凝胶图像蛋白点分割算法研究

发布时间：2018-04-24 10:19

本文选题：双向凝胶电泳 + 核模糊聚类算法　；参考：《山东师范大学》2017年硕士论文

【摘要】：蛋白质是生命活动的基础,对其功能及表达水平等方面的研究具有重要的价值,而快速发展的蛋白质组学成为生物系统研究的一个新方向。将蛋白质有效的分离出来是对其表达模式分析的前提,它是利用双向凝胶电泳技术平台来操作运行,依靠蛋白质在凝胶中水平维度上等电点的不同和竖直维度上分子量之间的差异特征实现,最终输出的是一幅数字灰度图像。在这幅图像中含有成千上万个形状和大小各不相同的蛋白质点,其中每个点代表了一个蛋白质,并且这些点都是以不同灰度级的形式表现出来。凝胶图像的分析过程主要包括:蛋白点的分割与检测,蛋白点的表达量化、匹配等过程,其中,蛋白点分割占据着非常重要的地位,不准确的分割会严重影响后续对蛋白质表达变化的分析,因此,本论文重点研究了对凝胶图像的分割算法。本论文的主要工作如下:(1)研究了不同的滤波技术主要包括空域滤波、改进的NL-means算法和引导滤波器,概述了其相关原理并分别对凝胶图像进行实验仿真。同时,与处理后的凝胶图像剖面图相结合,进一步对上述滤波算法处理的效果做出对比分析,并作为选择滤波算法的依据。(2)将传统的图像分割算法(基于阈值、分水岭和水平集算法)和基于模糊聚类的算法应用于凝胶图像的分割中,并利用真实的凝胶图像对上述几种算法进行实验,通过对比观察实验结果来分析各种算法的分割效果,为下一步的研究提供基础。(3)凝胶图像上蛋白点并不是均匀分布且部分点的边界灰度与背景之间的对比并不是特别明显,由此将核模糊聚类算法引入到凝胶图像的分割中并对其进行改进。首先将引导滤波器与形态学方法结合起来,一方面用于对图像进行降噪处理以降低噪声的干扰,另一方面用于提高图像中蛋白点与背景之间的对比差异;然后在核函数中引入一个权值向量,与此同时利用样本方差来合理的计算核参数值,使核参数具有一定的自适应度;最后将改进后的核函数引入到模糊聚类算法里面,从而最终完成凝胶图像的聚类分割。在此过程中,分别利用模拟和真实的凝胶图像进行实验,并与其他算法做对比分析,验证了本论文算法的分割效果明显优于对比算法,在一定程度上能够分割出更多的微弱蛋白点,提高了凝胶图像分割的精度和准确性。
[Abstract]:Protein is the basis of life activity, so it is of great value to study its function and expression level. The rapid development of proteomics has become a new direction of biological system research. The efficient separation of proteins is a prerequisite for the analysis of their expression patterns, which are operated on a two-dimensional gel electrophoresis platform. Depending on the difference of isoelectric point in the horizontal dimension and the molecular weight in the vertical dimension of the gel, the final output is a digital gray image. The image contains thousands of protein spots of different shapes and sizes, each of which represents a protein, and these dots are represented by different grayscale levels. The analysis process of gel image mainly includes: protein point segmentation and detection, protein point expression quantification, matching and so on. Among them, protein point segmentation occupies a very important position. Inaccurate segmentation will seriously affect the subsequent analysis of protein expression changes. Therefore, this paper focuses on the segmentation algorithm of gel image. The main work of this thesis is as follows: (1) the main filtering techniques include spatial filtering, improved NL-means algorithm and bootstrapping filter. At the same time, combining with the processed gel image profile, the effect of the above filtering algorithm is compared and analyzed, and the traditional image segmentation algorithm (based on the threshold value) is used as the basis for selecting the filtering algorithm. Watershed and level set algorithm) and fuzzy clustering algorithm are applied to the segmentation of gel image, and the real gel image is used to experiment the above algorithms. The segmentation effect of these algorithms is analyzed by comparing and observing the experimental results. To provide the basis for further research.) the protein spots on the gel image are not uniformly distributed and the contrast between the boundary grayscale of some points and the background is not particularly obvious. Therefore, kernel fuzzy clustering algorithm is introduced into gel image segmentation and improved. Firstly, the guided filter is combined with the morphological method, which is used to reduce the noise interference of the image, on the one hand, to improve the contrast between the protein points and the background in the image. Then a weight vector is introduced into the kernel function, and at the same time, the kernel parameter value is reasonably calculated by using the sample variance, so that the kernel parameter has a certain degree of adaptability. Finally, the improved kernel function is introduced into the fuzzy clustering algorithm. Finally, the clustering segmentation of gel image is completed. In this process, simulation and real gel images are used to carry out experiments, and compared with other algorithms, it is verified that the segmentation effect of this algorithm is obviously better than that of contrast algorithm. To some extent, more weak protein points can be segmented, and the accuracy and accuracy of gel image segmentation are improved.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：Q51;TP391.41

【参考文献】