A robust permutation test for quantitative SILAC proteomics experiments
Stable Isotope Labeling by Amino Acids in Cell Culture (SILAC) along with other relative quantitation methods in proteomics have become important tools in the analysis of cellular and subcelluar functions. Although numerous experimental applications of SILAC have been developed, there is no consensus on the use of statistical procedures to analyze the resulting experimental data. SILAC experiments output relative abundance ratios for proteins to quantify dierences in cell populations. These ratios have traditionally been analyzed with fold-change methods and hypothesis testing procedures under Gaussian distribution assumptions.
We find that the normality assumption is invalid and can lead to inaccurate quantitation of the significance of differences between cell populations. As a solution, a permutation based hypothesis test as an alternative for assessing significance is introduced. We develop distribution-free permutation testing methods for assessing SILAC experiments. These tests generate p-values which can be easily interpreted and if necessary, the
false discovery rate of these p-values can be easily controlled. To compare the permutation test against competing methodology, we used a set of simulations based upon a theoretical model of SILAC ratio data.
Through the simulation studies, we find that the permutation test is generally superior to the competing hypothesis tests across the range of simulation scenarios. We also find that the permutation test is typically
more powerful and accurate than the competing methods at the five percent level of signicance and averaged over the spectrum of signicance levels. Because of the broad superiority of the permutation test and the ease
of implementation, we propose the use of the permutation test as a standard measure of protein signicance in SILAC experiments.