Abstract
Researchers are increasingly turning to label-free MS1 intensity-based quantification strategies within HPLC-ESI-MS/MS workflows to reveal biological variation at the molecule level. Unfortunately, HPLC-ESI-MS/MS workflows using these strategies produce results with poor repeatability and reproducibility, primarily due to systematic bias and complex variability. While current global normalization strategies can mitigate systematic bias, they fail when faced with complex variability stemming from transient stochastic events during HPLC-ESI-MS/MS analysis. To address these problems, we developed a novel local normalization method, proximity-based intensity normalization (PIN), based on the analysis of compositional data. We evaluated PIN against common normalization strategies. PIN outperforms them in dramatically reducing variance and in identifying 20% more proteins with statistically significant abundance differences that other strategies missed. Our results show the PIN enables the discovery of statistically significant biological variation that otherwise is falsely reported or missed.