Quantification and differential analysis of mass spectrometry proteomics data with probabilistic recovery of information from missing values

Mass spectrometry (MS) is the technology standard for expression proteomics, but statistical analysis of the resulting data is complicated by the occurrence of missing values. Missing values remain ubiquitous in MS-based proteomics data and are especially frequent in the emerging fields of single-cell and spatial proteomics. The limpa package implements new methods for quantification and differential expression analysis of MS proteomics data, including probabilistic information recovery from missing values. limpa summarises peptide-level data to estimate an expression value for every protein in every sample. Expression values that are supported by fewer detected peptides or involve more missing values are treated as less precise and are downweighted in the differential expression analysis, maximising statistical power while avoiding false discoveries. limpa produces a linear model object suitable for downstream analysis with the limma package, allowing complex experimental designs and other downstream tasks such as the gene ontology or pathway analysis.