Automated combined assignment of NOESY spectra and three-dimensional protein structure determination

A procedure for automated protein structure determination is presented that is based on an iterative procedure during which the NOESY peak list assignment and the structure calculation are performed simultaneously. The input consists of a list of NOESY peak positions and a list of chemical shifts as obtained from sequence-specific resonance assignment. For the present applications of this approach the previously introduced NOAH routine was implemented in the distance geometry program DIANA. As an illustration, experimental 2D and 3D NOESY cross-peak lists of six proteins have been analyzed, for which complete sequence-specific ¹H assignments are available for the polypeptide backbone and the amino acid side chains. The automated method assigned 70–90% of all NOESY cross peaks, which is on average 10% less than with the interactive approach, and only between 0.8% and 2.4% of the automatically assigned peaks had a different assignment than in the corresponding manually assigned peak lists. The structures obtained with NOAH/DIANA are in close agreement with those from manually assigned peak lists, and with both approaches the residual constraint violations correspond to high-quality NMR structure determinations. Systematic comparisons of the bundles of conformers that represent corresponding automatically and interactively determined structures document the absence of significant bias in either approach, indicating that an important step has been made towards automation of structure determination from NMR spectra.     

Robust and highly accurate automatic NOESY assignment and structure determination with Rosetta

We have developed a novel and robust approach for automatic and unsupervised simultaneous nuclear Overhauser effect (NOE) assignment and structure determination within the CS-Rosetta framework. Starting from unassigned peak lists and chemical shift assignments, autoNOE-Rosetta determines NOE cross-peak assignments and generates structural models. The approach tolerates incomplete and raw NOE peak lists as well as incomplete or partially incorrect chemical shift assignments, and its performance has been tested on 50 protein targets ranging from 50 to 200 residues in size. We find a significantly improved performance compared to established programs, particularly for larger proteins and for NOE data obtained on perdeuterated protein samples. X-ray crystallographic structures allowed comparison of Rosetta and conventional, PDB-deposited, NMR models in 20 of 50 test cases. The unsupervised autoNOE-Rosetta models were often of significantly higher accuracy than the corresponding expert-supervised NMR models deposited in the PDB. We also tested the method with unrefined peak lists and found that performance was nearly as good as for refined peak lists. Finally, demonstrating our method’s remarkable robustness against problematic input data, we provided correct models for an incorrect PDB-deposited NMR solution structure.     

Constant-time NOESY: An aid in the analysis of protein NMR spectra

Summary A constant-time version of the homonuclear NOESY experiment (CT-NOESY) is described. The experiment yields simplified protein spectra, in which cross peaks arising from protons with zero or small couplings are differentiated from other cross peaks, thus partially overcoming the problem of signal overlap. In addition, the CT-NOESY spectrum provides information on the magnitude of³J_NH- and³J coupling' constants, and is thus useful to determine torsion angle constraints and to perform stereospecific assignments of CHH protons in the case of³J constants.     

Protein NMR structure determination with automated NOE-identification in the NOESY spectra using the new software ATNOS

Novel algorithms are presented for automated NOESY peak picking and NOE signal identification in homonuclear 2D and heteronuclear-resolved 3D [¹H,¹H]-NOESY spectra during de novoprotein structure determination by NMR, which have been implemented in the new software ATNOS (automated NOESY peak picking). The input for ATNOS consists of the amino acid sequence of the protein, chemical shift lists from the sequence-specific resonance assignment, and one or several 2D or 3D NOESY spectra. In the present implementation, ATNOS performs multiple cycles of NOE peak identification in concert with automated NOE assignment with the software CANDID and protein structure calculation with the program DYANA. In the second and subsequent cycles, the intermediate protein structures are used as an additional guide for the interpretation of the NOESY spectra. By incorporating the analysis of the raw NMR data into the process of automated de novoprotein NMR structure determination, ATNOS enables direct feedback between the protein structure, the NOE assignments and the experimental NOESY spectra. The main elements of the algorithms for NOESY spectral analysis are techniques for local baseline correction and evaluation of local noise level amplitudes, automated determination of spectrum-specific threshold parameters, the use of symmetry relations, and the inclusion of the chemical shift information and the intermediate protein structures in the process of distinguishing between NOE peaks and artifacts. The ATNOS procedure has been validated with experimental NMR data sets of three proteins, for which high-quality NMR structures had previously been obtained by interactive interpretation of the NOESY spectra. The ATNOS-based structures coincide closely with those obtained with interactive peak picking. Overall, we present the algorithms used in this paper as a further important step towards objective and efficient de novoprotein structure determination by NMR.     

