矩阵填充算法及其在流感病毒抗原性研究中的运用

发布时间：2018-04-25 15:00

本文选题：低秩矩阵填充 + 抗原距离　；参考：《浙江理工大学》2017年硕士论文

【摘要】：流行性感冒是流感病毒引起的急性呼吸道感染疾病,它传染性强、传播速度快,流感的预防一直受到世界各国的高度重视。目前,血凝素抑制试验(HI)是获得流感病毒抗原性特征的常规手段之一,但是,由于受到实验条件的干扰和限制,导致有些数据不准或有些数据无法得到,为了获取可靠的、完整的血凝素抑制试验数据,本文基于低秩矩阵填充算法,提出了一种新的融合了生物信息的低秩矩阵填充算法,带边信息的生物矩阵填充算法(BMCSI)。本文应用这个模型到1968-2003年的H3N2流感病毒数据,通过将HA蛋白序列的相似性信息整合进矩阵填充模型,用来预测流感病毒的抗原性。首先对血凝素抑制试验数据中已有的低反应数据进行了重估计,对缺失的数据进行了恢复。通过10倍交叉验证法评估该算法,与已有的计算方法比较,该算法的均方根误差减少了37%。利用这些重估后的数据构造抗原图和遗传图,结果显示H3N2流感病毒的抗原进化图呈“S”型,遗传进化图呈半圆型。进一步地,本研究通过分析H3N2流感病毒的遗传进化与抗原进化之间的关系,发现遗传距离与抗原距离的斯皮尔曼相关系数是0.83,此结果说明遗传进化与抗原进化之间虽然有局部差异但全局具有高度的一致性。本文根据上述算法BMCSI提出了一种选择疫苗株的方法,并对历史疫苗株的选择进行了评估。另外,我们还将这个模型运用到CCLE中491种癌细胞系与24种抗癌药物反应的敏感性数据上,利用矩阵的行和列之间的相关性以及基因表达特征来预测抗癌药物敏感值,然后通过抗癌药物敏感性数据寻找与药物敏感性关联的基因。
[Abstract]:Influenza is an acute respiratory infection caused by influenza virus. At present, hemagglutinin inhibition test (HIH) is one of the conventional methods to obtain the antigenicity characteristics of influenza virus. However, due to the interference and limitation of experimental conditions, some data are inaccurate or some data cannot be obtained, in order to obtain reliable, Based on the low rank matrix filling algorithm, a new low rank matrix filling algorithm with edge information is proposed. In this paper, the H3N2 influenza virus data from 1968-2003 were used to predict the antigenicity of influenza viruses by integrating the similarity information of HA protein sequences into the matrix filling model. Firstly, the low response data of hemagglutinin inhibition test were reestimated and the missing data were recovered. Compared with the existing methods, the RMS error of this algorithm is reduced by 37 times. Using these reestimated data to construct the antigen map and genetic map, the results show that the H3N2 influenza virus antigen evolution map is "S" type, genetic evolution map is semi-circular type. Further, this study analyzed the relationship between genetic evolution and antigen evolution of H3N2 influenza virus. It is found that the Spelman correlation coefficient between genetic distance and antigen distance is 0.83, which indicates that there is a high consistency between genetic evolution and antigen evolution, although there is a local difference between genetic evolution and antigen evolution. Based on the above algorithm, BMCSI, a method for selecting vaccine strains is proposed, and the selection of historical vaccine strains is evaluated. In addition, we applied the model to the sensitivity data of 491 cancer cell lines and 24 anticancer drugs in CCLE, and predicted the sensitivity of anticancer drugs by using the correlation between the rows and columns of the matrix and the characteristics of gene expression. The genes associated with drug sensitivity were then identified by using anti-cancer drug sensitivity data.
【学位授予单位】：浙江理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O151.21;R373

【参考文献】