Analysis of Two-Dimensional Electrophoretic Gels

SUMMARY The paper comprises a study on automatic analysis of two-dimensional (2D) electrophoretic gels by means of image analysis. The data considered are gels showing polypeptides from human embryonal lung fibroblasts digitized to approximately 2,000 x 2,000 picture elements (pixels). The matching between the reference gel and the test gel is done in several steps. First, the disparity between related points in the two gels are estimated by minimizing sums of squared differences, which basically is equivalent to a cross-correlation technique. Based on these disparities, the test gel is transformed ("warped") so that it will match the reference gel as well as possible. Second, the polypeptide spots in the two gels are identified by a series of (morphological) filterings. The methods include noise removal by means of median filtering, grey level erosion, edge detection using "second-derivative filters," and binary erosion. The operations are done at different resolutions in order to detect objects of different sizes. The identification of polypeptides in the test gel may now be assessed by superimposing the spots from the reference gel on the "warped" spots from the test gel. A visual inspection shows that this automated approach may be of value in the further development of 2D electrophoresis programs.