Research on Protein Spots Matching in2D Gel Electrophoresis Images
|School||Nanchang University of Aeronautics and|
|Course||Detection Technology and Automation|
|Keywords||two-dimensional gel image spot matching tagging feature point hierarchical strategy|
The Extraction of variant protein spots between2D gel images is an importantmethod in the proteomic research, and the homologous protein spots matching in2D gelimages is the key technology in extraction of variant protein points. For the problem ofheavy workload and error-prone matching by hand in the spot matching of gel images,how to match mass protein spots automatically, rapidly and effectively in2D gel imagesis the difficulty and the emphasis in the comparative analysis of protein. Basing on theproject of Natural Science Foundation of China and the project of Natural ScienceFoundation of JiangXi Province, this paper studied the protein spots matching methodin2D gel images in depth. The primary research work and achievements are as follows:(1) The base theory about protein spots matching in gel images was systematicallystudied in this paper, and some factors of image matching were elaborated in detail,such as feature of image matching, similarity measures, search space and search strategyin image matching.(2) A protein spots matching algorithm based on manually tagging feature pointswas presented. Firstly, feature points were extracted by manual mode in the proteinspots which were detected in gel images. Then the model of nonlinear transformationwas set up between gel images by the method of least-squares solution for thepolynomial fitting. Finally, according to the geometric projective relationship, theuniform dimensional reference frame of protein spots was built in gel images, and theprotein spots matching was accomplished by the matching algorithm of minimumdistance metric. The effect of spots matching errors in deformation images was reducedusing the method of manually tagging feature points to determine the relationship ofnonlinear transformation in gel images. Experimental results showed that the proposedmatching algorithm was simple and effective, and had the better robustness, highermatching accuracy.(3) An auto-matching algorithm based on hierarchical strategy was presented.Firstly, according to the different image intensity, protein spots were divided intocorresponding gray level in each gel image based on the hierarchical strategy of gray.Then a local normalized correlation method with overlap constraint of gel images wasused to coarse matching in protein spots which were from the same gray level. Finally,to set protein spot pairs in coarse matching as feature points, the precise matching in the rest of protein spots, which were not matched in the process of coarse matching, wereaccomplished by the method of geometric correlation and similarity criterion. Theproposed matching algorithm improved the matching efficiency when the coarsematching was accomplished by the normalized correlation method combined withoverlap constraint of gel images. And the algorithm also had higher matching accuracyin the case of the precise matching which was proceed by the geometric correlation ofthe rest of protein spots based on the result of the coarse matching. Experimental resultsshowed that the algorithm had better matching effect, the matching error was less than4%, and better stability.