We present a novel method of statistical analysis for the comparison of ele
ctrophoretic data. The method is based on the squared Euclidian distance of
normalized signal data vectors of electrophoretic lanes. The differences i
n the electrophoretic patterns are evaluated by a statistical test based on
Hubert's statistics which measures the significance of the signal grouping
. We demonstrate the validity and applicability of the method in a large da
ta set derived from automated fluorescent mRNA differential display analysi
s of the expression of acute-phase proteins during experimental Escherichia
coli infection in mice. The current testing method is capable of finding t
heoretically similar natural groupings to be similar in a statistically sig
nificant way whereas theoretically dissimilar or random groupings can be re
cognized to be artifactual. We also show how the calculated pairwise signal
distances can be utilized in methodological problem solving. These analyti
cal methods can be applied to the study of other related problems of simila
rity analysis of electrophoretic patterns, and also provide useful tools fo
r the development of automated recognition of differentially expressed mRNA
s.