The Clinical Applied Research Of Serum Protein Fingerprint In Thyroid Carcinoma

Thyroid carcinoma is a common malignant tumor in head and neck and the mostcommon endocrine malignancy, which represents 1% of all malignant diseases but91.5% of all endocrine malignancies. Recent statistic shows that thyroid carcinoma'sincidence is on the rise. More than 90% of primary thyroid carcinomas aredifferentiated papillary or follic, ular types, 5 years survival rate of which would be ashigh as 75% when patients could receive early operation therapy. Therefore, earlydetection and diagnosis is very important to prognosis of thyroid carcinoma. However,effective screening of thyroid carcinoma and preoperatively distinction betweenbenign and malignant lesions nowadays remain a difficult task by using imaginetechnique, fine-needle aspiration. So it is highly demanded to develop more accurateinitial diagnostic tests and find new biomarkers for thyroid carcinoma detection anddiagnosis.Advance in proteomics study present new horizon and novel techniques fordiagnosis of carcinoma and detection of biomarker. Because of the multifactor natureof the thyroid carcinoma, it is very likely that a combination of several markers willbe necessary to effectively detect and diagnose thyroid carcinoma. Proteomicmethods detect the functioning unit of expressed genes, through biochemical analysisof cellular proteins, to provide a protein fingerprint. The proteomic reflects both the intrinsic genetic programmer of the cell and the impact of its immediate environmentand is therefore valuable in biomarker discovery. Distinct changes that occur at theprotein level during the transformation of a normal cell into a ceroplastic cell includealtered expression,differential protein modification, changes in specific activity, andinappropriatelicalization,all of which may affect cellular function. Surface enhancedlaser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF-MS)andproteinchip technology, a novel proteomic approach,was developed by CiphergenBiosystem,Inc in 2002.This system is a sensitive,high throughput, and rapid methodfor analyzing complex mixture of protein and peptide. SELDI-TOF-MS coupled withbioinformatics approach, has successfully used to analyze the complex serum proteinsto explore the cancer-specific fingerprints or patterns.We report the use of SELDI-TOF-MS with Weak cation exchange (WCX2)proteinchip arrays analyze 108 serum samples to find new biomarkers and to establishserum protein fingerprint models for early detection and diagnosis of thyroidcarcinoma.Material and methodA total of 108 serum samples from patients with thyroid carcinoma, benignthyroid nodes and healthy individuals were obtained from the First Affiliated Hospitalof Zhengzhou University. All thyroid carcinoma consisting of 40 serum samples wereconfirmed pathologically with the median age of 41 years (range, years).Both controlgroups include 36 serum samples from patients with benign thyroid nodes which wereconfirmed pathologically and 32 serum samples from healthy individuls. Healthyindividuals were age and sex matched with thyroid carcinoma group. All sampleswere obtained in early morning and stored at -80℃until use.Samples were randomly divided into training set (n=87, 32 thyroid carcinoma,30 benign thyroid nodes,25 healthy individuals) and test set (n=21) in test. Allsamples were detected by SELDI-TOF-MS. Noise of spectra was filtrated and peakswere detected with an automatic peak detection pass (signal-to-noise ratio＞2). Peakclusters were completed to cluster the peaks in different samples that had similar masses (defined by a mass window in 0.3% mass error). The power of each peak indiscriminating different group samples Were estimated by the P value of wilcoxonrank sum test. The lower P value shows the higher relative importance value of abilityto accurately distinguish the different groups. The data of spectra were analyzed bythe bioinformatics tools, support vector machine.Results1. Biomarkers selection based on thyroid carcinoma versus healthy group347 qualified peaks were selected after filtrating noise and completing peakclusters by ProteinChip Software 3.1. These peaks were ranked by the P value ofwilcoxon rank sum test. The top ten peaks were selected and randomly grouped toinput SVM. Accuracy of all models was calculated and diagnostic model whichachieved the highest Youden's index was selected as final model to separate twogroups. It include three potential biomarkers with m/z of 6651, 6452,6984Da whichwere highly expressed in healthy individuals but weakly expressed in thyroidcarcinoma. The model combined with three peaks was evaluated by leave-onecross-validation with specificity of 90% and sensitivity of 100%.The model was tested to distinguish thyroid carcinoma from healthy individualsin blind test set of 15 serum samples with specificity of 86% and sensitivity of 100%.2. Biomarkers selection based on thyroid carcinoma versus benign thyroid nodes347 qualified peaks were selected and ranked by the P value of wilcoxon ranksum test. The top ten peaks were selected and input SVM for further analysis. Fivepeaks were selected as potential biomarkers and model combined with them achievedthe highest Youden's index. Three potential biomarkers with m/z of 3400,3690,2902.6 were highly expressed in thyroid carcinoma but weakly expressed in benignthyroid nodes. Two potential biomarkers with m/z of 3937,2687.8 appeared to beexpressed in a contrary way. The model with five peaks was evaluated by leave-onecross-validation with specificity of 100% and sensitivity of 80%.The model was tested to distinguish thyroid carcinoma from benign thyroidnodes in blind test set of 14 serum samples with specificity of 83% and sensitivity of 88%.3. Biomarkers selection based on gradeⅠ-Ⅱthyroid carcinoma versus gradeⅢ-Ⅳthyroid carcinomaTo identify biomarkers with potential for detection of gradeⅠ-Ⅱthyroidcarcinoma from gradeⅢ-Ⅳones, separability between two groups was achieved byusing discriminant analysis linered combination of 206 peaks. Four peaks with m/z of2285,2182,2068,2196 were selected as potential biomarkers,which were highlyexpressed inⅠ-Ⅱthyroid carcinoma but weakly expressed inⅢ-Ⅳthyroid carcinoma.A total accuracy of 80%, an accuracy ofⅠ-Ⅱthyroid carcinoma was 77%, anaccuracy ofⅢ-Ⅳthyroid carcinoma was 86% were obtained when comparinggradeⅠ-Ⅱthyroid carcinoma versus gradeⅢ-Ⅳones.ConclusionThe serum protein fingerprint models of thyroid carcinoma have a high valuein screening thyroid carcinoma and differentiation between benign and malignantlesions.The combination of SELDI-TOF-MS with bioinformatics tools is a noveleffective method for thyroid cancer detection and diagnosis.