The discovery of biomarkers unique to multiple myeloma (MM) is of great importance to clinical practice. This study was designed to identify serum tumor marker candidates of MM in the mass range of 700-10000 Da. Serum samples from 48 MM patients and 74 healthy controls were collected and classified into a training dataset (MM/controls: 26/26) and a testing dataset (MM/controls: 22/48). Weak cation exchange magnetic beads, MALDI-TOF MS and analytic software in the CLINPROT system were used to do serum sample pre-fractionation, data acquisition and data analysis. Peak statistics were performed using Welch's t test. Mass spectra from the two model generation cohorts in the training dataset were analyzed by the Supervised Neural Network Algorithm (SNNA) in ClinProTools((TM)) to identify the mass peaks with the highest separation power. The resulting diagnostic model was subsequently validated in the testing dataset. A total of 89 discriminating mass peaks were detected by ClinProTools((TM)) in the range of 700-10000 Da using a signal to noise threshold of 3.0. Of these, 49 peaks had statistical significance (P < 0.0001) and four peaks with the highest separation power were picked up by SNNA to form a diagnostic model. This model achieved high sensitivity (86.36 %) and specificity (87.5 %) in the validation in the testing dataset. Using CLINPROT system and MB-WCX we found four novel biomarker candidates. The diagnostic model built by the four peaks achieved high sensitivity and specificity in validation. CLINPROT system is a powerful and reliable tool for clinical proteomic research.