Characterization and Management of Noise in HDX-MS Data Modeling

Abstract

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.

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

“…The ability to extract k obs directly from isotopically labeled peptide pairs was evaluated using previously obtained experimental HDX-MS data for alpha lactalbumin. This data set has been shown to have a high correspondence with model HDX-MS projections obtained following optimization and is therefore expected to yield subtracted profiles that are described well by an exponential function with minimal error. , The data set also has many instances where different peptide pairs permit k obs calculations for the same residue providing an additional test for the efficacy of this approach since k obs calculations should be independent of the peptide pairs used. Time-dependent isotope levels were obtained from different peptide pairs by data subtraction (Methods).…”

“…15 Where data lack natural breaks, the preparation of smaller data subsections can be facilitated by deletion of specific peptides selected by consideration of peptide error and occupancy scores. 16 Frequent evaluation runs consisting of 10 replicate optimizations are recommended with the R-matrix outputs used as guides on the preparation of input data for production runs. HDXmodeller will now output a new constraint file following any optimization.…”

“…Users should follow recommended guidelines, and where possible, data should be subdivided into smaller subsections each submitted as unique jobs . Where data lack natural breaks, the preparation of smaller data subsections can be facilitated by deletion of specific peptides selected by consideration of peptide error and occupancy scores . Frequent evaluation runs consisting of 10 replicate optimizations are recommended with the R-matrix outputs used as guides on the preparation of input data for production runs.…”

“…Our own solution to this problem has been the development of HDXmodeller an online optimizer for HDX-MS with a novel validation method based on a covariance matrix (eqs S1–S3). , The validation method calculates the pairwise correlation coefficients between optimization replicates and reports the mean value or R-matrix which appears to relay information on the quality of the restraints. There is a very strong connection between R-matrix and the accuracy of the optimization, and since this term is derived ab initio , it provides essential feedback of only the likely accuracy of modeled outputs.…”

Computational Modeling of Molecular Structures Guided by Hydrogen-Exchange Data

“…76,77 Nevertheless, recent developments in this area have demonstrated the possibility of validating NLP outcomes a priori, providing a more robust strategy to model exchange rates in HDX-MS data. 78,79 Note that other ideas have been suggested to obtain HDX-MS data at the amino acid level. 80−83 An additional and often overlooked limitation of HDX-MS stems from the phenomenon of extraneous exchange, which encompasses both back exchange and forward exchange.…”