目的 用统计学和生物信息学找到胃癌早期诊断的生物标志，建立诊断模型。方法 回顾性分析解放军总医院2009年至2010年消化科的住院患者179例，根据胃镜检查结果分为胃癌组（139例）和不典型增生组（40例）。用统计学的多参数法分析胃癌和不典型增生患者的血清学参数。用SPSS18.0统计学软件分析两组血清肿瘤标志物、细胞因子以及常规生化指标的差异，建立分类模型。结果 在所有的血清学指标中,两组有20项有差异（P＜0.05）,其中12项具有显著差异（P<0.01）。单一指标的诊断准确率均较低, IL-6和CEA的诊断水平最高，曲线下面积（AUC）分别为0.824和0.721。Logistic回归分析显示，CEA、CA72-4、IL-6、IMA、SOD和ApoA1在胃癌的鉴别诊断中有重要的统计学意义，分类预测准确度达85.3%。结论 用统计学软件对大量数据进行挖掘，为胃癌的早期诊断提供了有前景的研究方向。CEA、CA72-4、IL-6、IMA、SOD和ApoA1在胃癌和不典型增生的鉴别诊断中有着重要的意义。

Objective To set up a diagnostic model of gastric cancer by using biomarkers for differential diagnosis of early gastric cancer through statistics and bioinformatics technologies. Methods We used retrospective cohort study to screening 179 cases of gastric associated diseases.139 cases of gastric cancer and 40 cases of atypical hyperplasia were separated by the diagnosis of gastroscopy. The serological parameter differences were analyzed by statistical multi-parameter analysis between gastric cancer and atypical hyperplasia. SPSS18.0 software was adopted for analysis of tumor markers cytokines and biochemical indicators in two groups to set up a diagnostic model. Results There were significant difference in 20 parameters between gastric cancer and atypical hyperplasia (P < 0. 05). The diagnosis accuracy of single marker was low. The diagnosis horizontal of the IL-6 and CEA were highest. The AUC was 0.824 and 0.721, respectively .The results of logistic regression analysis showed CEA, CA72-4, IL-6, IMA, SOD and ApoA1 had significant differences in the differential diagnosis for gastric cancer, and the prediction accuracy of two groups was 85.3%. Conclusion This study provides a promising strategy for differential diagnosis of gastric cancer with statistics software. The parameters of CEA，CA72-4，IMA，SOD and ApoA1 may have important roles in the predicted classification of gastric cancer and atypical hyperplasia with linear pattern.