Toward a computer assisted analysis of NOESY spectra: a multivariate data analysis of an RNA NOESY spectrum

A multivariate data-representation of a portion of the H-NOESY spectrum of an RNA octamer duplex was used to explore the possibility of using Principal Component Analysis and Partial Least Squares Discrimination for pattern recognition. In this case, it is found that the methods can: (i) distinguish slices containing signal from those containing only noise, (ii) locate slices containing overlapping signals, and (iii) in some cases to segregate slices with unique aspects such as those from terminal nucleotides, overlapping signals, purine-H8, pyrimidine-H6 and adenine-H2 containing slices. These properties can easily be included in a scheme to automate spectral analysis. The formulation described here does not distinguish patterns needed to automate sequential assignment of resonances in NOESY spectra of RNA.     

Variability in automated assignment of NOESY spectra and three-dimensional structure determination: A test case on three small disulfide-bonded proteins

Three independent runs of automatic assignment and structure calculations were performed on three small proteins, calcicludine from the venom of the green mamba Dendroaspis angusticeps, -conotoxin PVIIA from the purple cone Conus purpurascens and HsTX1, a short scorpion toxin from the venom of Heterometrus spinnifer. At the end of all the runs, the number of cross peaks which remained unassigned (0.6%, 1.4% and 2% for calcicludine, -conotoxin and HsTX1, respectively), as well as the number of constraints which were rejected as producing systematic violations (2.7%, 1.0%, and 1.4% for calcicludine, -conotoxin and HsTX1, respectively) were low. The conformation of the initial model used in the procedure (linear model or constructed by homology) has no influence on the final structures. Mainly two parameters control the procedure: the chemical shift tolerance and the cut-off distance. Independent runs of structure calculations, using the same parameters, yield structures for which the rmsd between averaged structures and the rmsd around each averaged structure were of the same order of magnitude. A different cut-off distance and a different chemical shift tolerance yield rmsd values on final average structures which did not differ more than 0.5 Å compared to the rmsd obtained around the averaged structure for each calculation. These results show that the procedure is robust when applied to such a small disulfide-bonded protein.     

The optimization of protein secondary structure determination with infrared and circular dichroism spectra.

We have used the circular dichroism and infrared spectra of a specially designed 50 protein database [Oberg, K.A., Ruysschaert, J.M. & Goormaghtigh, E. (2003) Protein Sci. 12, 2015-2031] in order to optimize the accuracy of spectroscopic protein secondary structure determination using multivariate statistical analysis methods. The results demonstrate that when the proteins are carefully selected for the diversity in their structure, no smaller subset of the database contains the necessary information to describe the entire set. One conclusion of the paper is therefore that large protein databases, observing stringent selection criteria, are necessary for the prediction of unknown proteins. A second important conclusion is that only the comparison of analyses run on circular dichroism and infrared spectra independently is able to identify failed solutions in the absence of known structure. Interestingly, it was also found in the course of this study that the amide II band has high information content and could be used alone for secondary structure prediction in place of amide I.     

Photobleaching step analysis for robust determination of protein complex stoichiometries

The counting of discrete photobleaching steps in fluorescence microscopy is ideally suited to study protein complex stoichiometry in situ. The counting range of photobleaching step analysis has been significantly improved with more-sophisticated algorithms for step detection, albeit at an increasing computational cost and with the necessity for high-quality data. Here, we address concerns regarding robustness, automation, and experimental validation, optimizing both data acquisition and analysis. To make full use of the potential of photobleaching step analysis, we evaluate various labeling strategies with respect to their molecular brightness, photostability, and photoblinking. The developed analysis algorithm focuses on automation and computational efficiency. Moreover, we validate the developed methods with experimental data acquired on DNA origami labeled with defined fluorophore numbers, demonstrating counting of up to 35 fluorophores. Finally, we show the power of the combination of optimized trace acquisition and automated data analysis by counting labeled nucleoporin 107 in nuclear pore complexes of intact U2OS cells. The successful in situ application promotes this framework as a new resource enabling cell biologists to robustly determine the stoichiometries of molecular assemblies at the single-molecule level in an automated manner.     

