2D gel image processing and analysis for proteomics

Abstract

In proteomics, two-dimensional gel electrophoresis (2DGE) is the most commonly used technique to separate the complex mixtures of proteins, and image processing and analysis plays an important role in 2DGE. We found that some spots which correspond to proteins might be missed when the watershed algorithm was used to detect the spots. Based on the properties of such spots, we proposed a clustering based method for spots detection. This method regards the pixels in the 2DGE image as cluster data and employs the subtractive clustering technique to detect the cluster centers, which can be used as the internal markers in watershed segmentation. With the new markers, more potential protein spots can be detected. To model the saturated regions of protein spots, we proposed a new method which uses axis-parallel ellipses as covering models on the saturated region. Particle Swarm Optimization (PSO) and subtractive clustering are used to construct the model. By using the clustering method, we could obtain good estimation of the positions of the potential merging spots in a saturated spot. PSO is used to minimize the covering error and find the best covering ellipses for those protein spots. The Combination of all the detected ellipses makes up the model of the saturated spot region. Our simulations will show satisfying results of the new covering model.