Bimodal peptide collision cross section distribution reflects two stable conformations in the gas phase

Recent high throughput applications to shotgun proteomics have shown great benefits of coupling ion mobility spectrometry (IMS) to mass spectrometry. IMS adds a separation dimension by differentiating biomolecules by their size and shape. We (and others) find that the distribution of peptide collision cross section (CCS) is often bimodal, which limits the utility of current machine learning predictions for peptide identification. Molecular dynamics simulations indicate that the peptides in the drift tube can adopt multiple stable conformations and that the two modes correspond to predominantly extended (mostly helical) and more compact (globular and less ordered) conformations. Most peptides have a charge-dependent strong preference for one of the two conformations, while some can adapt both, as evidenced by a simple geometric model of the CCS data. We suggest a novel two-valued CCS predictor allowing for multiple peptide conformations. Its integration into data-independent acquisition proteomics increases identification rates of peptides compared to single-value predictors.