Automated assignment and 3D structure calculations using combinations of 2D homonuclear and 3D heteronuclear NOESY spectra

The NOAH/DIAMOD suite uses feedback filtering and self-correcting distance geometry to generate 3D structures from unassigned NOESY spectra. In this study we determined the minimum set of experiments needed to generate a high quality structure bundle. Different combinations of 3D ¹⁵N-edited, ¹³C-edited HSQC-NOESY and 2D homonuclear ¹H-¹H NOESY spectra of the 77 amino acid protein, myeloid progenitor inhibitory factor-1 (MPIF-1) were used as input for NOAH/DIAMOD calculations. The quality of the assignments of NOESY cross peaks and the accuracy of the automatically generated 3D structures were compared to those obtained with a conventional manual procedure. Combining data from two types of experiments synergistically increased the number of peaks assigned unambiguously in both individual spectra. As a general trend for the accuracy of the structures we observed structural variations in the backbone fold of the final structures of about 2 Å for single spectral data, of 1 Å to 1.5 Å for double spectral data, and of 0.6 Å for triple spectral data sets. The quality of the assignments and 3D structures from the optimal data using all three spectra were similar to those obtained from traditional assignment methods with structural variations within the bundle of 0.6 Å and 1.3 Å for backbone and heavy atoms, respectively. Almost all constraints (97%) of the automatic NOESY cross peak assignments were cross compatible with the structures from the conventional manual assignment procedure, and an even larger proportion (99%) of the manually derived constraints were compatible with the automatically determined 3D structures. The two mean structures determined by both methods differed only by 1.3 Å rmsd for the backbone atoms in the well-defined regions of the protein. Thus NOAD/DIAMOD analysis of spectra from labeled proteins provides a reliable method for high throughput analysis of genomic targets.     

K2D2: Estimation of protein secondary structure from circular dichroism spectra

Background  

Circular dichroism spectroscopy is a widely used technique to analyze the secondary structure of proteins in solution. Predictive methods use the circular dichroism spectra from proteins of known tertiary structure to assess the secondary structure contents of a protein with unknown structure given its circular dichroism spectrum.     

Automated robust and accurate assignment of protein resonances for solid state NMR

The process of resonance assignment represents a time-consuming and potentially error-prone bottleneck in structural studies of proteins by solid-state NMR (ssNMR). Software for the automation of this process is therefore of high interest. Procedures developed through the last decades for solution-state NMR are not directly applicable for ssNMR due to the inherently lower data quality caused by lower sensitivity and broader lines, leading to overlap between peaks. Recently, the first efforts towards procedures specifically aimed for ssNMR have been realized (Schmidt et al. in J Biomol NMR 56(3):243–254, 2013). Here we present a robust automatic method, which can accurately assign protein resonances using peak lists from a small set of simple 2D and 3D ssNMR experiments, applicable in cases with low sensitivity. The method is demonstrated on three uniformly ¹³C, ¹⁵N labeled biomolecules with different challenges on the assignments. In particular, for the immunoglobulin binding domain B1 of streptococcal protein G automatic assignment shows 100 % accuracy for the backbone resonances and 91.8 % when including all side chain carbons. It is demonstrated, by using a procedure for generating artificial spectra with increasing line widths, that our method, GAMES_ASSIGN can handle a significant amount of overlapping peaks in the assignment. The impact of including different ssNMR experiments is evaluated as well.     

Evaluation of the information content in infrared spectra for protein secondary structure determination

Fourier-transform infrared spectroscopy is a method of choice for the experimental determination of protein secondary structure. Numerous approaches have been developed during the past 15 years. A critical parameter that has not been taken into account systematically is the selection of the wavenumbers used for building the mathematical models used for structure prediction. The high quality of the current Fourier-transform infrared spectrometers makes the absorbance at every single wavenumber a valid and almost noiseless type of information. We address here the question of the amount of independent information present in the infrared spectra of proteins for the prediction of the different secondary structure contents. It appears that, at most, the absorbance at three distinct frequencies of the spectra contain all the nonredundant information that can be related to one secondary structure content. The ascending stepwise method proposed here identifies the relevance of each wavenumber of the infrared spectrum for the prediction of a given secondary structure and yields a particularly simple method for computing the secondary structure content. Using the 50-protein database built beforehand to contain as little fold redundancy as possible, the standard error of prediction in cross-validation is 5.5% for the alpha-helix, 6.6% for the beta-sheet, and 3.4% for the beta-turn.     

