hxsparse: A Feature Extraction Algorithm for Hydrogen Exchange-Mass Spectrometry Based on a Linear Deuteration Model and LASSO Regularization

Extraction of average masses of deuterated peptides from LC-MS data in hydrogen exchange-mass spectrometry (HX-MS) is a challenging task that is not well-suited to traditional LC-MS feature extraction algorithms. While commercial software is available, no feature extraction algorithms specific to HX-MS data analysis have been fully described in the literature. The present work describes such an algorithm. hxsparse uses an exact model that describes the deuteration of each peptide in terms of linear combinations of discrete deuteration states. Sparsity is introduced into the model using least absolute shrinkage and selection operator (LASSO) regularization. We demonstrate that automated feature extraction with hxsparse outperforms commercial software and nearly matches the results obtained from expert curation of HX-MS data.