【作者】 尹雪莉；

【导师】 李昕；

【作者基本信息】 燕山大学 ， 生物医学工程， 2006， 硕士

【摘要】 癌症是危害人类健康的常见病、多发病,癌症的早期发现对其治疗及愈后判断均具有非常重要的意义。血清蛋白组学模式诊断技术是一种新型的蛋白质组学平台,在此平台上通过高维质谱所获得的蛋白质质谱图可用做疾病的诊断标准。此技术在早期癌症的检测中具有一定的应用前景。目前发现通过蛋白指纹技术所获得的生物标记物,大多数是在特异性癌症微环境中所产生的低分子质量蛋白碎片,通过多种癌症的检测表明,其敏感性和特异性均优于传统的癌症标记物,对某些癌症的敏感性已达100%,特异性也超过95%,因而在癌症早期诊断和早期预警中具有重要临床应用价值。本文主要聚焦于蛋白质质谱数据的卵巢癌诊断。首先,本文介绍了蛋白质组学、蛋白质芯片,重点介绍了表面加强激光解析电离质谱(SELDI-TOF)技术的基本原理、机理、临床应用及存在的问题。其次,进行卵巢癌质谱数据的预处理过程。预处理包括质谱数据的载入,重新取样,基线校正,去噪;用主成分分析进行数据降维以便于进行后面的分类工作。最后,进行卵巢癌数据分类,本文使用的是线性判别分析和K-最近邻域法对降维后的数据进行了分类,并比较了在选用不同训练集和测试集的情况下,得出分类结果的错误率和敏感性及特异性,通过比较得到运用线性判别作为分类器能取得较好的分类效果。更多还原

【Abstract】 Cancer is ordinary and frequent illness that do great harm to human being’s healthiness. Detecting cancer early has great significance to its therapy and judgement from cure. Serum proteomic pattern diagnosis technology is a new proteomic flat.Gaining protein spectrum through high dimension mass spectrum in this flat can be as a criterion to diagnose illness. This technology has great application foreground in detecting cancer early. At present, most of biomarkers gaining from SELDI-TOF-MS are low molecule mass protein pieces produced from tiny environment of special cancer.Through inspecting cancer suggests its sensitivity and specialty are better traditional cancer marker, its sensitivity is nearly 100% and specialty is over 95%. So, it has important clinic application value in inspecting cancer early and warning early, this paper is focused on proteomic mass spectrum ovarian diagnosising.First, proteomics and protein chip are introduced. The technology of SELDI-TOF is stressed, including its basic principle, mechanism, clinic application and existent problem. Second, preprocess ovarian mass spectrum, including loading data, resampling the spectrum, baseline correction, noise reduction. Use LDA to reduce data dimensionality in order to classify in the end.In the end, data is classified. LDA and KNN are used to classify data reduced dimensionality in this paper. Comparing sensitivity and specialty and choosing different training set and testing set, LDA is a better classifier that can be learned from the paper.更多还原