Rapid and accurate structure determination of coiled-coil domains using NMR dipolar couplings: application to cGMP-dependent protein kinase Ialpha

Coiled-coil motifs play essential roles in protein assembly and molecular recognition, and are therefore the targets of many ongoing structural and functional studies. However, owing to the dynamic nature of many of the smaller coiled-coil domains, crystallization for X-ray studies is very challenging. Determination of elongated structures using standard NMR approaches is inefficient and usually yields low-resolution structures due to accumulation of small errors over long distances. Here we describe a solution NMR approach based on residual dipolar couplings (RDCs) for rapid and accurate structure determination of coiled-coil dimers. Using this approach, we were able to determine the high-resolution structure of the coiled-coil domain of cGMP-dependent protein kinase Ialpha, a protein of previously unknown structure that is critical for physiological relaxation of vascular smooth muscle. This approach can be extended to solve coiled-coil structures with higher order assemblies.     

pParse: a method for accurate determination of monoisotopic peaks in high-resolution mass spectra

Determining the monoisotopic peak of a precursor is a first step in interpreting mass spectra, which is basic but non-trivial. The reason is that in the isolation window of a precursor, other peaks interfere with the determination of the monoisotopic peak, leading to wrong mass-to-charge ratio or charge state. Here we propose a method, named pParse, to export the most probable monoisotopic peaks for precursors, including co-eluted precursors. We use the relationship between the position of the highest peak and the mass of the first peak to detect candidate clusters. Then, we extract three features to sort the candidate clusters: (i) the sum of the intensity, (ii) the similarity of the experimental and the theoretical isotopic distribution, and (iii) the similarity of elution profiles. We showed that the recall of pParse, MaxQuant, and BioWorks was 98-98.8%, 0.5-17%, and 1.8-36.5% at the same precision, respectively. About 50% of tandem mass spectra are triggered by multiple precursors which are difficult to identify. Then we design a new scoring function to identify the co-eluted precursors. About 26% of all identified peptides were exclusively from co-eluted peptides. Therefore, accurately determining monoisotopic peaks, including co-eluted precursors, can greatly increase peptide identification rate.     

Fast, robust, and accurate determination of transmission electron microscopy contrast transfer function

Transmission electron microscopy, as most imaging devices, introduces optical aberrations that in the case of thin specimens are usually modeled in Fourier space by the so-called contrast transfer function (CTF). Accurate determination of the CTF is crucial for its posterior correction. Furthermore, the CTF estimation must be fast and robust if high-throughput three-dimensional electron microscopy (3DEM) studies are to be carried out. In this paper we present a robust algorithm that fits a theoretical CTF model to the power spectrum density (PSD) measured on a specific micrograph or micrograph area. Our algorithm is capable of estimating the envelope of the CTF which is absolutely needed for the correction of the CTF amplitude changes.     

Estimation of protein secondary structure and error analysis from circular dichroism spectra

The estimation of protein secondary structure from circular dichroism spectra is described by a multivariate linear model with noise (Gauss-Markoff model). With this formalism the adequacy of the linear model is investigated, paying special attention to the estimation of the error in the secondary structure estimates. It is shown that the linear model is only adequate for the alpha-helix class. Since the failure of the linear model is most likely due to nonlinear effects, a locally linearized model is introduced. This model is combined with the selection of the estimate whose fractions of secondary structure summate to approximately one. Comparing the estimation from the CD spectra with the X-ray data (by using the data set of W.C. Johnson Jr., 1988, Annu. Rev. Biophys. Chem. 17, 145-166) the root mean square residuals are 0.09 (alpha-helix), 0.12 (anti-parallel beta-sheet), 0.08 (parallel beta-sheet), 0.07 (beta-turn), and 0.09 (other). These residuals are somewhat larger than the errors estimated from the locally linearized model. In addition to alpha-helix, in this model the beta-turn and "other" class are estimated adequately. But the estimation of the antiparallel and parallel beta-sheet class remains unsatisfactory. We compared the linear model and the locally linearized model with two other methods (S. W. Provencher and J. Gl?ckner, 1981, Biochemistry 20, 1085-1094; P. Manavalan and W. C. Johnson Jr., 1988, Anal. Biochem. 167, 76-85). The locally linearized model and the Provencher and Gl?ckner method provided the smallest residuals. However, an advantage of the locally linearized model is the estimation of the error in the secondary structure estimates.     

