A high confidence, manually validated human blood plasma protein reference set
1 Department of Biochemistry and Molecular Biology, Bioinformatics, University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark
2 University of Bergen, P.O. Box 7800, 5020 Bergen, Norway
3 Proteomics and signal transduction, Max-Planck Institute for Biochemistry, Am Klopferspitz 18, 81152 Martinsried, Germany
BMC Medical Genomics 2008, 1:41 doi:10.1186/1755-8794-1-41Published: 15 September 2008
The immense diagnostic potential of human plasma has prompted great interest and effort in cataloging its contents, exemplified by the Human Proteome Organization (HUPO) Plasma Proteome Project (PPP) pilot project. Due to challenges in obtaining a reliable blood plasma protein list, HUPO later re-analysed their own original dataset with a more stringent statistical treatment that resulted in a much reduced list of high confidence (at least 95%) proteins compared with their original findings. In order to facilitate the discovery of novel biomarkers in the future and to realize the full diagnostic potential of blood plasma, we feel that there is still a need for an ultra-high confidence reference list (at least 99% confidence) of blood plasma proteins.
To address the complexity and dynamic protein concentration range of the plasma proteome, we employed a linear ion-trap-Fourier transform (LTQ-FT) and a linear ion trap-Orbitrap (LTQ-Orbitrap) for mass spectrometry (MS) analysis. Both instruments allow the measurement of peptide masses in the low ppm range. Furthermore, we employed a statistical score that allows database peptide identification searching using the products of two consecutive stages of tandem mass spectrometry (MS3). The combination of MS3 with very high mass accuracy in the parent peptide allows peptide identification with orders of magnitude more confidence than that typically achieved.
Herein we established a high confidence set of 697 blood plasma proteins and achieved a high 'average sequence coverage' of more than 14 peptides per protein and a median of 6 peptides per protein. All proteins annotated as belonging to the immunoglobulin family as well as all hypothetical proteins whose peptides completely matched immunoglobulin sequences were excluded from this protein list. We also compared the results of using two high-end MS instruments as well as the use of various peptide and protein separation approaches. Furthermore, we characterized the plasma proteins using cellular localization information, as well as comparing our list of proteins to data from other sources, including the HUPO PPP dataset.
Superior instrumentation combined with rigorous validation criteria gave rise to a set of 697 plasma proteins in which we have very high confidence, demonstrated by an exceptionally low false peptide identification rate of 0.29%.