OBJECTIVE: To establish a diagnostic model for systemic lupus erythematosus (SLE) using proteiomic fingerprint techology. METHODS: Blood samples were collected from 64 cases of SLE, 30 cases of rheumatoid arthritis (RA), 30 cases of Sjogren's syndrome (SS), 25 cases of systemic sclerosis (SSc), as well as 83 healthy controls. Proteomic spectra of these 232 serum samples were generated by proteomic fingerprint technology. Diagnostic model was established by a machine learning algorithm called decision boosting. The sensitivity and specificity of the diagnostic model was validated with a blinded testing set. RESULTS: Sixty differential protein peaks (P<0.05) between SLE and control subjects were indicated, 28 of them were up regulated and 32 were down regulated in SLE patients. The algorithm identified a cluster pattern segregating SLE from non-SLE with sensitivity of 91% and specificity of 92%. The discriminatory diagnostic pattern correctly identified SLE. A sensitivity of 78% and specificity of 96% for the blinded test were obtained when comparing SLE vs non-SLE. CONCLUSION: This diagnostic model using proteiomic fingerprint techology appears to be a promising tools with high sensitivity and specificity in diagnosis of SLE.
OBJECTIVE: To establish a diagnostic model for systemic lupus erythematosus (SLE) using proteiomic fingerprint techology. METHODS: Blood samples were collected from 64 cases of SLE, 30 cases of rheumatoid arthritis (RA), 30 cases of Sjogren's syndrome (SS), 25 cases of systemic sclerosis (SSc), as well as 83 healthy controls. Proteomic spectra of these 232 serum samples were generated by proteomic fingerprint technology. Diagnostic model was established by a machine learning algorithm called decision boosting. The sensitivity and specificity of the diagnostic model was validated with a blinded testing set. RESULTS: Sixty differential protein peaks (P<0.05) between SLE and control subjects were indicated, 28 of them were up regulated and 32 were down regulated in SLEpatients. The algorithm identified a cluster pattern segregating SLE from non-SLE with sensitivity of 91% and specificity of 92%. The discriminatory diagnostic pattern correctly identified SLE. A sensitivity of 78% and specificity of 96% for the blinded test were obtained when comparing SLE vs non-SLE. CONCLUSION: This diagnostic model using proteiomic fingerprint techology appears to be a promising tools with high sensitivity and specificity in diagnosis of SLE.