Construction of protein profile classification model and screening of proteomic signature of acute leukemia.

The French-American-British (FAB) and WHO classifications provide important guidelines for the diagnosis, treatment, and prognostic prediction of acute leukemia, but are incapable of accurately differentiating all subtypes, and not well correlated with the clinical outcomes. In this study, we performed the protein profiling of the bone marrow mononuclear cells from the patients with acute leukemia and the health volunteers (control) by surface enhanced laser desorption/ionization-time of flight mass spectrometry (SELDI_TOF_MS). The patients with acute leukemia were analyzed as unitary by the profiling that were grouped into acute promyelocytic leukemia (APL), acute myeloid leukemia-granulocytic (AML-Gran), acute myeloid leukemia-monocytic (AML-Mon) acute lymphocytic leukemia (ALL), and control. Based on 109 proteomic signatures, the classification models of acute leukemia were constructed to screen the predictors by the improvement of the proteomic signatures and to detect their expression characteristics. According to the improvement and the expression characteristics of the predictors, the proteomic signatures (M3829, M1593, M2121, M2536, M1016) characterized successively each group (CON, APL, AML-Gra, AML-Mon, ALL) were screened as target molecules for identification. Meanwhile, the proteomic-based class of determinant samples could be made by the classification models. The credibility of the proteomic-based classification passed the evaluation of Biomarker Patterns Software 5.0 (BPS 5.0) scoring and validated application in clinical practice. The results suggested that the proteomic signatures characterized by different blasts were potential for developing new treatment and monitoring approaches of leukemia blasts. Moreover, the classification model was potential in serving as new diagnose approach of leukemia.