A novel method for protein secondary structure prediction using dual-layer SVM and profiles

A high-performance method was developed for protein secondary structure prediction based on the dual-layer support vector machine (SVM) and position-specific scoring matrices (PSSMs). SVM is a new machine learning technology that has been successfully applied in solving problems in the field of bioinformatics. The SVM's performance is usually better than that of traditional machine learning approaches. The performance was further improved by combining PSSM profiles with the SVM analysis. The PSSMs were generated from PSI-BLAST profiles, which contain important evolution information. The final prediction results were generated from the second SVM layer output. On the CB513 data set, the three-state overall per-residue accuracy, Q3, reached 75.2%, while segment overlap (SOV) accuracy increased to 80.0%. On the CB396 data set, the Q3 of our method reached 74.0% and the SOV reached 78.1%. A web server utilizing the method has been constructed and is available at http://www.bioinfo.tsinghua.edu.cn/pmsvm.     

Better prediction of the location of alpha-turns in proteins with support vector machine

We have developed a novel method named AlphaTurn to predict alpha-turns in proteins based on the support vector machine (SVM). The prediction was done on a data set of 469 nonhomologous proteins containing 967 alpha-turns. A great improvement in prediction performance was achieved by using multiple sequence alignment generated by PSI-BLAST as input instead of the single amino acid sequence. The introduction of secondary structure information predicted by PSIPRED also improved the prediction performance. Moreover, we handled the very uneven data set by combining the cost factor j with the "state-shifting" rule. This further promoted the prediction quality of our method. The final SVM model yielded a Matthews correlation coefficient (MCC) of 0.25 by a 10-fold cross-validation. To our knowledge, this MCC value is the highest obtained so far for predicting alpha-turns. An online Web server based on this method has been developed and can be freely accessed at http://bmc.hust.edu.cn/bioinformatics/ or http://210.42.106.80/.     

Removal of the N-terminal hexapeptide from human beta2-microglobulin facilitates protein aggregation and fibril formation

The solution structure and stability of N-terminally truncated beta2-microglobulin (deltaN6beta2-m), the major modification in ex vivo fibrils, have been investigated by a variety of biophysical techniques. The results show that deltaN6beta2-m has a free energy of stabilization that is reduced by 2.5 kcal/mol compared to the intact protein. Hydrogen exchange of a mixture of the truncated and full-length proteins at microM concentrations at pH 6.5 monitored by electrospray mass spectrometry reveals that deltaN6beta2-m is significantly less protected than its wild-type counterpart. Analysis of deltaN6beta2-m by NMR shows that this loss of protection occurs in beta strands I, III, and part of II. At mM concentration gel filtration analysis shows that deltaN6beta2-m forms a series of oligomers, including trimers and tetramers, and NMR analysis indicates that strand V is involved in intermolecular interactions that stabilize this association. The truncated species of beta2-microglobulin was found to have a higher tendency to self-associate than the intact molecule, and unlike wild-type protein, is able to form amyloid fibrils at physiological pH. Limited proteolysis experiments and analysis by mass spectrometry support the conformational modifications identified by NMR and suggest that deltaN6beta2-m could be a key intermediate of a proteolytic pathway of beta2-microglobulin. Overall, the data suggest that removal of the six residues from the N-terminus of beta2-microglobulin has a major effect on the stability of the overall fold. Part of the tertiary structure is preserved substantially by the disulfide bridge between Cys25 and Cys80, but the pairing between beta-strands far removed from this constrain is greatly perturbed.     

Cooperative effects in hydrogen-bonding of protein secondary structure elements: a systematic analysis of crystal data using Secbase

