Rnm = Inm/Mean(Im)
Where n represents the sample and m represents the protein. The raw MS data have been uploaded to the ProteomeXchange Consortium (PXD039974) via the iProX partner[8].
Mass spectrometry quality control detection and sample repeatability test
To test the the quantitative repeatability of samples was tested by three statistical analysis methods, including Pearson’s correlation coefficient (PCC), principal component analysis (PCA), and coefficient of variation (CV).
Bioinformatic analyses
Single sample gene set enrichment analysis (ssGSEA)
The estimation scores of the specific signaling pathways was acquired by R package GSVA package with ssGSEA algorithm[9]. The differences among the three groups were identified by the Kruskal-Wallis test.
Assessment of protein expression pattern by fuzzy c-means
The overall identified proteins and the differentially expressed proteins with similar expression patterns were grouped into different clusters using ClusterGVis package in R (https://github.com/junjunlab/ClusterGVis). Metascape and DAVID databases were applied for functional enrichment. The identified transcription factors were screened according to the earlier-published review[10].
Weighted gene co-expression network analysis (WGCNA)
R package WGCNA is applied to identify the interested protein sets and establish the association between protein modules and clinical traits[11].