These scripts implement the combined test for feature selection in a multi-class comparison from data with missing values, presented in Plancade, Berland et al (2022). The test combines a GLMM model for the association between classes missingness, and a LMM on non-missing values. This metaproteomics data set extracted from the ProteoCardis cohort includes the XIC intensities at protein and peptide for 99 patients belonging to two classes, as well as 7 technical replicates for 8 biological samples. All details can be found in the readme file. Protein and peptide are not annotated as biological analysis is in progress.