Annotation of Human Exome Gene Variants with Consensus Pathogenicity

A novel approach is developed to address the challenge of annotating with phenotypic effects those exome variants for which relevant empirical data are lacking or minimal. The predictive annotation method is implemented as a stacked ensemble of supervised base-learners, including distributed random forest and gradient boosting machines. Ensemble models were trained and cross-validated on evidence-based categorical variant effect annotations from the ClinVar database, and were applied to 84 million non-synonymous single nucleotide variants (SNVs). The consensus model combined 39 functional mutation impacts, cross-species conservation score, and gene indispensability score. The indispensability score, accounting for differences in variant pathogenicities including in essential and mutation-tolerant genes, considerably improved the predictions. The consensus combination is consistent with as many input scores as possible while minimizing false predictions. The input scores are ranked based on their ability to predict effects. The score rankings and categorical phenotypic variant effect predictions are aimed for direct use in clinical and biological applications to prioritize human exome variants and mutations.

Expitope 2.0: a tool to assess immunotherapeutic antigens for their potential cross-reactivity against naturally expressed proteins in human tissues

BACKGROUND

Adoptive immunotherapy offers great potential for treating many types of cancer but its clinical application is hampered by cross-reactive T cell responses in healthy human tissues, representing serious safety risks for patients. We previously developed a computational tool called Expitope for assessing cross-reactivity (CR) of antigens based on tissue-specific gene expression. However, transcript abundance only indirectly indicates protein expression. The recent availability of proteome-wide human protein abundance information now facilitates a more direct approach for CR prediction. Here we present a new version 2.0 of Expitope, which computes all naturally possible epitopes of a peptide sequence and the corresponding CR indices using both protein and transcript abundance levels weighted by a proposed hierarchy of importance of various human tissues.

RESULTS

We tested the tool in two case studies: The first study quantitatively assessed the potential CR of the epitopes used for cancer immunotherapy. The second study evaluated HLA-A*02:01-restricted epitopes obtained from the Immune Epitope Database for different disease groups and demonstrated for the first time that there is a high variation in the background CR depending on the disease state of the host: compared to a healthy individual the CR index is on average two-fold higher for the autoimmune state, and five-fold higher for the cancer state.

CONCLUSIONS

The ability to predict potential side effects in normal tissues helps in the development and selection of safer antigens, enabling more successful immunotherapy of cancer and other diseases.

Peak picking NMR spectral data using non-negative matrix factorization

BACKGROUND

Simple peak-picking algorithms, such as those based on lineshape fitting, perform well when peaks are completely resolved in multidimensional NMR spectra, but often produce wrong intensities and frequencies for overlapping peak clusters. For example, NOESY-type spectra have considerable overlaps leading to significant peak-picking intensity errors, which can result in erroneous structural restraints. Precise frequencies are critical for unambiguous resonance assignments.

RESULTS

To alleviate this problem, a more sophisticated peaks decomposition algorithm, based on non-negative matrix factorization (NMF), was developed. We produce peak shapes from Fourier-transformed NMR spectra. Apart from its main goal of deriving components from spectra and producing peak lists automatically, the NMF approach can also be applied if the positions of some peaks are known a priori, e.g. from consistently referenced spectral dimensions of other experiments.

CONCLUSIONS

Application of the NMF algorithm to a three-dimensional peak list of the 23 kDa bi-domain section of the RcsD protein (RcsD-ABL-HPt, residues 688-890) as well as to synthetic HSQC data shows that peaks can be picked accurately also in spectral regions with strong overlap.