A systematic analysis of the hydrogen-bonding geometry in helices and beta sheets has been performed. The distances and angles between the backbone carbonyl O and amide N atoms were correlated considering more than 1500 protein chains in crystal structures determined to a resolution better than 1.5 A. They reveal statistically significant trends in the H-bond geometry across the different secondary structural elements. The analysis has been performed using Secbase, a modular extension of Relibase (Receptor Ligand Database) which integrates information about secondary structural elements assigned to individual protein structures with the various search facilities implemented into Relibase. A comparison of the mean hydrogen-bond distances in alpha helices and 3(10) helices of increasing length shows opposing trends. Whereas in alpha helices the mean H-bond distance shrinks with increasing helix length and turn number, the corresponding mean dimension in 3(10) helices expands in a comparable series. Comparing similarly the hydrogen-bond lengths in beta sheets there is no difference to be found between the mean H-bond length in antiparallel and parallel beta sheets along the strand direction. In contrast, an interesting systematic trend appears to be given for the hydrogen bonds perpendicular to the strands bridging across an extended sheet. With increasing number of accumulated strands, which results in a growing number of back-to-back piling hydrogen bonds across the strands, a slight decrease of the mean H-bond distance is apparent in parallel beta sheets whereas such trends are obviously not given in antiparallel beta sheets. This observation suggests that cooperative effects mutually polarizing spatially well-aligned hydrogen bonds are present either in alpha helices and parallel beta sheets whereas such influences seem to be lacking in 3(10) helices and antiparallel beta sheets.     

Sequence specificity, statistical potentials, and three-dimensional structure prediction with self-correcting distance geometry calculations of beta-sheet formation in proteins

A statistical analysis of a representative data set of 169 known protein structures was used to analyze the specificity of residue interactions between spatial neighboring strands in beta-sheets. Pairwise potentials were derived from the frequency of residue pairs in nearest contact, second nearest and third nearest contacts across neighboring beta-strands compared to the expected frequency of residue pairs in a random model. A pseudo-energy function based on these statistical pairwise potentials recognized native beta-sheets among possible alternative pairings. The native pairing was found within the three lowest energies in 73% of the cases in the training data set and in 63% of beta-sheets in a test data set of 67 proteins, which were not part of the training set. The energy function was also used to detect tripeptides, which occur frequently in beta-sheets of native proteins. The majority of native partners of tripeptides were distributed in a low energy range. Self-correcting distance geometry (SECODG) calculations using distance constraints sets derived from possible low energy pairing of beta-strands uniquely identified the native pairing of the beta-sheet in pancreatic trypsin inhibitor (BPTI). These results will be useful for predicting the structure of proteins from their amino acid sequence as well as for the design of proteins containing beta-sheets.     

Strand-loop-strand motifs: prediction of hairpins and diverging turns in proteins

Beta-sheet proteins have been particularly challenging for de novo structure prediction methods, which tend to pair adjacent beta-strands into beta-hairpins and produce overly local topologies. To remedy this problem and facilitate de novo prediction of beta-sheet protein structures, we have developed a neural network that classifies strand-loop-strand motifs by local hairpins and nonlocal diverging turns by using the amino acid sequence as input. The neural network is trained with a representative subset of the Protein Data Bank and achieves a prediction accuracy of 75.9 +/- 4.4% compared to a baseline prediction rate of 59.1%. Hairpins are predicted with an accuracy of 77.3 +/- 6.1%, diverging turns with an accuracy of 73.9 +/- 6.0%. Incorporation of the beta-hairpin/diverging turn classification into the ROSETTA de novo structure prediction method led to higher contact order models and somewhat improved tertiary structure predictions for a test set of 11 all-beta-proteins and 3 alphabeta-proteins. The beta-hairpin/diverging turn classification from amino acid sequences is available online for academic use (Meiler and Kuhn, 2003; www.jens-meiler.de/turnpred.html).     

Expanded turn conformations: characterization and sequence-structure correspondence in alpha-turns with implications in helix folding

Like the beta-turns, which are characterized by a limiting distance between residues two positions apart (i, i+3), a distance criterion (involving residues at positions i and i+4) is used here to identify alpha-turns from a database of known protein structures. At least 15 classes of alpha-turns have been enumerated based on the location in the phi,psi space of the three central residues (i+1 to i+3)-one of the major being the class AAA, where the residues occupy the conventional helical backbone torsion angles. However, moving towards the C-terminal end of the turn, there is a shift in the phi,psi angles towards more negative phi, such that the electrostatic repulsion between two consecutive carbonyl oxygen atoms is reduced. Except for the last position (i+4), there is not much similarity in residue composition at different positions of hydrogen and non-hydrogen bonded AAA turns. The presence or absence of Pro at i+1 position of alpha- and beta-turns has a bearing on whether the turn is hydrogen-bonded or without a hydrogen bond. In the tertiary structure, alpha-turns are more likely to be found in beta-hairpin loops. The residue composition at the beginning of the hydrogen bonded AAA alpha-turn has similarity with type I beta-turn and N-terminal positions of helices, but the last position matches with the C-terminal capping position of helices, suggesting that the existence of a "helix cap signal" at i+4 position prevents alpha-turns from growing into helices. Our results also provide new insights into alpha-helix nucleation and folding.     

