HDXmodeller: an online webserver for high-resolution HDX-MS with auto-validation

Deep Learning Enables Automatic Correction of Experimental HDX-MS Data with Applications in Protein Modeling

Observed mass shifts associated with deuterium incorporation in hydrogen-deuterium exchange mass spectrometry (HDX-MS) frequently deviate from the initial signals due to back and forward exchange. In typical HDX-MS experiments, the impact of these disparities on data interpretation is generally low because relative and not absolute mass changes are investigated. However, for more advanced data processing including optimization, experimental error correction is imperative for accurate results. Here the potential for automatic HDX-MS data correction using models generated by deep neural networks is demonstrated. A multilayer perceptron (MLP) is used to learn a mapping between uncorrected HDX-MS data and data with mass shifts corrected for back and forward exchange. The model is rigorously tested at various levels including peptide level mass changes, residue level protection factors following optimization, and ability to correctly identify native protein folds using HDX-MS guided protein modeling. AI is shown to demonstrate considerable potential for amending HDX-MS data and improving fidelity across all levels. With access to big data, online tools may eventually be able to predict corrected mass shifts in HDX-MS profiles. This should improve throughput in workflows that require the reporting of real mass changes as well as allow retrospective correction of historic profiles to facilitate new discoveries with these data.

Mapping HDX-MS Data to Protein Conformations through Training Ensemble-Based Models

An original approach that adopts machine learning inference to predict protein structural information using hydrogen-deuterium exchange mass spectrometry (HDX-MS) is described. The method exploits an in-house optimization program that increases the resolution of HDX-MS data from peptides to amino acids. A system is trained using Gradient Tree Boosting as a type of machine learning ensemble technique to assign a protein secondary structure. Using limited training data we generate a discriminative model that uses optimized HDX-MS data to predict protein secondary structure with an accuracy of 75%. This research could form the basis for new methods exploiting artificial intelligence to model protein conformations by HDX-MS.

Artificial intelligence-based HDX (AI-HDX) prediction reveals fundamental characteristics to protein dynamics: Mechanisms on SARS-CoV-2 immune escape

Three-dimensional structure and dynamics are essential for protein function. Advancements in hydrogen-deuterium exchange (HDX) techniques enable probing protein dynamic information in physiologically relevant conditions. HDX-coupled mass spectrometry (HDX-MS) has been broadly applied in pharmaceutical industries. However, it is challenging to obtain dynamics information at the single amino acid resolution and time consuming to perform the experiments and process the data. Here, we demonstrate the first deep learning model, artificial intelligence-based HDX (AI-HDX), that predicts intrinsic protein dynamics based on the protein sequence. It uncovers the protein structural dynamics by combining deep learning, experimental HDX, sequence alignment, and protein structure prediction. AI-HDX can be broadly applied to drug discovery, protein engineering, and biomedical studies. As a demonstration, we elucidated receptor-binding domain structural dynamics as a potential mechanism of anti-severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) antibody efficacy and immune escape. AI-HDX fundamentally differs from the current AI tools for protein analysis and may transform protein design for various applications.

High-Resolution Hydrogen-Deuterium Protection Factors from Sparse Mass Spectrometry Data Validated by Nuclear Magnetic Resonance Measurements

Experimental measurement of time-dependent spontaneous exchange of amide protons with deuterium of the solvent provides information on the structure and dynamical structural variation in proteins. Two experimental techniques are used to probe the exchange: NMR, which relies on different magnetic properties of hydrogen and deuterium, and MS, which exploits the change in mass due to deuteration. NMR provides residue-specific information, that is, the rate of exchange or, analogously, the protection factor (i.e., the unitless ratio between the rate of exchange for a completely unstructured state and the observed rate). MS provides information that is specific to peptides obtained by proteolytic digestion. The spatial resolution of HDX-MS measurements depends on the proteolytic pattern of the protein, the fragmentation method used, and the overlap between peptides. Different computational approaches have been proposed to extract residue-specific information from peptide-level HDX-MS measurements. Here, we demonstrate the advantages of a method recently proposed that exploits self-consistency and classifies the possible sets of protection factors into a finite number of alternative solutions compatible with experimental data. The degeneracy of the solutions can be reduced (or completely removed) by exploiting the additional information encoded in the shape of the isotopic envelopes. We show how sparse and noisy MS data can provide high-resolution protection factors that correlate with NMR measurements probing the same protein under the same conditions.

Computational Modeling of Molecular Structures Guided by Hydrogen-Exchange Data

Data produced by hydrogen-exchange monitoring experiments have been used in structural studies of molecules for several decades. Despite uncertainties about the structural determinants of hydrogen exchange itself, such data have successfully helped guide the structural modeling of challenging molecular systems, such as membrane proteins or large macromolecular complexes. As hydrogen-exchange monitoring provides information on the dynamics of molecules in solution, it can complement other experimental techniques in so-called integrative modeling approaches. However, hydrogen-exchange data have often only been used to qualitatively assess molecular structures produced by computational modeling tools. In this paper, we look beyond qualitative approaches and survey the various paradigms under which hydrogen-exchange data have been used to quantitatively guide the computational modeling of molecular structures. Although numerous prediction models have been proposed to link molecular structure and hydrogen exchange, none of them has been widely accepted by the structural biology community. Here, we present as many hydrogen-exchange prediction models as we could find in the literature, with the aim of providing the first exhaustive list of its kind. From purely structure-based models to so-called fractional-population models or knowledge-based models, the field is quite vast. We aspire for this paper to become a resource for practitioners to gain a broader perspective on the field and guide research toward the definition of better prediction models. This will eventually improve synergies between hydrogen-exchange monitoring and molecular modeling.

