Research On Protein Spots Segmentation For Two-dimensional Gel Electrophoresis Images

Protein is the basis of biological life activities,its expression level and function has important research value.And the rapid development of biological system research provides a new perspective.The effective separation of proteins is a prerequisite for the analysis of their expression patterns.It is operated by a two-dimensional gel electrophoresis platform,which depends on the difference of the isoelectric point on the horizontal dimension and the molecular weight on the vertical dimension.The final output is a digital gray scale image,which containing tens of thousands of protein spots with various shape and size.In the image each point represents a protein in the form of different gray levels.And the corresponding analysis procedure involves the following steps: protein segmentation and detection,protein expression of the quantification,matching and other processing techniques.The segmentation of protein spots occupies a very important position in the gel image analysis stage and inaccurate segmentation will seriously affect the analysis results of change of protein expression.This thesis mainly studied the segmentation of the gel electrophoresis image.The main work and achievements of our thesis are as follows:(1)Different filtering techniques of the 2DE were studied,including airspace filtering,improved NL-means algorithm and guided image filtering.The related principles were summarized and the simulation of the images was carried out.Meanwhile,combined with the cross-sectional view of the processed image,the effect of different filtering algorithms was analyzed and the results were compared as the basis of the selection filter algorithm.(2)The traditional image segmentation algorithm(based on threshold,watershed and level set algorithm)and fuzzy clustering algorithm were applied to the gel images.The real gel images were used for simulation and the experimental results were compared to analyze the algorithm effect,which provides the basis of the research of the algorithm.(3)Considering that the protein spots on the gel image are not evenly distributed,the contrast between the gray scale of the border and the background is not particularly obvious.So the kernel fuzzy clustering algorithm was introduced into the segmentation of the gel image and an improved kernel fuzzy clustering gel image segmentation algorithm was proposed.Firstly,the guided image filtering was combined with the morphological method.On the one hand,it was used to reduce the noise interference.On the other hand,it was used to improve the contrast between the protein spot and the background.Then a weight vector was introduced in the kernel function,meanwhile the sample variance was used to calculate the kernel parameter and so that the nuclear parameter can have a certain degree of self-adaptability.Finally,the improved kernel function was introduced into the fuzzy clustering algorithm to achieve the final gel image segmentation.In this process,the experiment was carried out using simulated and real gel images,and the results with other algorithms were also analyzed.It is verified that the proposed algorithm is better than the other algorithms.To some extent,it can be divided into more weak protein spots and the accuracy of the gel image segmentation can be therefore improved.