msqrob2TMT: Robust Linear Mixed Models for Inferring Differential Abundant Proteins in Labeled Experiments With Arbitrarily Complex Design

Labeling strategies in mass spectrometry-based proteomics enhance sample throughput by enabling the acquisition of multiplexed samples within a single run. However, contemporary experiments often involve increasingly complex designs, where the number of samples exceeds the capacity of a single run, resulting in a complex correlation structure that must be addressed for accurate statistical inference and reliable biomarker discovery. To this end, we introduce msqrob2TMT, a suite of mixed model-based workflows specifically designed for differential abundance analysis in labeled mass spectrometry-based proteomics data. msqrob2TMT accommodates both sample-specific and feature-specific (e.g., peptide or protein) covariates, facilitating inference in experiments with arbitrarily complex designs and allowing for explicit correction of feature-specific covariates. We benchmark our innovative workflows against state-of-the-art tools, including DEqMS, MSstatsTMT, and msTrawler, using two spike-in studies. Our findings demonstrate that msqrob2TMT offers greater flexibility, improved modularity, and enhanced performance, particularly through the application of robust ridge regression. Finally, we demonstrate the practical relevance of msqrob2TMT in a real mouse study, highlighting its capacity to effectively account for the complex correlation structure in the data.