Common problems encountered during proteomics workflows - the analysis of all proteins in a biological sample - are identified in a paper online in Nature Methods this week. Proteomics is most often carried out using mass spectrometry technology, which has gained a reputation of being poorly reproducible. This work could therefore be used as a basis for looking into how to overcome these issues.
John Bergeron and colleagues from several other laboratories created test samples consisting of 20 highly purified proteins present at equal concentrations. They sent these samples to 27 different labs, asking the members of these labs to identify the 20 proteins by the mass spectrometry instrumentation and workflows they routinely use. Initially, only 7 of the 27 labs correctly reported all 20 proteins. However, when the study designers reanalyzed the labs’ data, they found that almost all of the labs had generated sufficient data to identify the proteins but for various reasons were unable to report the correct protein identities. Although some of the labs made simple mistakes in sample handling, most of the problems stemmed from difficulties in applying proteomic database search tools to interpret the data rather than from the mass spectrometry technology itself.In the end, with coaching from the study designers, members of all 27 labs were able to identify all 20 proteins.
Though 20 proteins is, of course, not representative of a complex biological proteome, the study demonstrates that high-quality, reproducible data can be generated by different labs using different mass spectrometry-based proteomics workflows. However, in an accompanying News & Views, Ruedi Aebersold emphasizes that proper training is very important for correctly applying this technology.