Learning the shotgun proteomics measurement process enables global absolute protein quantification

Abstract

Accurate measurement of protein concentrations is essential for quantitative understanding of biological systems. Mass spectrometry-based proteomics is uniquely suited for identifying and quantifying thousands of proteins in a single experiment. However, it yields relative rather than absolute protein concentrations, preventing comparisons across experiments and species and constraining quantitative modeling of biological systems. This limitation arises because proteins are measured through their constituent peptides, whose signals are distorted by peptide-specific response factors and non-random missingness. Here, I first develop an improved sample preparation protocol that enhances proteolytic digestion efficiency, reducing arginine missed cleavages from ~30% to <1%. Next, I develop a Bayesian framework for absolute protein quantification (BAPQuant) that integrates sequence-dependent peptide ionization efficiency and peptide detection probability to infer absolute protein concentrations from MS data. Using orthogonally derived reference data, BAPQuant achieves high accuracy (R² = 0.95, median absolute percent error ~24%) and provides protein-specific credible intervals. These advances enable proteome-wide comparisons of protein abundance across species. I leverage this capability to compare protein abundances between Xenopus laevis and Ambystoma mexicanum (axolotl) and investigate how a roughly tenfold difference in genome size influences cell size.