Evaluation of Point-matching Algorithms for Registration of 2D Electrophoresis Gels

Many point-matching algorithms have been developed and applied in a number of image registration problems. Comparison of 2D Electrophoresis (2DE) gels is an important problem in biological research and registration of gels provides a case study in the technical difficulties that beset point-matching: large numbers of points with variable densities, large non-rigid transformations between point sets, paucity of structural information and large numbers of unmatchable points in either set. Here we evaluate a number of algorithmic approaches for accuracy and robustness in the face of these difficulties. We separate three different components of the point-matching problem: correspondence assignment, distance metrics and image transformation. In each case we evaluate candidate algorithms using synthesised point sets, applying controlled image distortion and random outlier proportions. A version of Iterated Closest Point (ICP) using a non-Euclidean distance metric and robust estimation of transform parameters turns in a very good performance in the face of well-regarded opposition.