Bootstrap testing of 2D electrophoresis gels across groups

Abstract

Statistical procedures are often used to pre-screen spots that may be differentially expressed across groups of 2-dimensional electrophoresis (2DE) gels. Commonly used inference procedures are univariate and have poor detection capabilities due to their reliance on unrealistic assumptions and result in a large number of false positives, thereby reducing the efficiency and costing effectiveness of the statistical pre-screening procedures. We develop a data-driven statistical approach that provides accurate identification of statistically significant differences in protein expression profiles across groups. Our methodology relies on a spatial bootstrap technique, and we develop a novel testing procedure using modified Kolmogorov-Smirnov type statistic that is appropriate for spatial dependence in data such as in gel images. Simulation study based on synthetic gels shows that the testing procedure has high efficiency in detecting group differences. The methodology is illustrated via a set of real 2DE gels.