into the core. Here we present evidence for an alternate view of core formation by taking advantage of our serendipitous observation of large membrane-enclosed structures in HIV-1 supernatants from infected cells. Cryo-electron tomographic studies show that these structures, which contain ordered arrays of what is likely the membrane-associated matrix protein, contain multiple cores that can be captured at different stages of maturation. Our studies suggest that HIV maturation involves a non-diffusional phase transition in which the detaching layer of the cleaved CA lattice is gradually converted into a roll that ultimately forms the surface of the mature conical core. Investigating the global organization of a binding site network Ligand binding sites are critical in regulating an enzyme activity and signal transduction. Comparing similarities between binding sites have been widely used for drug discovery, protein engineering, and function prediction. However, unlike protein domain structures, the global organization of binding sites has not yet been address clearly; how diverse are they or do similar binding sites guarantee similar ligands or functions? We constructed a global weighted network of binding sites based on their all-to-all similarities and detected the communities of similar sites. We found that all binding sites can be categorized into 303 communities. The largest community consists mainly of metal binding sites; especially zinc-binding sites, and plays a role of structural hub of other sites. We also found that the sizes of communities follow a power-law distribution; a few dominantly large and many small communities co-exist, which implies that the metal-binding sites may be the most ancient binding sites. Prediction of the variability of spatial restraints in template-based modeling by random forest In template-based modeling when using a single template, inter-atomic distances of an unknown protein structure are assumed to be distributed by Gaussian probability density functions, whose center peaks are located at the distances between corresponding atoms in the template structure. The width of the Gaussian distribution, the variability of a spatial restraint, is closely related to the reliability of the restraint information extracted from a template, and it should be accurately estimated for successful template-based protein structure modeling. To predict the variability of the spatial restraints in template-based modeling, we have devised a prediction model, Sigma-RF, by using the random forest (RF) algorithm. The benchmark results on 22 CASP9 targets show that the variability values from Sigma-RF are of higher correlations with the true distance deviation than those from Modeller. We assessed the effect of new sigma values by performing the single-domain homology modeling of 22 CASP9 targets and 24 CASP10 targets. For most of the targets tested, we could obtain more accurate 3D models from the identical alignments by using the Sigma-RF results than by using Modeller ones. We find that the average alignment quality of residues located between and at two aligned residues, quasi-local information, is the most contributing factor, by investigating the importance of input features used in the RF machine learning. This average alignment quality is shown to be more important than the previously identified quantity of local information: the product of alignment qualities at two aligned residues.
Showing the most recent 10 out of 21 publications
