We present here a new method for automatic meta-analysis of proteomic articles using assessment of frequency of individual protein names in the text. The list of all possible human protein names including synonyms was retrieved from UniProt knowledgebase. The retrieved names were searched in full-texts of peer-reviewed publications from electronic version of "Proteomics" journal and from PubMedCentral. In the automatic mode we have confirmed the earlier list of proteins [Petrak et al., Proteomics (2008) 8, 1744] most frequently reported as differentially expressed (DEPs) in human tissues. We have also verified, that the most recurrent proteins were reported in proteomic papers regardless of tissue, experimental goals or, to some extent, experimental methods employed. Frequently reported DEPs were: annexins, peroxiredoxins, alpha-enolase, triosephosphate isomerase, and HSP60. Besides, serum albumin, cathepsin D and vimentin were observed with relatively high frequency. The DEPs were reported in papers related to oncological, cardiovascular and neuronal diseases, and were involved in such biological processes as inflammation, cell regulation, immune responce and signal transduction. We conclude that automatic meta-analysis of proteomic papers enabled extraction of frequently reported proteins that are most likely the differentially expressed ones.