Exploiting the Propagation of Constrained Variables for Enhanced HDX-MS Data Optimization

Nonlinear programming has found useful applications in protein biophysics to help understand the microscopic exchange kinetics of data obtained using hydrogen-deuterium exchange mass spectrometry (HDX-MS). Finding a microscopic kinetic solution for HDX-MS data provides a window into local protein stability and energetics allowing them to be quantified and understood. Optimization of HDX-MS data is a significant challenge, however, due to the requirement to solve a large number of variables simultaneously with exceptionally large variable bounds. Modeled rates are frequently uncertain with an explicate dependency on the initial guess values. In order to enhance the search for a minimum solution in HDX-MS optimization, the ability of selected constrained variables to propagate throughout the data is considered. We reveal that locally bound constrained optimization induces a global effect on all variables. The global response to local constraints is large and surprisingly long-range, but the outcome is unpredictable, unexpectedly decreasing the overall accuracy of certain data sets depending on the stringency of the constraints. Utilizing previously described in-house validation criteria based on covariance matrices, a method is described that is able to accurately determine whether constraints benefit or impair the optimization of HDX-MS data. From this, we establish a new two-stage method for our online optimizer HDXmodeller that can effectively leverage locally bound variables to enhance HDX-MS data modeling.

Probing Universal Protein Dynamics Using Hydrogen-Deuterium Exchange Mass Spectrometry-Derived Residue-Level Gibbs Free Energy

Hydrogen-deuterium exchange mass spectrometry (HDX-MS) is a powerful technique to monitor protein intrinsic dynamics. The technique provides high-resolution information on how protein intrinsic dynamics are altered in response to biological signals, such as ligand binding, oligomerization, or allosteric networks. However, identification, interpretation, and visualization of such events from HDX-MS data sets is challenging as these data sets consist of many individual data points collected across peptides, time points, and experimental conditions. Here, we present PyHDX, an open-source Python package and webserver, that allows the user to batch extract the universal quantity Gibbs free energy at residue levels over multiple protein conditions and homologues. The output is directly visualized on a linear map or 3D structures or is exported as .csv files or PyMOL scripts.

Advances in Hydrogen/Deuterium Exchange Mass Spectrometry and the Pursuit of Challenging Biological Systems

Solution-phase hydrogen/deuterium exchange (HDX) coupled to mass spectrometry (MS) is a widespread tool for structural analysis across academia and the biopharmaceutical industry. By monitoring the exchangeability of backbone amide protons, HDX-MS can reveal information about higher-order structure and dynamics throughout a protein, can track protein folding pathways, map interaction sites, and assess conformational states of protein samples. The combination of the versatility of the hydrogen/deuterium exchange reaction with the sensitivity of mass spectrometry has enabled the study of extremely challenging protein systems, some of which cannot be suitably studied using other techniques. Improvements over the past three decades have continually increased throughput, robustness, and expanded the limits of what is feasible for HDX-MS investigations. To provide an overview for researchers seeking to utilize and derive the most from HDX-MS for protein structural analysis, we summarize the fundamental principles, basic methodology, strengths and weaknesses, and the established applications of HDX-MS while highlighting new developments and applications.

Characterization and Management of Noise in HDX-MS Data Modeling

Quantification of hydrogen deuterium exchange (HDX) kinetics can provide information on the stability of individual amino acids in proteins by finding the degree to which the local backbone environment corresponds to that of a random coil. When characterized by mass spectrometry, extraction of HDX kinetics is not possible because different residue exchange rates become merged depending on the peptides that are formed during proteolytic digestion. We have recently developed an advanced programming tool called HDXmodeller, which enables the exchange rates of individual amino acids to be understood by optimization of low-resolution HDX-mass spectrometry (MS) data. HDXmodeller is also uniquely able to appraise each optimization and quantify the accuracy of modeled exchange rates ab initio using a novel autovalidation method based on a covariance matrix. Here, we address the noise-handling capabilities of HDXmodeller and demonstrate the effectiveness of the algorithm on self-inconsistent datasets. Reference intervals for experimental HDX-MS data are also derived, and this information is presented in an updated online workflow for HDXmodeller, allowing users to evaluate the consistency of their data. The development of a modified version of HDXmodeller is also discussed with enhanced noise-handling capability brought about through loss function optimization. Changes in optimizer accuracy with different loss functions are also demonstrated along with the effectiveness of HDXmodeller to select the most effective optimizer for different data using currently embedded autovalidation criteria.