Prediction of protein continuum secondary structure with probabilistic models based on NMR solved structures

Background: The structure of proteins may change as a result of the inherent flexibility of some protein regions. We develop and explore probabilistic machine learning methods for predicting a continuum secondary structure, i.e. assigning probabilities to the conformational states of a residue. We train our methods using data derived from high-quality NMR models. Results: Several probabilistic models not only successfully estimate the continuum secondary structure, but also provide a categorical output on par with models directly trained on categorical data. Importantly, models trained on the continuum secondary structure are also better than their categorical counterparts at identifying the conformational state for structurally ambivalent residues. Conclusion: Cascaded probabilistic neural networks trained on the continuum secondary structure exhibit better accuracy in structurally ambivalent regions of proteins, while sustaining an overall classification accuracy on par with standard, categorical prediction methods.

Identifying sequence regions undergoing conformational change via predicted continuum secondary structure

Motivation: Conformational flexibility is essential to the function of many proteins, e.g. catalytic activity. To assist efforts in determining and exploring the functional properties of a protein, it is desirable to automatically identify regions that are prone to undergo conformational changes. It was recently shown that a probabilistic predictor of continuum secondary structure is more accurate than categorical predictors for structurally ambivalent sequence regions, suggesting that such models are suited to characterize protein flexibility. Results: We develop a computational method for identifying regions that are prone to conformational change directly from the amino acid sequence. The method uses the entropy of the probabilistic output of an 8-class continuum secondary structure predictor. Results for 171 unique amino acid sequences with well-characterized variable structure (identified in the 'Macromolecular movements database') indicate that the method is highly sensitive at identifying flexible protein regions, but false positives remain a problem. The method can be used to explore conformational flexibility of proteins (including hypothetical or synthetic ones) whose structure is yet to be determined experimentally.

Translation of the flavivirus Kunjin NS3 gene in cis but not its RNA sequence or secondary structure is essential for efficient RNA packaging

Our previous studies using trans-complementation analysis of Kunjin virus (KUN) full-length cDNA clones harboring in-frame deletions in the NS3 gene demonstrated the inability of these defective complemented RNAs to be packaged into virus particles (W. J. Liu, P. L. Sedlak, N. Kondratieva, and A. A. Khromykh, J. Virol. 76:10766-10775). In this study we aimed to establish whether this requirement for NS3 in RNA packaging is determined by the secondary RNA structure of the NS3 gene or by the essential role of the translated NS3 gene product. Multiple silent mutations of three computer-predicted stable RNA structures in the NS3 coding region of KUN replicon RNA aimed at disrupting RNA secondary structure without affecting amino acid sequence did not affect RNA replication and packaging into virus-like particles in the packaging cell line, thus demonstrating that the predicted conserved RNA structures in the NS3 gene do not play a role in RNA replication and/or packaging. In contrast, double frameshift mutations in the NS3 coding region of full-length KUN RNA, producing scrambled NS3 protein but retaining secondary RNA structure, resulted in the loss of ability of these defective RNAs to be packaged into virus particles in complementation experiments in KUN replicon-expressing cells. Furthermore, the more robust complementation-packaging system based on established stable cell lines producing large amounts of complemented replicating NS3-deficient replicon RNAs and infection with KUN virus to provide structural proteins also failed to detect any secreted virus-like particles containing packaged NS3-deficient replicon RNAs. These results have now firmly established the requirement of KUN NS3 protein translated in cis for genome packaging into virus particles.

Identification of the essential histidine residue for high-affinity binding of AlbA protein to albicidin antibiotics

