基于血清蛋白组表达谱分析模型的癌症分期预测研究

Cancer presents high mortality and morbidity globally, largely due to its complex and heterogeneous nature as well as the lack of effective biomarkers. Thus, identification of clinically relevant biomarkers from serum proteomic profiling for detecting cancer type and stage is a fundamental work in bioinformatics..In this project, our researches will focus on serum proteomic profiling data of human. Using regularization models and algorithms, we standardize human serum proteomic profiling data. And then the "spectrum" model of serum proteomic spectra expression profile data will be suggested to character quantitatively the information of these data. Some methods will be introduced to identify biomarker protein, which can distinguish early-stage and late-stage cancer patients from disease-free individuals..Furthermore, applying the fuzzy theory and matrix factorization, we will suggest some mathematical descriptions corresponding to information related to tumor diagnosis. Based on the information mining, phylogeny analysis, and statistical method, the information of serum proteomic profiling data, related to cancer early and late stages, of tumor biomarker protein expression level are extracted. We will construct the comparison and cluster models of "spectrum graph" to classify normal human’s, early-stage, and late-stage cancer patients’ serum protein expression profile data. .In addition, we will design the multivariate statistical algorithm and the mixed model to improve the existing tumor biomarker prediction model. Based on the support vector machine and other machine learning model, the prediction algorithm of biomarker protein related to tumor stages, types, and grades will be developed. .This project has important scientific significance for diagnosis and theoretical researches of cancer.

由于恶性肿瘤的复杂性而且缺乏有效的生物标记，从血清蛋白表达谱数据获取与肿瘤分期、分类相关的生物标志是一个重要的基础性研究。.本项目从人的血清蛋白组表达谱数据出发，利用正则化模型和算法对血清蛋白组表达谱数据的信息度量标准化，研究人的血清蛋白表达谱数据的“谱图”模型与定量化描述，寻找血清蛋白组中与肿瘤早期、癌变期相关的标志物，改进已有肿瘤标志物预测模型；利用模糊理论和矩阵分解构造血清蛋白表达谱中与肿瘤诊断相关信息对应的数学描述，建立基于血清蛋白质组表达谱信息的系统发育分析及其统计校验方法，提取肿瘤发生-发展与标志物表达水平的关系信息，构造正常人、早期、癌变期血清蛋白组表达谱数据的“谱图”比较与分类模型；构造多元统计算法以及混合模型，基于支持向量机、随机森林等机器学习方法设计恶性肿瘤的分期、分类、分级预测以及与其相关肿瘤标志物预测算法。.此项目对恶性肿瘤的诊断和理论研究都有重要的科学意义。