The original expression values often need to be transformed and/or normalized in order to ensure that samples are comparable and assumptions on the data for analysis are met [Allison et al., 2006]. These are essential requirements for carrying out a meaningful analysis. The raw expression values often exhibit a strong dependency of the variance on the mean, and it may be preferable to remove this by log-transforming the data. Furthermore, the sets of expression values in the different samples in an experiment may exhibit systematic differences that are likely due to differences in sample preparation and array processing, rather being the result of the underlying biology. These noise effects should be removed before statistical analysis is carried out.
Transformation and normalization
When you perform transformation and normalization, the original expression values will be kept, and the new values will be added. If you select an experiment (), the new values will be added to the experiment (not the original samples). And likewise if you select a sample ( () or ()) - in this case the new values will be added to the sample (the original values are still kept on the sample).