Logarithmic Scale and Fold Changes
A lot of statistical methods assume that data is normal distributed, and gene expression data often tends to be normally distributed when expressed in logarithmic scale than in linear scale. Also, we are often interested in fold changes in expression rather than absolute changes or percentages. Therefore, copy numbers are often converted to logarithmic scale. The Pre-processing menu in the Data editor offers five options to convert into logarithmic scale.
The first three options differ only in the base (2, e, and 10). They give the same relative ratios (quotients between two genes or two samples), although the log2 is most commonly used. Using log10(1+value) makes it possible to convert negative samples (containing zero copies) to logarithmic scale, but on the other hand it introduces an error. A negative sample will be treated as containing 50 % of the number of molecules in a sample containing a single copy, and 33% of the molecules in a sample containing two copies. Hence, the correction is not perfect, but the error is usually negligible.