Serum Protein Profiling with Mass Spectrometry for the Diagnosis of Myelodysplastic Syndromes

• Published in: Blood 2004-11, Vol.104 (11), p.2362-2362
• Main Authors: Aivado, Manuel, Spentzos, Dimitrios, Germing, Ulrich, Alterovitz, Gil, Meng, Xiao-Ying, Grall, Franck, Giagounidis, Aristoteles A.N., Klement, Giannoula, Steidl, Ulrich, Otu, Hasan H., Iking-Konert, Christof, Czibere, Akos, Prall, Wolf C., Shayne, Michelle, Ramoni, Marco F., Gattermann, Norbert, Mitsiades, Constantine S., Haas, Rainer, Fung, Eric T., Libermann, Towia A.
• Format: Article
• Language: English
• ISSN: 0006-4971; 1528-0020
• DOI: 10.1182/blood.V104.11.2362.2362

Surface enhanced laser desorption/ionization-time of flight mass spectrometry (SELDI-TOF MS) has facilitated disease-specific serum protein profiles, which may become instrumental for diseases that are difficult to diagnose. The diagnosis of MDS can be very difficult since bone marrow dysplasia and peripheral cytopenia, the hallmarks of MDS, can be observed due to many reasons other than MDS. Moreover, cytomorphological evaluation of dysplasia requires extensive experience and is thus dependent on the examiner. Between 2–16% of the patients have a fibrotic bone marrow, which can impede appropriate bone marrow aspiration and smear. Finally, while cytogenetic examination can strongly support the diagnosis of MDS, only 40–50% of the patients have an abnormal karyotype. In the present study, we employed a SELDI-TOF MS-based procedure termed Pattern Track™ in order to analyze a total of 218 serum samples. We generated serum protein profiles from a first sample set comprising 74 patients with MDS, 39 control patients with cytopenia for reasons other than MDS, and 24 healthy persons. We fractionated their serum by means of anion exchange chromatography and applied the resulting serum fractions to weak cationic exchange as well as to reversed phase chromatography ProteinChip™ arrays. We randomly split this dataset into a learning (n=72) and a first independent validation set (n=41). Then, we used a k-nearest-neighbor algorithm to build a class predicting profile that consisted of 81 protein peaks. That profile was tested by leave-one-out cross validation and predicted the diagnosis MDS with an accuracy of 81.9% in the learning set (Fisher's test, P=0.0000003). Then, we tested the profile on our first independent validation set and obtained a similar accuracy of 80.5% (P=0.0002). Its diagnostic performance and long-term reproducibility were confirmed by successfully applying it to a prospectively collected second independent validation set consisting of 81 new samples (P=0.0000006). Eventually, following serial chromatography, 1D gel electrophoresis, and tryptic peptide fingerprinting, we discovered the identity of 2 members of the profile. We conclude that our predicting serum protein profile represents a novel, non-invasive aid in distinguishing patients with MDS from patients with cytopenia for reasons other than MDS.