Prediction of turn types in protein structure by machine-learning classifiers

We present machine learning approaches for turn prediction from the amino acid sequence. Different turn classes and types were considered based on a novel turn classification scheme. We trained an unsupervised (self-organizing map) and two kernel-based classifiers, namely the support vector machine and a probabilistic neural network. Turn versus non-turn classification was carried out for turn families containing intramolecular hydrogen bonds and three to six residues. Support vector machine classifiers yielded a Matthews correlation coefficient (mcc) of approximately 0.6 and a prediction accuracy of 80%. Probabilistic neural networks were developed for beta-turn type prediction. The method was able to distinguish between five types of beta-turns yielding mcc > 0.5 and at least 80% overall accuracy. We conclude that the proposed new turn classification is distinct and well-defined, and machine learning classifiers are suited for sequence-based turn prediction. Their potential for sequence-based prediction of turn structures is discussed.     

Determination of alpha-helix and beta-sheet stability in the solid state: a solid-state NMR investigation of poly(L-alanine)

The relative stability of alpha-helix and beta-sheet secondary structure in the solid state was investigated using poly(L-alanine) (PLA) as a model system. Protein folding and stability has been well studied in solution, but little is known about solid-state environments, such as the core of a folded protein, where peptide packing interactions are the dominant factor in determining structural stability. (13)C cross-polarization with magic angle spinning (CPMAS) NMR spectroscopy was used to determine the backbone conformation of solid powder samples of 15-kDa and 21.4-kDa PLA before and after various sample treatments. Reprecipitation from helix-inducing solvents traps the alpha-helical conformation of PLA, although the method of reprecipitation also affects the conformational distribution. Grinding converts the secondary structure of PLA to a final steady-state mixture of 55% beta-sheet and 45% alpha-helix at room temperature regardless of the initial secondary structure. Grinding PLA at liquid nitrogen temperatures leads to a similar steady-state mixture with 60% beta-sheet and 40% alpha-helix, indicating that mechanical shear force is sufficient to induce secondary structure interconversion. Cooling the sample in liquid nitrogen or subjecting it to high pressure has no effect on secondary structure. Heating the sample without grinding results in equilibration of secondary structure to 50% alpha-helix/50% beta-sheet at 100 degrees C when starting from a mostly alpha-helical state. No change was observed upon heating a beta-sheet sample, perhaps due to kinetic effects and the different heating rate used in the experiments. These results are consistent with beta-sheet approximately 260 J/mol more stable than alpha-helix in solid-state PLA.     

C-terminal region of protein kinase CK2 alpha: How the structure can affect function and stability of the catalytic subunit

A novel mutant of the catalytic alpha subunit of human protein kinase CK2 (CK2 alpha) was designed in an attempt to clarify the role of the carboxylic-terminal segment characteristic of vertebrates, excluding fish. Starting from the sequence alignments, we constructed a phylogenetic tree of the primary structure of CK2 alpha. On this basis, we substituted two distal prolines with alanines (PA 382-384). Theoretical calculations and spectropolarimetry measurements, performed both on native and mutant subunits, indicate an increased content of alpha-helix after this double amino acidic substitution. In order to clarify the structure/function relationship of the C-terminal region, we verified if the structural change affects the catalytic activity of CK2 alpha. The mutant exhibits slightly increased phosphorylation efficiency, but reduced ability to transfer phosphate in comparison with the native subunit. At last, we compared the thermal stability of the mutant with respect to the native subunit and we tested the proteolytic degradability. The observation that the PA 382-384 mutant exhibits an increased thermal and proteolytic stability suggests that this mutant could be employed to solve the three-dimensional (3D) structure of human CK2 alpha and to overcome difficulties in crystallizing the native form.     

Contribution of secondary structure to the heat capacity and enthalpy distribution of the unfolded state in proteins.

