As a microarray experiment, the 2-DE is due to experimental errors that are a source of bias. For this reason, it is necessary to apply multiple preprocessing steps before the proper data analysis. Most of the methods for quality control and preprocessing of microarray experiment can be applied also on 2-DE data. Like in microarray data, 2-DE data does not follow the normal distribution. The log transformation commonly used in microarray experiment is of use here too, at the same time helping to stabilize the variance.
Another important preprocessing step is the calibration of all expression values for all samples, to ensure, that noticed differences are caused just by biological variation. During calibration we remove both spatial and dye effect.
Next part of preprocessing 2-DE data is substitution of missing values. Usually there are large numbers of missing values and some statistical methods need complete dataset. Such missings can be substitute by averaging the expression of the same protein from the rest gels. Or there are more elaborated algorithms like methods based on singular value decompositions or K-nearest neighbours, which uses k most similar expression pattern in comparison with sample with missing value and then it calculates the missing value as weighted mean across the k nearest samples