Determination of protein secondary structure using factor analysis of infrared spectra

A method is presented for determining the secondary structural composition of a protein in aqueous solution from its infrared spectrum. A factor analysis approach is used to analyze the infrared spectra of 18 proteins whose crystal structures are known from X-ray studies. Factor analysis followed by multiple linear regression identifies those eigenspectra that correlate with the variation in properties described by the calibration set. The properties of interest in this study are % alpha-helix, % beta-sheet, and % turns. In the analysis of an unknown, the factor loadings required to reproduce its spectrum are substituted in the regression equation for each property to predict its secondary structural composition. The accuracy of the method was determined by removing each standard, in turn, from the calibration set and using a calibration set generated from the remainder to predict its composition. By this method we obtain standard errors of prediction of 3.9% for alpha-helix, 8.3% for beta-sheet, and 6.6% for turns. The method may also be applied to the spectra of proteins in 2H2O. The method has important advantages over those currently in use for the quantitative analysis of the infrared spectra of proteins. Manipulation of the spectrum is kept to a minimum, no curve-fitting is necessary, and the several amide I band components need not be assigned.     

Spectral editing of two-dimensional magic-angle-spinning solid-state NMR spectra for protein resonance assignment and structure determination

Several techniques for spectral editing of 2D ¹³C?C¹³C correlation NMR of proteins are introduced. They greatly reduce the spectral overlap for five common amino acid types, thus simplifying spectral assignment and conformational analysis. The carboxyl (COO) signals of glutamate and aspartate are selected by suppressing the overlapping amide N?CCO peaks through ¹³C?C¹⁵N dipolar dephasing. The sidechain methine (CH) signals of valine, lecuine, and isoleucine are separated from the overlapping methylene (CH₂) signals of long-chain amino acids using a multiple-quantum dipolar transfer technique. Both the COO and CH selection methods take advantage of improved dipolar dephasing by asymmetric rotational-echo double resonance (REDOR), where every other ??-pulse is shifted from the center of a rotor period t_r by about 0.15 t_r. This asymmetry produces a deeper minimum in the REDOR dephasing curve and enables complete suppression of the undesired signals of immobile segments. Residual signals of mobile sidechains are positively identified by dynamics editing using recoupled ¹³C?C¹H dipolar dephasing. In all three experiments, the signals of carbons within a three-bond distance from the selected carbons are detected in the second spectral dimension via ¹³C spin exchange. The efficiencies of these spectral editing techniques range from 60?% for the COO and dynamic selection experiments to 25?% for the CH selection experiment, and are demonstrated on well-characterized model proteins GB1 and ubiquitin.     

TASSER_WT: a protein structure prediction algorithm with accurate predicted contact restraints for difficult protein targets

To improve the prediction accuracy in the regime where template alignment quality is poor, an updated version of TASSER_2.0, namely TASSER_WT, was developed. TASSER_WT incorporates more accurate contact restraints from a new method, COMBCON. COMBCON uses confidence-weighted contacts from PROSPECTOR_3.5, the latest version, PROSPECTOR_4, and a new local structural fragment-based threading algorithm, STITCH, implemented in two variants depending on expected fragment prediction accuracy. TASSER_WT is tested on 622 Hard proteins, the most difficult targets (incorrect alignments and/or templates and incorrect side-chain contact restraints) in a comprehensive benchmark of 2591 nonhomologous, single domain proteins ≤200 residues that cover the PDB at 35% pairwise sequence identity. For 454 of 622 Hard targets, COMBCON provides contact restraints with higher accuracy and number of contacts per residue. As contact coverage with confidence weight ≥3 (F^wt≥3_cov) increases, the more improved are TASSER_WT models. When F^wt≥3_cov > 1.0 and > 0.4, the average root mean-square deviation of TASSER_WT (TASSER_2.0) models is 4.11 Å (6.72 Å) and 5.03 Å (6.40 Å), respectively. Regarding a structure prediction as successful when a model has a TM-score to the native structure ≥0.4, when F^wt≥3_cov > 1.0 and > 0.4, the success rate of TASSER_WT (TASSER_2.0) is 98.8% (76.2%) and 93.7% (81.1%), respectively.