We have recently shown that one can construct the enthalpy distribution for protein molecules from experimental knowledge of the temperature dependence of the heat capacity. For many proteins the enthalpy distribution evaluated at the midpoint of the denaturation transition (corresponding to the maximum in the heat capacity vs temperature curve) is broad and biphasic, indicating two different populations of molecules (native and unfolded) with distinctly different enthalpies. At temperatures above the denaturation point, the heat capacity for the unfolded state in many proteins is quite large and using the analysis just mentioned, we obtain a gaussian-like enthalpy distribution that is very broad. A large value of the heat capacity indicates that there are structural changes going on in the unfolded state above the transition temperature. In the present paper we investigate the origin of this large heat capacity by considering the presence of changing amounts of secondary structure (specifically, alpha-helix) in the unfolded state. For this purpose we use the empirical estimates of the Zimm-Bragg sigma and s factors for all of the native amino acids in water as determined by Scheraga and co-workers. Using myoglobin as an example, we calculate probability profiles and distribution functions for the total number of helix states in the specific-sequence molecule. Given the partition function for the specific-sequence molecule, we can then calculate a set of enthalpy moments for the molecule from which we obtain a good estimate of the enthalpy distribution in the unfolded state. This distribution turns out to be quite narrow when compared with the distribution obtained from the raw heat capacity data. We conclude that there must be other major structural changes (backbone and solvent) that are not accounted for by the inclusion of alpha-helix in the unfolded state.     

The interaction of CK2alpha and CK2beta, the subunits of protein kinase CK2, requires CK2beta in a preformed conformation and is enthalpically driven

The protein kinase CK2 (former name: "casein kinase 2") predominantly occurs as a heterotetrameric holoenzyme composed of two catalytic chains (CK2alpha) and two noncatalytic subunits (CK2beta). The CK2beta subunits form a stable dimer to which the CK2alpha monomers are attached independently. In contrast to the cyclins in the case of the cyclin-dependent kinases CK2beta is no on-switch of CK2alpha; rather the formation of the CK2 holoenzyme is accompanied with an overall change of the enzyme's profile including a modulation of the substrate specificity, an increase of the thermostability, and an allocation of docking sites for membranes and other proteins. In this study we used C-terminal deletion variants of human CK2alpha and CK2beta that were enzymologically fully competent and in particular able to form a heterotetrameric holoenzyme. With differential scanning calorimetry (DSC) we confirmed the strong thermostabilization effect of CK2alpha on CK2beta with an upshift of the CK2alpha melting temperature of more than 9 degrees . Using isothermal titration calorimetry (ITC) we measured a dissociation constant of 12.6 nM. This high affinity between CK2alpha and CK2beta is mainly caused by enthalpic rather than entropic contributions. Finally, we determined a crystal structure of the CK2beta construct to 2.8 A resolution and revealed by structural comparisons with the CK2 holoenzyme structure that the CK2beta conformation is largely conserved upon association with CK2alpha, whereas the latter undergoes significant structural adaptations of its backbone.     

Self-cycling operation increases productivity of recombinant protein in Escherichia coli

Self-cycling fermentation (SCF), a cyclical, semi-continuous process that induces cell synchrony, was incorporated into a recombinant protein production scheme. Escherichia coli CY15050, a lac(-) mutant lysogenized with temperature-sensitive phage λ modified to over-express β-galactosidase, was used as a model system. The production scheme was divided into two de-coupled stages. The host cells were cultured under SCF operation in the first stage before being brought to a second stage where protein production was induced. In the first stage, the host strain demonstrated a stable cycling pattern immediately following the first cycle. This reproducible pattern was maintained over the course of the experiments and a significant degree of cell synchrony was obtained. By growing cells using SCF, productivity increased 50% and production time decreased by 40% compared to a batch culture under similar conditions. In addition, synchronized cultures induced from the end of a SCF cycle displayed shorter lysis times and a more complete culture-wide lysis than unsynchronized cultures. Finally, protein synthesis was influenced by the time at which the lytic phase was induced in the cell life cycle. For example, induction of a synchronized culture immediately prior to cell division resulted in the maximum protein productivity, suggesting protein production can be optimized with respect to the cell life cycle using SCF.     

