The Clinical Applied Research Of SELDI Protein Fingerprint Models In Breast Cancer

Breast cancer remains the most common malignancy affecting women.Its incidence increases year by year in China.Even the diagnosis and treatment of breast cancer are much improved in these years,but they are not being satisfied yet.The early detection is the most reasonable route to improve the survival rate of breast cancer.Therefore,it is urged to find specific and convenient new biomarkers.Surface enhanced laser desorption/ionization-time of flight-mass spectrometry(SELDI-TOF-MS) is one of the developed sophisticated technologies,which,based on mass spectrogram and protein chips,is specifically powerful for analyzing the complex biological samples and carved out such a new path to proteomics research on tumor markers.ObjectiveTo screen serum tumor biomarkers and found diagnostic models of breast cancer by SELDI-TOF-MS and bioinformatics tools and to screen novel tumor biomarkers of breast cancer that are fit for diagnosis in early stage,estimation of prognosis before operation and prediction of axillary lymph nodes metastasis.MethodsSerum samples from 66 cases of breast cancer patients,49 cases of patients of benign breast disease and 29 cases healthy women were analyzed by SELDI-TOF on a ProteinChip reader.PBSII-C.Protein profiles were generated using CM10 protein chips.Protein peak clustering and classification analyses were performed utilizing the Biomarker Wizard and Biomarker Pattern software packages,respectively.Serum samples from 34 cases of gradeⅠ-Ⅱbreast cancer patients,31 cases of gradeⅢ-Ⅳbreast cancer patients were analyzed and protein profiles were generated.Protein peak clustering and classification analyses were performed utilizing the Biomarker Wizard and Biomarker Pattern software packages, respectively.Serum samples from breast cancer patients with or without ALN metastasis were analyzed by SELDI-TOF on a ProteinChip reader,PBSII-C.Protein profiles were generated using CM10 protein chips.Protein peak clustering and classification analyses were performed utilizing the Biomarker Wizard and Biomarker Pattem software packages, respectively.Results1.Thirteen discrepant proteins peaks were screened between the breast cancer patients and the healthy women(P＜0.05).M2306.58,M3091.15,M4804.47 and M5476.80 was used to found the diagnostic models.The accuracy,sensitivity and specificity of cross verification of this model were 74.4%,74.2%and 75.9%,respectively.2.Twenty two discrepant proteins peaks were screened between the breast cancer patients and the patients of benign breast diseases(P＜0.05).The best combination was composed by M2939.19,M2952.86,M4147.16 and M4845.3.The model's accuracy, sensitivity and specificity of cross validation were 73.0%,74.2%and 71.4%,respectively.3.Nineteen discrepant proteins peaks were screened between the breast cancer patients and the non-breast cancer(P＜0.05) and the model composed by five protein peeks (M2306.58,M2952.86,M4102.03,M4147.16 and M5333.92) could do the best in the division of the two groups.The diagnostic accuracy sensitivity and specificity were 77.1%,75.8%和78.2%.4.These model all had the similar abilities to diagnose the breast cancer.5.Eleven discrepant proteins peaks were screened between two groups(P＜0.05).The diagnostic model composed with the M/Z of M2042.87,M2459.83,M3881.37,M4804.47, M6683.24 and M6706.06 could do the best in the division of the two groups.The accuracy of cross verification of this model was 80.0%,the gradeⅠ-Ⅱbreast cancer accuracy was 82.4%and the gradeⅢ-Ⅳbreast cancer accuracy was 77.4%.6.The data of breast cancer patients with or without metastasis in ALN were compared and eleven discrepant proteins peaks were found(P＜0.05).The model composed by three protein peeks(M2164.16,M3269.90 and M3272.31) could do the best in the division of the two groups.The diagnostic accuracy was 81.8%.7.The data of breast cancer patients with≤3 metastatic node versus＞3 metastatic nodes were compared and thirteen discrepant proteins peaks were found(P＜0.05).The model composed by five protein peeks(M2414.20,M2727.38,M2772.35,M2949.77 and M3155.8) could do the best in the division of the two groups.The diagnostic accuracy was 86.1%.Conclusions1.The model composed by four protein peeks(M2306.58,M3091.15,M4804.47 and M5476.80) could do the best in separation of the breast cancer patients and the healthy women with accuracy of 74.4%.This model had the similar abilities to diagnose the breast cancer.2.The best combination was composed by M2939.19,M2952.86,M4147.16 and M4845.3 to separate the breast cancer patients from the patients of benign breast diseases (P＜0.05).The model's accuracy was 73.0%and had the similar abilities to diagnose the breast cancer.3.The model composed by five protein peeks(M2306.58,M2952.86,M4102.03, M4147.16 and M5333.92) could do the best in the division of the breast cancer patients and the non-breast cancer.The model's accuracy was 77.1%and had the similar abilities to diagnose the breast cancer.4.The diagnostic model composed with the M/Z of M2042.87,M2459.83,M3881.37, M4804.47,M6683.24 and M6706.06 could do the best in the division of gradeⅠ-Ⅱbreast and gradeⅢ-Ⅳones.The accuracy of cross verification of this model was 80.0%. 5.The model composed by three protein peeks(M2164.16,M3269.90 and M3272.31) could do the best in the division of breast cancer patients with or without metastasis in ALN.6.The model composed by five protein peeks(M2414.20,M2727.38,M2772.35, M2949.77 and M3155.8) could do the best in the division of breast cancer patients with≤3 metastatic node versus＞3 metastatic nodes.The diagnostic accuracy was 86.1%.