Preparation and characterization of bisphenolato magnesium derivatives: an efficient catalyst for the ring-opening polymerization of epsilon-caprolactone and L-lactide.

A series of magnesium complexes have been prepared and characterized, in which [(EDBP-Me)Mg(mu-OBn)](2) (4) has shown high activity toward the ring-opening polymerization (ROP) of epsilon-caprolactone and L-lactide. 2,2'-Ethylidene-bis(4,6-di-tert-butylphenol)-monomethyl ether ([EDBP-(Me)H]) is prepared by the reaction of 2,2'-ethylidene-bis(4,6-di-tert-butylphenol) (EDBP-H(2)) with dimethyl sulfate. The reaction of [EDBP-(Me)H] with (n)Bu(2)Mg yields [(EDBP-Me)Mg(n)Bu](2), which further reacts with benzyl alcohol and N,N-dimethylethylenediamine giving complexes [(EDBP-Me)Mg(mu-OBn)](2) (4) and {[EDBP(Me)]Mg(mu-Me(2)NCH(2)CH(2)NH)}(2) (5), respectively. Experimental results indicate that the activity of complex 4 toward cyclic esters is higher than that of [(mu-EDBP)Mg](2)/(BnOH).

How well can the accuracy of comparative protein structure models be predicted?

Comparative structure models are available for two orders of magnitude more protein sequences than are experimentally determined structures. These models, however, suffer from two limitations that experimentally determined structures do not: They frequently contain significant errors, and their accuracy cannot be readily assessed. We have addressed the latter limitation by developing a protocol optimized specifically for predicting the Calpha root-mean-squared deviation (RMSD) and native overlap (NO3.5A) errors of a model in the absence of its native structure. In contrast to most traditional assessment scores that merely predict one model is more accurate than others, this approach quantifies the error in an absolute sense, thus helping to determine whether or not the model is suitable for intended applications. The assessment relies on a model-specific scoring function constructed by a support vector machine. This regression optimizes the weights of up to nine features, including various sequence similarity measures and statistical potentials, extracted from a tailored training set of models unique to the model being assessed: If possible, we use similarly sized models with the same fold; otherwise, we use similarly sized models with the same secondary structure composition. This protocol predicts the RMSD and NO3.5A errors for a diverse set of 580,317 comparative models of 6174 sequences with correlation coefficients (r) of 0.84 and 0.86, respectively, to the actual errors. This scoring function achieves the best correlation compared to 13 other tested assessment criteria that achieved correlations ranging from 0.35 to 0.71.

Protein structure modeling with MODELLER.

Genome sequencing projects have resulted in a rapid increase in the number of known protein sequences. In contrast, only about one-hundredth of these sequences have been characterized using experimental structure determination methods. Computational protein structure modeling techniques have the potential to bridge this sequence-structure gap. This chapter presents an example that illustrates the use of MODELLER to construct a comparative model for a protein with unknown structure. Automation of similar protocols (correction of protcols) has resulted in models of useful accuracy for domains in more than half of all known protein sequences.

Comparative protein structure modeling using MODELLER.

Functional characterization of a protein sequence is a common goal in biology, and is usually facilitated by having an accurate three-dimensional (3-D) structure of the studied protein. In the absence of an experimentally determined structure, comparative or homology modeling can sometimes provide a useful 3-D model for a protein that is related to at least one known protein structure. Comparative modeling predicts the 3-D structure of a given protein sequence (target) based primarily on its alignment to one or more proteins of known structure (templates). The prediction process consists of fold assignment, target-template alignment, model building, and model evaluation. This unit describes how to calculate comparative models using the program MODELLER and discusses all four steps of comparative modeling, frequently observed errors, and some applications. Modeling lactate dehydrogenase from Trichomonas vaginalis (TvLDH) is described as an example. The download and installation of the MODELLER software is also described.

Statistical potential for assessment and prediction of protein structures.