The geometry and efficacy of cation-pi interactions in a diagonal position of a designed beta-hairpin

Cation-pi interactions are common in proteins, but their contribution to the stability and specificity of protein structure has not been well established. In this study, we examined the impact of cation-pi interactions in a diagonal position of a beta-hairpin peptide through comparison of the interaction of Phe or Trp with Lys or Arg. The diagonal interactions ranged from -0.20 to -0.48 kcal/mole. Our experimental values for the diagonal cation-pi interactions are similar to those found in alpha-helical studies. Upfield shifting of the Lys and Arg side chains indicates that the geometries of cation-pi interactions adopted in the 12-residue beta-hairpin are comparable to those found in protein structures. The Lys was found to interact through the polarized Cepsilon, and the Arg is stacked against the aromatic ring of Phe or Trp. Folding of these peptides was found to be enthalpically favorable (DeltaH degrees equals approximately -3 kcal/mole) and entropically unfavorable (DeltaS degrees equals approximately -8 cal mole(-1) K(-1)).     

Crystal structure of cyclic (APGVGV)2, an analog of elastin, and a suggested mechanism for elongation/contraction of the molecule

Tropoelastin is a complex polymeric protein composed primarily of repeating segments of Val-Pro-Gly-Gly, Val-Pro-Gly-Val-Gly, and Ala-Pro-Gly-Val-Gly-Val that occurs in connective tissue and arteries. It has rubber-like extensible properties. A synthetic cyclic dodecapeptide, with a double repeat of the hexapeptide sequence, has been shown to undergo a reversible inverse temperature transition; that is, crystals grow at 60 degrees C and dissolve in the mother liquor upon cooling. An x-ray crystal structure analysis established that the cyclic backbone formed an elongated loop with a Pro-Gly, type II beta turn at both ends. Six internal cross strand NH...OC hydrogen bonds form between six NH donors and four O=C acceptors where two of the carbonyl O atoms are bifurcated acceptors. As a result, the molecule is pulled up into a corrugated profile. The corrugated loops form extended beta-sheets by additional intermolecular hydrogen bonds. An analysis of the dome region in a corrugated sheet suggests a reversible mechanism for extending and contracting the length of the whole molecule, akin to the motion of opening and closing an umbrella, caused by the motion of a water molecule with its associated hydrogen bonds acting as spokes. Crystal parameters: C44H72N12O12.3H2O, sp. gr. P2(1)2(1)2(1), a = 9.212 angstroms, b = 19.055 angstroms, c = 32.247 angstroms, d = 1.157 g/cm3.     

Prediction of transmembrane regions of beta-barrel proteins using ANN- and SVM-based methods

This article describes a method developed for predicting transmembrane beta-barrel regions in membrane proteins using machine learning techniques: artificial neural network (ANN) and support vector machine (SVM). The ANN used in this study is a feed-forward neural network with a standard back-propagation training algorithm. The accuracy of the ANN-based method improved significantly, from 70.4% to 80.5%, when evolutionary information was added to a single sequence as a multiple sequence alignment obtained from PSI-BLAST. We have also developed an SVM-based method using a primary sequence as input and achieved an accuracy of 77.4%. The SVM model was modified by adding 36 physicochemical parameters to the amino acid sequence information. Finally, ANN- and SVM-based methods were combined to utilize the full potential of both techniques. The accuracy and Matthews correlation coefficient (MCC) value of SVM, ANN, and combined method are 78.5%, 80.5%, and 81.8%, and 0.55, 0.63, and 0.64, respectively. These methods were trained and tested on a nonredundant data set of 16 proteins, and performance was evaluated using "leave one out cross-validation" (LOOCV). Based on this study, we have developed a Web server, TBBPred, for predicting transmembrane beta-barrel regions in proteins (available at http://www.imtech.res.in/raghava/tbbpred).     

