A Shrinkage-based statistical method for testing group mean differences in quantitative bottom-up proteomics

A statistical method for testing group mean differences in quantitative bottom-up proteomics is proposed. A probabilistic graphical model is introduced that explains the performance degradation of traditional statistical testing methods by random variations in ionization efficiency. The proposed method enhances a traditional feature-based test statistic by integrating nonparametric shrinkage estimation of covariance matrices and bootstrap estimation of degrees-of-freedom. Numerical experiments, based on simulated and real quantitative tandem mass spectrometry data, illustrate that the proposed method surpasses traditional statistical methods in specificity and sensitivity, particularly under conditions of small sample sizes.