Protein structures in the Protein Data Bank provide a wealth of data about the interactions that determine the native states of proteins. Using the probability theory, we derive an atomic distance-dependent statistical potential from a sample of native structures that does not depend on any adjustable parameters (Discrete Optimized Protein Energy, or DOPE). DOPE is based on an improved reference state that corresponds to noninteracting atoms in a homogeneous sphere with the radius dependent on a sample native structure; it thus accounts for the finite and spherical shape of the native structures. The DOPE potential was extracted from a nonredundant set of 1472 crystallographic structures. We tested DOPE and five other scoring functions by the detection of the native state among six multiple target decoy sets, the correlation between the score and model error, and the identification of the most accurate non-native structure in the decoy set. For all decoy sets, DOPE is the best performing function in terms of all criteria, except for a tie in one criterion for one decoy set. To facilitate its use in various applications, such as model assessment, loop modeling, and fitting into cryo-electron microscopy mass density maps combined with comparative protein structure modeling, DOPE was incorporated into the modeling package MODELLER-8.

Structural modeling of protein interactions by analogy: application to PSD-95.

We describe comparative patch analysis for modeling the structures of multidomain proteins and protein complexes, and apply it to the PSD-95 protein. Comparative patch analysis is a hybrid of comparative modeling based on a template complex and protein docking, with a greater applicability than comparative modeling and a higher accuracy than docking. It relies on structurally defined interactions of each of the complex components, or their homologs, with any other protein, irrespective of its fold. For each component, its known binding modes with other proteins of any fold are collected and expanded by the known binding modes of its homologs. These modes are then used to restrain conventional molecular docking, resulting in a set of binary domain complexes that are subsequently ranked by geometric complementarity and a statistical potential. The method is evaluated by predicting 20 binary complexes of known structure. It is able to correctly identify the binding mode in 70% of the benchmark complexes compared with 30% for protein docking. We applied comparative patch analysis to model the complex of the third PSD-95, DLG, and ZO-1 (PDZ) domain and the SH3-GK domains in the PSD-95 protein, whose structure is unknown. In the first predicted configuration of the domains, PDZ interacts with SH3, leaving both the GMP-binding site of guanylate kinase (GK) and the C-terminus binding cleft of PDZ accessible, while in the second configuration PDZ interacts with GK, burying both binding sites. We suggest that the two alternate configurations correspond to the different functional forms of PSD-95 and provide a possible structural description for the experimentally observed cooperative folding transitions in PSD-95 and its homologs. More generally, we expect that comparative patch analysis will provide useful spatial restraints for the structural characterization of an increasing number of binary and higher-order protein complexes.

Minimalist representations and the importance of nearest neighbor effects in protein folding simulations.

In order to investigate the level of representation required to simulate folding and predict structure, we test the ability of a variety of reduced representations to identify native states in decoy libraries and to recover the native structure given the advanced knowledge of the very broad native Ramachandran basin assignments. Simplifications include the removal of the entire side-chain or the retention of only the Cbeta atoms. Scoring functions are derived from an all-atom statistical potential that distinguishes between atoms and different residue types. Structures are obtained by minimizing the scoring function with a computationally rapid simulated annealing algorithm. Results are compared for simulations in which backbone conformations are sampled from a Protein Data Bank-based backbone rotamer library generated by either ignoring or including a dependence on the identity and conformation of the neighboring residues. Only when the Cbeta atoms and nearest neighbor effects are included do the lowest energy structures generally fall within 4 A of the native backbone root-mean square deviation (RMSD), despite the initial configuration being highly expanded with an average RMSD > or = 10 A. The side-chains are reinserted into the Cbeta models with minimal steric clash. Therefore, the detailed, all-atom information lost in descending to a Cbeta-level representation is recaptured to a large measure using backbone dihedral angle sampling that includes nearest neighbor effects and an appropriate scoring function.

A composite score for predicting errors in protein structure models.

Reliable prediction of model accuracy is an important unsolved problem in protein structure modeling. To address this problem, we studied 24 individual assessment scores, including physics-based energy functions, statistical potentials, and machine learning-based scoring functions. Individual scores were also used to construct approximately 85,000 composite scoring functions using support vector machine (SVM) regression. The scores were tested for their abilities to identify the most native-like models from a set of 6000 comparative models of 20 representative protein structures. Each of the 20 targets was modeled using a template of <30% sequence identity, corresponding to challenging comparative modeling cases. The best SVM score outperformed all individual scores by decreasing the average RMSD difference between the model identified as the best of the set and the model with the lowest RMSD (DeltaRMSD) from 0.63 A to 0.45 A, while having a higher Pearson correlation coefficient to RMSD (r=0.87) than any other tested score. The most accurate score is based on a combination of the DOPE non-hydrogen atom statistical potential; surface, contact, and combined statistical potentials from MODPIPE; and two PSIPRED/DSSP scores. It was implemented in the SVMod program, which can now be applied to select the final model in various modeling problems, including fold assignment, target-template alignment, and loop modeling.

