An approach to completely automatic comparison of two-dimensional electrophoresis gels

The problem addressed is that of determining the similarities and differences appearing in a sequence of two-dimensional polyacrylamide gels where we have no a priori knowledge of the intensity and spatial distribution of the protein spots. It is assumed that the gels are not in precise registration. An attempt is being made to develop a completely automatic program for use in genetic studies, which will compare a sequence of three gels run on samples from father, mother, and child. The program constructs a graph by using as nodes those spot cues that exceed a given intensity threshold. The graphs are then compared to determine an initial subset of spots that are common to all three gels. From this subset of common spots the program then determines whether the remaining graph differences are real or result from quantitative variation causing spots to fall below threshold.