Computer analysis of two‐dimensional gels: Automatic matching

Abstract

Abstract A program which automatically mathches the spot patterns resolved on two‐dimensional gel electrophoretograms is described. The program does not require handmatched, landmark matches for initial alignment of the spot patterns. The matching algorithm is based on a hierarchical nearest neighbor analysis. Starting with the most intense spots in the two films, the program determines if spots are equivalent by attempting to match a list of all nearby spots (a “cluster”) from each film. “Clusters” are considered to be aligned if there is a low probability that the pairing of spots between them is due to a random process. The match is further tested by requiring that secondary clusters ( i. e. those clusters that surround the spots matched between the central clusters) can also be aligned. Pairings found by cluster matching are checked for consistency and matches that are out of alignment with the majority of other mathces are eliminated. Finally, the program tries to match the remaining spots by mapping the coordinates of an unmatched spot in one gel into the coordinate system of the other using the matched spots in the cluster as landmarks. The formulas used for the transformation allow for localized rotation and stretching of the coordinate systems. The accuracy and robustness of the program are discussed. If high quality gels are used, the program will find, on average, 95% of the spots that are mathcable between the gels, and requires about 4.3 seconds per mathch. Most of the errors are found at the extreme edges of the gels.