Protein complex compositions predicted by structural similarity.

Proteins function through interactions with other molecules. Thus, the network of physical interactions among proteins is of great interest to both experimental and computational biologists. Here we present structure-based predictions of 3387 binary and 1234 higher order protein complexes in Saccharomyces cerevisiae involving 924 and 195 proteins, respectively. To generate candidate complexes, comparative models of individual proteins were built and combined together using complexes of known structure as templates. These candidate complexes were then assessed using a statistical potential, derived from binary domain interfaces in PIBASE (http://salilab.org/pibase). The statistical potential discriminated a benchmark set of 100 interface structures from a set of sequence-randomized negative examples with a false positive rate of 3% and a true positive rate of 97%. Moreover, the predicted complexes were also filtered using functional annotation and sub-cellular localization data. The ability of the method to select the correct binding mode among alternates is demonstrated for three camelid VHH domain-porcine alpha-amylase interactions. We also highlight the prediction of co-complexed domain superfamilies that are not present in template complexes. Through integration with MODBASE, the application of the method to proteomes that are less well characterized than that of S.cerevisiae will contribute to expansion of the structural and functional coverage of protein interaction space. The predicted complexes are deposited in MODBASE (http://salilab.org/modbase).

MODBASE: a database of annotated comparative protein structure models and associated resources.

MODBASE (http://salilab.org/modbase) is a database of annotated comparative protein structure models for all available protein sequences that can be matched to at least one known protein structure. The models are calculated by MODPIPE, an automated modeling pipeline that relies on MODELLER for fold assignment, sequence-structure alignment, model building and model assessment (http:/salilab.org/modeller). MODBASE is updated regularly to reflect the growth in protein sequence and structure databases, and improvements in the software for calculating the models. MODBASE currently contains 3 094 524 reliable models for domains in 1 094 750 out of 1 817 889 unique protein sequences in the UniProt database (July 5, 2005); only models based on statistically significant alignments and models assessed to have the correct fold despite insignificant alignments are included. MODBASE also allows users to generate comparative models for proteins of interest with the automated modeling server MODWEB (http://salilab.org/modweb). Our other resources integrated with MODBASE include comprehensive databases of multiple protein structure alignments (DBAli, http://salilab.org/dbali), structurally defined ligand binding sites and structurally defined binary domain interfaces (PIBASE, http://salilab.org/pibase) as well as predictions of ligand binding sites, interactions between yeast proteins, and functional consequences of human nsSNPs (LS-SNP, http://salilab.org/LS-SNP).

Comparative protein structure modeling using Modeller.

Functional characterization of a protein sequence is one of the most frequent problems in biology. This task is usually facilitated by accurate three-dimensional (3-D) structure of the studied protein. In the absence of an experimentally determined structure, comparative or homology modeling can sometimes provide a useful 3-D model for a protein that is related to at least one known protein structure. Comparative modeling predicts the 3-D structure of a given protein sequence (target) based primarily on its alignment to one or more proteins of known structure (templates). The prediction process consists of fold assignment, target-template alignment, model building, and model evaluation. This unit describes how to calculate comparative models using the program MODELLER and discusses all four steps of comparative modeling, frequently observed errors, and some applications. Modeling lactate dehydrogenase from Trichomonas vaginalis (TvLDH) is described as an example. The download and installation of the MODELLER software is also described.

A simple method for faster nonbonded force evaluations.

