Regression models to Detect and Quantify Peptides from Mass Spectra

Abstract

Data-independent acquisition (DIA) mass spectrometry-based proteomics aims to quantify every peptide and its derivatives in a sample by systematically sampling every ion. However, much of the signal in the resulting spectra is difficult to interpret because they represent complex mixtures of ions, preventing the accurate quantification of every peptide. I propose regularized linear regression approaches to jointly account for mixtures of multiple peptides and their relationships in DIA spectra to deconvolve spectra precursor and fragment spectra, remove the problem of interference, and improve the sensitivity and precision of peptide detection and quantification. The deconvolution extracts information invisible to current methods and provides a framework to detect and quantify more peptides.