Application of ATR‐FTIR spectroscopy to the study of thermally induced changes in secondary structure of protein molecules in solid state

FTIR spectroscopy in combination with ATR sampling technique is the most accessible analytical technique to study secondary structure of proteins both in solid and aqueous solution. Although several studies have demonstrated the applications of ATR‐FTIR to study conformational changes of solid dried proteins due to dehydration, there are no reports that demonstrate the application of ATR‐FTIR in the study of thermally induced changes of secondary structure of biomolecules directly on the solid state. In this study, four biomolecules of pharmaceutical interest, lysozyme, myoglobine, chymotripsin and human growth hormone (hGH), were studied on the solid state before and after different thermal treatments in order to relate changes of secondary structure to partial or total thermal denaturation processes. The results obtained provide experimental evidence that protein thermal denaturation in the solid state can be detected by displacement of carbonyl bands which correspond to conformational transformations between α–helix to β‐sheet or intermolecular β‐sheet; the molecules studied undergo this transformation when exposed to a temperature close to their denaturation temperature which may become irreversible depending on the extent of the heating treatment. These findings demonstrate that ATR‐FTIR is an effective and time efficient technique that allows the monitoring of the protein thermal denaturation process of solid samples without further reconstitution or prior sample preparation. © 2015 Wiley Periodicals, Inc. Biopolymers 103: 574–584, 2015.     

Beta-sheet and associated turn signatures in vibrational Raman optical activity spectra of proteins.

We have measured the aqueous solution vibrational Raman optical activity (ROA) spectra of concanavalin A, alpha-chymotrypsin, and beta-lactoglobulin, all of which are rich in beta-sheet, together with that of the model beta-turn peptide L-pro-L-leu-gly-NH2. Possible ROA signatures of antiparallel beta-sheet include a strong sharp positive band at approximately 1,313 cm-1 associated with backbone amide III C alpha H and NH deformations, and an amide I couplet, negative at low wavenumber and positive at high, centered at approximately 1,658 cm-1. Negative ROA bands in the range approximately 1,340-1,380 cm-1, which might originate in glycine CH2 deformations, appear to be characteristic of beta-turns. Our results provide further evidence that ROA is a more incisive probe of protein conformation than conventional vibrational spectroscopy, infrared, or Raman, because only those few vibrational coordinates within a given normal mode that sample the skeletal chirality directly contribute to the corresponding ROA band intensity.     

Synthesis, folding, and structure of the beta-turn mimic modified B1 domain of streptococcal protein G

The mechanism of beta-sheet formation remains a fundamental issue in our understanding of the protein folding process, but is hampered by the often encountered kinetic competition between folding and aggregation. The role of local versus nonlocal interactions has been probed traditionally by mutagenesis of both turn and strand residues. Recently, rigid organic molecules that impose a correct chain reversal have been introduced in several small peptides to isolate the importance of the long-range interactions. Here, we present the incorporation of a well-studied beta-turn mimic, designated as the dibenzofuran-based (DBF) amino acid, in the B1 domain of streptococcal protein G (B1G), and compare our results with those obtained upon insertion of the same mimic into the N-terminal beta-hairpin of B1G (O Melnyk et al., 1998, Lett Pept Sci 5:147-150). The DBF-B1G domain conserves the structure and the functional and thermodynamical properties of the native protein, whereas the modified peptide does not adopt a native-like conformation. The nature of the DBF flanking residues in the modified B1G domain prevents the beta-turn mimic from acting as a strong beta-sheet nucleator, which reinforces the idea that the native beta-hairpin formation is not driven by the beta-turn formation, but by tertiary interactions.     

Initial structural and dynamic characterization of the M2 protein transmembrane and amphipathic helices in lipid bilayers

Amphipathic helices in membrane proteins that interact with the hydrophobic/hydrophilic interface of the lipid bilayer have been difficult to structurally characterize. Here, the backbone structure and orientation of an amphipathic helix in the full-length M2 protein from influenza A virus has been characterized. The protein has been studied in hydrated DMPC/DMPG lipid bilayers above the gel to liquid-crystalline phase transition temperature by solid-state NMR spectroscopy. Characteristic PISA (Polar Index Slant Angle) wheels reflecting helical wheels have been observed in uniformly aligned bilayer preparations of both uniformly 15N labeled and amino acid specific labeled M2 samples. Hydrogen/deuterium exchange studies have shown the very slow exchange of some residues in the amphipathic helix and more rapid exchange for the transmembrane helix. These latter results clearly suggest the presence of an aqueous pore. A variation in exchange rate about the transmembrane helical axis provides additional support for this claim and suggests that motions occur about the helical axes in this tetramer to expose the entire backbone to the pore.