Accurate approximations are introduced for the evaluation of inverse interparticle distances such that square root or division operations are not required in the computation of interparticle interactions and forces. These generally applicable approximations are illustrated by incorporation into the protein simulation package TINKER along with several other speed enhancement strategies. With these modifications, implicit solvent Langevin dynamics simulations of proteins are performed factors of 4.6 times faster than the modified open source distributed program. Programming speedups are obtained by extensive vectorization, simplification of the inner loop to avoid IF statements, and by using lookup tables for the distance dependent "dielectric constant" in implicit solvent models. Benchmarks are provided for the all-atom, implicit solvent dynamics of Met-enkephalin, the villin headpiece, the B1 domain of protein-G, and barnase. We also discuss the more general applicability of the approximation methods to explicit solvent simulations and of look-up tables for other implicit solvent models such as the generalized Born models.

Investigations into sequence and conformational dependence of backbone entropy, inter-basin dynamics and the Flory isolated-pair hypothesis for peptides.

The populations and transitions between Ramachandran basins are studied for combinations of the standard 20 amino acids in monomers, dimers and trimers using an implicit solvent Langevin dynamics algorithm and employing seven commonly used force-fields. Both the basin populations and inter-conversion rates are influenced by the nearest neighbor's conformation and identity, contrary to the Flory isolated-pair hypothesis. This conclusion is robust to the choice of force-field, even though the use of different force-fields produces large variations in the populations and inter-conversion rates between the dominant helical, extended beta, and polyproline II basins. The computed variation of conformational and dynamical properties with different force-fields exceeds the difference between explicit and implicit solvent calculations using the same force-field. For all force-fields, the inter-basin transitions exhibit a directional dependence, with most transitions going through extended beta conformation, even when it is the least populated basin. The implications of these results are discussed in the context of estimates for the backbone entropy of single residues, and for the ability of all-atom simulations to reproduce experimental protein folding data.

Folding and misfolding of the papillomavirus E6 interacting peptide E6ap.

All-atom Langevin dynamics simulations have been performed to study the folding pathways of the 18-residue binding domain fragment E6ap of the human papillomavirus E6 interacting peptide. Six independent folding trajectories, with a total duration of nearly 2 micros, all lead to the same native state in which the E6ap adopts a fluctuating alpha-helix structure in the central portion (Ser-4-Leu-13) but with very flexible N and C termini. Simulations starting from different core configurations exhibit the E6ap folding dynamics as either a two- or three-state folder with an intermediate misfolded state. The essential leucine hydrophobic core (Leu-9, Leu-12, and Leu-13) is well conserved in the native-state structure but absent in the intermediate structure, suggesting that the leucine core is not only essential for the binding activity of E6ap but also important for the stability of the native structure. The free energy landscape reveals a significant barrier between the basins separating the native and misfolded states. We also discuss the various underlying forces that drive the peptide into its native state.

Large-scale context in protein folding: villin headpiece.

The villin headpiece folds autonomously in vitro forming three alpha-helical regions. Local propensities, however, strongly disfavor the formation of the C-terminal helix because most native residue pairs in that helix are hydrophobic/polar mismatches. Even the N-terminal helix is unfavored according to the AGADIR criterion. Our coarse-grained ab initio simulations reveal three-body correlations in which hydrophobic residues position to protect amide-carbonyl hydrogen bonds from attack by water, thus inducing the growth of the C-terminal helix and guiding the folding process. Similar correlations are also found in all-atom simulations with an implicit solvent model that accurately reproduces the results of simulations with explicit solvent molecules. The correlations establish a large-scale, many-body context that may be probed experimentally by introducing mutations of certain nonobvious residues that reside outside the native hydrophobic core but that are predicted to affect the folding rates and dynamics dramatically.

All-atom fast protein folding simulations: the villin headpiece.

We provide a fast folding simulation using an all-atom solute, implicit solvent method that eliminates the need for treating solvent degrees of freedom. The folding simulations for the 36-residue villin headpiece exhibit close correspondence with the landmark all-atom explicit solvent molecular dynamics simulations by Duan and Kollman (Duan & Kollman, Science 1998;282:740-744; Duan, Wang, & Kollman, Proc Natl Acad Sci USA 1998;95:9897-9902). Our implicit solvent approach uses only an entry-level single CPU PC with comparable throughput ( approximately 4 nsec/day) to the DK supercomputer simulation. The native state is shown to be stable. Our 200-nsec folding trajectory agrees with the DK simulation in displaying a burst phase, a rapid initial shrinkage, a highly native-like binding site structure, and more.