The albA gene from Klebsiella oxytoca encodes a protein that binds albicidin phytotoxins and antibiotics with high affinity. Previously, it has been shown that shifting pH from 6 to 4 reduces binding activity of AlbA by about 30%, indicating that histidine residues might be involved in substrate binding. In this study, molecular analysis of the albA coding region revealed sequence discrepancies with the albA sequence reported previously, which were probably due to sequencing errors. The albA gene was subsequently cloned from K oxytoca ATCC 13182(T) to establish the revised sequence. Biochemical and molecular approaches were used to determine the functional role of four histidine residues (His(78), HiS(125), HiS(141) and His(189)) in the corrected sequence for AlbA. Treatment of AlbA with diethyl pyrocarbonate (DEPC), a histidine-specific alkylating reagent, reduced binding activity by about 95%. DEPC treatment increased absorbance at 240-244 nm by an amount indicating conversion to N-carbethoxyhistidine of a single histidine residue per AlbA molecule. Pretreatment with albicidin protected AlbA against modification by DEPC, with a 1 : 1 molar ratio of albicidin to the protected histidine residues. Based on protein secondary structure and amino acid surface probability indices, it is predicted that HiS125 might be the residue required for albicidin binding. Mutation of HiS125 to either alanine or leucine resulted in about 32% loss of binding activity, and deletion of HiS125 totally abolished binding activity. Mutation of HiS125 to arginine and tyrosine had no effect. These results indicate that HiS125 plays a key role either in an electrostatic interaction between AlbA and albicidin or in the conformational dynamics of the albicidin-binding site.

West nile virus core protein: Tetramer structure and ribbon formation

We have determined the crystal structure of the core (C) protein from the Kunjin subtype of West Nile virus (WNV), closely related to the NY99 strain of WNV, currently a major health threat in the U.S. WNV is a member of the Flaviviridae family of enveloped RNA viruses that contains many important human pathogens. The C protein is associated with the RNA genome and forms the internal core which is surrounded by the envelope in the virion. The C protein structure contains four a. helices and forms dimers that are organized into tetramers. The tetramers form extended filamentous ribbons resembling the stacked alpha helices seen in HEAT protein structures.

Metal Clips Induce Folding of a Short Unstructured Peptide into an a-Helix via Turn Conformations in Water. Kinetic versus Thermodynamic Products

Short peptides corresponding to two to four a-helical turns of proteins are not thermodynamically stable helices in water. Unstructured octapeptide Ac-His1*-Ala2-Ala3-His4*-His5*-Glu6-Leu7-His8*-NH2 (1) reacts with two [Pd ((NH2)-N-15(CH2)(2) (NH2)-N-15)(NO3)(2)] in water to form a kinetically stable intermediate, [{Pden}(2)-{(1,4)(5,8)-peptide}](2), in which two 19-membered metallocyclic rings stabilize two peptide turns. Slow subsequent folding to a thermodynamically more stable two-turn a-helix drives the equilibrium to [{Pden}(2)-{(1,5)(4,8)-peptide}] (3), featuring two 22-membered rings. This transformation from unstructured peptide via turns to an a-helix suggests that metal clips might be useful probes for investigating peptide folding.

Single turn peptide alpha helices with exceptional stability in water

Cyclic pentapepticles are not known to exist in a-helical conformations. CD and NMR spectra show that specific 20-membered cyclic pentapepticles, Ac-(cyclo-1,5) [KxxxD]-NH2 and Ac-(cyclo-2,6)R[KxxxD]-NH2, are highly a-helical structures in water and independent of concentration, TFE, denaturants, and proteases. These are the smallest a-helical peptides in water.

Prediction of subcellular localization using sequence-biased recurrent networks

Motivation: Targeting peptides direct nascent proteins to their specific subcellular compartment. Knowledge of targeting signals enables informed drug design and reliable annotation of gene products. However, due to the low similarity of such sequences and the dynamical nature of the sorting process, the computational prediction of subcellular localization of proteins is challenging. Results: We contrast the use of feed forward models as employed by the popular TargetP/SignalP predictors with a sequence-biased recurrent network model. The models are evaluated in terms of performance at the residue level and at the sequence level, and demonstrate that recurrent networks improve the overall prediction performance. Compared to the original results reported for TargetP, an ensemble of the tested models increases the accuracy by 6 and 5% on non-plant and plant data, respectively.

The applicability of recurrent neural networks for biological sequence analysis

Selection of machine learning techniques requires a certain sensitivity to the requirements of the problem. In particular, the problem can be made more tractable by deliberately using algorithms that are biased toward solutions of the requisite kind. In this paper, we argue that recurrent neural networks have a natural bias toward a problem domain of which biological sequence analysis tasks are a subset. We use experiments with synthetic data to illustrate this bias. We then demonstrate that this bias can be exploitable using a data set of protein sequences containing several classes of subcellular localization targeting peptides. The results show that, compared with feed forward, recurrent neural networks will generally perform better on sequence analysis tasks. Furthermore, as the patterns within the sequence become more ambiguous, the choice of specific recurrent architecture becomes more critical.