Effects of NMR spectral resolution on protein structure calculation

Adequate digital resolution and signal sensitivity are two critical factors for protein structure determinations by solution NMR spectroscopy. The prime objective for obtaining high digital resolution is to resolve peak overlap, especially in NOESY spectra with thousands of signals where the signal analysis needs to be performed on a large scale. Achieving maximum digital resolution is usually limited by the practically available measurement time. We developed a method utilizing non-uniform sampling for balancing digital resolution and signal sensitivity, and performed a large-scale analysis of the effect of the digital resolution on the accuracy of the resulting protein structures. Structure calculations were performed as a function of digital resolution for about 400 proteins with molecular sizes ranging between 5 and 33 kDa. The structural accuracy was assessed by atomic coordinate RMSD values from the reference structures of the proteins. In addition, we monitored also the number of assigned NOESY cross peaks, the average signal sensitivity, and the chemical shift spectral overlap. We show that high resolution is equally important for proteins of every molecular size. The chemical shift spectral overlap depends strongly on the corresponding spectral digital resolution. Thus, knowing the extent of overlap can be a predictor of the resulting structural accuracy. Our results show that for every molecular size a minimal digital resolution, corresponding to the natural linewidth, needs to be achieved for obtaining the highest accuracy possible for the given protein size using state-of-the-art automated NOESY assignment and structure calculation methods.

Estimating structure quality trends in the Protein Data Bank by equivalent resolution

The quality of protein structures obtained by different experimental and ab-initio calculation methods varies considerably. The methods have been evolving over time by improving both experimental designs and computational techniques, and since the primary aim of these developments is the procurement of reliable and high-quality data, better techniques resulted on average in an evolution toward higher quality structures in the Protein Data Bank (PDB). Each method leaves a specific quantitative and qualitative "trace" in the PDB entry. Certain information relevant to one method (e.g. dynamics for NMR) may be lacking for another method. Furthermore, some standard measures of quality for one method cannot be calculated for other experimental methods, e.g. crystal resolution or NMR bundle RMSD. Consequently, structures are classified in the PDB by the method used. Here we introduce a method to estimate a measure of equivalent X-ray resolution (e-resolution), expressed in units of Å, to assess the quality of any type of monomeric, single-chain protein structure, irrespective of the experimental structure determination method. We showed and compared the trends in the quality of structures in the Protein Data Bank over the last two decades for five different experimental techniques, excluding theoretical structure predictions. We observed that as new methods are introduced, they undergo a rapid method development evolution: within several years the e-resolution score becomes similar for structures obtained from the five methods and they improve from initially poor performance to acceptable quality, comparable with previously established methods, the performance of which is essentially stable.

Blind testing of routine, fully automated determination of protein structures from NMR data

The protocols currently used for protein structure determination by nuclear magnetic resonance (NMR) depend on the determination of a large number of upper distance limits for proton-proton pairs. Typically, this task is performed manually by an experienced researcher rather than automatically by using a specific computer program. To assess whether it is indeed possible to generate in a fully automated manner NMR structures adequate for deposition in the Protein Data Bank, we gathered 10 experimental data sets with unassigned nuclear Overhauser effect spectroscopy (NOESY) peak lists for various proteins of unknown structure, computed structures for each of them using different, fully automatic programs, and compared the results to each other and to the manually solved reference structures that were not available at the time the data were provided. This constitutes a stringent "blind" assessment similar to the CASP and CAPRI initiatives. This study demonstrates the feasibility of routine, fully automated protein structure determination by NMR.

Fast automated NMR spectroscopy of short-lived biological samples

Faster than death: NMR techniques that make use of nonlinear sampling and hyperdimensional processing enable the recording of complete NMR data sets for the automated assignment of the backbone and side-chain resonances of short-lived protein samples of cell lysates.

Protein structure validation by generalized linear model root-mean-square deviation prediction

Large-scale initiatives for obtaining spatial protein structures by experimental or computational means have accentuated the need for the critical assessment of protein structure determination and prediction methods. These include blind test projects such as the critical assessment of protein structure prediction (CASP) and the critical assessment of protein structure determination by nuclear magnetic resonance (CASD-NMR). An important aim is to establish structure validation criteria that can reliably assess the accuracy of a new protein structure. Various quality measures derived from the coordinates have been proposed. A universal structural quality assessment method should combine multiple individual scores in a meaningful way, which is challenging because of their different measurement units. Here, we present a method based on a generalized linear model (GLM) that combines diverse protein structure quality scores into a single quantity with intuitive meaning, namely the predicted coordinate root-mean-square deviation (RMSD) value between the present structure and the (unavailable) "true" structure (GLM-RMSD). For two sets of structural models from the CASD-NMR and CASP projects, this GLM-RMSD value was compared with the actual accuracy given by the RMSD value to the corresponding, experimentally determined reference structure from the Protein Data Bank (PDB). The correlation coefficients between actual (model vs. reference from PDB) and predicted (model vs. "true") heavy-atom RMSDs were 0.69 and 0.76, for the two datasets from CASD-NMR and CASP, respectively, which is considerably higher than those for the individual scores (-0.24 to 0.68). The GLM-RMSD can thus predict the accuracy of protein structures more reliably than individual coordinate-based quality scores.

Automatic assignment of protein backbone resonances by direct spectrum inspection in targeted acquisition of NMR data

The necessity to acquire large multidimensional datasets, a basis for assignment of NMR resonances, results in long data acquisition times during which substantial degradation of a protein sample might occur. Here we propose a method applicable for such a protein for automatic assignment of backbone resonances by direct inspection of multidimensional NMR spectra. In order to establish an optimal balance between completeness of resonance assignment and losses of cross-peaks due to dynamic processes/degradation of protein, assignment of backbone resonances is set as a stirring criterion for dynamically controlled targeted nonlinear NMR data acquisition. The result is demonstrated with the 12 kDa (13)C,(15) N-labeled apo-form of heme chaperone protein CcmE, where hydrolytic cleavage of 29 C-terminal amino acids is detected. For this protein, 90 and 98% of manually assignable resonances are automatically assigned within 10 and 40 h of nonlinear sampling of five 3D NMR spectra, respectively, instead of 600 h needed to complete the full time domain grid. In addition, resonances stemming from degradation products are identified. This study indicates that automatic resonance assignment might serve as a guiding criterion for optimal run-time allocation of NMR resources in applications to proteins prone to degradation.

Removal of a time barrier for high-resolution multidimensional NMR spectroscopy

We introduce the recursive multidimensional decomposition (R-MDD) method to speed recording of high-resolution NMR spectra. The measurement time is logarithmically dependent on the sizes of indirect spectral dimensions. R-MDD has the sensitivity and resolution advantages of optimized nonuniform acquisition schemes and is applicable to all types of biomolecular spectra. We demonstrated it for triple resonance experiments on three globular proteins (ubiquitin, azurin and the barstar-barnase complex) of 8-22 kDa.

Optimization of resolution and sensitivity of 4D NOESY using multi-dimensional decomposition

Highly resolved multi-dimensional NOE data are essential for rapid and accurate determination of spatial protein structures such as in structural genomics projects. Four-dimensional spectra contain almost no spectral overlap inherently present in lower dimensionality spectra and are highly amenable to application of automated routines for spectral resonance location and assignment. However, a high resolution 4D data set using conventional uniform sampling usually requires unacceptably long measurement time. Recently we have reported that the use of non-uniform sampling and multi-dimensional decomposition (MDD) can remedy this problem. Here we validate accuracy and robustness of the method, and demonstrate its usefulness for fully protonated protein samples. The method was applied to 11 kDa protein PA1123 from structural genomics pipeline. A systematic evaluation of spectral reconstructions obtained using 15-100% subsets of the complete reference 4D 1H-13C-13C-1H NOESY spectrum has been performed. With the experimental time saving of up to six times, the resolution and the sensitivity per unit time is shown to be similar to that of the fully recorded spectrum. For the 30% data subset we demonstrate that the intensities in the reconstructed and reference 4D spectra correspond with a correlation coefficient of 0.997 in the full range of spectral amplitudes. Intensities of the strong, middle and weak cross-peaks correlate with coefficients 0.9997, 0.9965, and 0.83. The method does not produce false peaks. 2% of weak peaks lost in the 30% reconstruction is in line with theoretically expected noise increase for the shorter measurement time. Together with good accuracy in the relative line-widths these translate to reliable distance constrains derived from sparsely sampled, high resolution 4D NOESY data.