NMR of conotoxins: structural features and an analysis of chemical shifts of post-translationally modified amino acids

Conotoxins are small conformationally constrained peptides found in the venom of marine snails of the genus Conus. They are usually cysteine rich and frequently contain a high degree of post-translational modifications such as C-terminal amidation, hydroxylation, carboxylation, bromination, epimerisation and glycosylation. Here we review the role of NMR in determining the three-dimensional structures of conotoxins and also provide a compilation and analysis of H-1 and C-13 chemical shifts of post-translationally modified amino acids and compare them with data from common amino acids. This analysis provides a reference source for chemical shifts of post-translationally modified amino acids. Copyright (C) 2006 John Wiley & Sons, Ltd.

Molecular dynamics simulations of the hydrophobin SC3 at a hydrophobic/hydrophilic interface

Hydrophobins are small (similar to 100 aa) proteins that have an important role in the growth and development of mycelial fungi. They are surface active and, after secretion by the fungi, self-assemble into amphipathic membranes at hydrophobic/hydrophilic interfaces, reversing the hydrophobicity of the surface. In this study, molecular dynamics simulation techniques have been used to model the process by which a specific class I hydrophobin, SC3, binds to a range of hydrophobic/ hydrophilic interfaces. The structure of SC3 used in this investigation was modeled based on the crystal structure of the class II hydrophobin HFBII using the assumption that the disulfide pairings of the eight conserved cysteine residues are maintained. The proposed model for SC3 in aqueous solution is compact and globular containing primarily P-strand and coil structures. The behavior of this model of SC3 was investigated at an air/water, an oil/water, and a hydrophobic solid/water interface. It was found that SC3 preferentially binds to the interfaces via the loop region between the third and fourth cysteine residues and that binding is associated with an increase in a-helix formation in qualitative agreement with experiment. Based on a combination of the available experiment data and the current simulation studies, we propose a possible model for SC3 self-assembly on a hydrophobic solid/water interface.

Structure of Petunia hybrida Defensin 1, a Novel Plant Defensin with Five Disulfide Bonds

The structure of a novel plant defensin isolated from the flowers of Petunia hybrida has been determined by H-1 NMR spectroscopy. P. hybrida defensin 1 (PhD1) is a basic, cysteine-rich, antifungal protein of 47 residues and is the first example of a new subclass of plant defensins with five disulfide bonds whose structure has been determined. PhD1 has the fold of the cysteine-stabilized alphabeta motif, consisting of an alpha-helix and a triple-stranded antiparallel beta-sheet, except that it contains a fifth disulfide bond from the first loop to the alpha-helix. The additional disulfide bond is accommodated in PhD1 without any alteration of its tertiary structure with respect to other plant defensins. Comparison of its structure with those of classic, four-disulfide defensins has allowed us to identify a previously unrecognized hydrogen bond network that is integral to structure stabilization in the family.

Protein contact prediction using patterns of correlation

We describe a new method for using neural networks to predict residue contact pairs in a protein. The main inputs to the neural network are a set of 25 measures of correlated mutation between all pairs of residues in two windows of size 5 centered on the residues of interest. While the individual pair-wise correlations are a relatively weak predictor of contact, by training the network on windows of correlation the accuracy of prediction is significantly improved. The neural network is trained on a set of 100 proteins and then tested on a disjoint set of 1033 proteins of known structure. An average predictive accuracy of 21.7% is obtained taking the best L/2 predictions for each protein, where L is the sequence length. Taking the best L/10 predictions gives an average accuracy of 30.7%. The predictor is also tested on a set of 59 proteins from the CASP5 experiment. The accuracy is found to be relatively consistent across different sequence lengths, but to vary widely according to the secondary structure. Predictive accuracy is also found to improve by using multiple sequence alignments containing many sequences to calculate the correlations. (C) 2004 Wiley-Liss, Inc.