Quantitative proteomics data quality and statistical power
for discerning treatment differences
To demonstrate the power of using DIA assay libraries for accurate
protein quantitation on a proteome-wide scale we analyzed the DIA data
for six replicates of kidneys from fish acclimated to brackish water
(BW) and freshwater (FW), each. A volcano plot was generated with
Skyline 20.0 as previously described (Li, Levitan, Gomez-Jimenez, &
Kültz, 2018) to visualize statistically significant proteins associated
with acclimation of fish to BW (Figure 2A). QC charts of the DIA data
illustrate that the majority of transitions had a mass error of less
than 10 ppm (Figure 2B) and there was a perfect correlation of measured
retention time (RT) with the predicted RT based on intrinsic RT (iRT)
standards (Escher et al., 2012) with no outliers (Figure 2C). In
addition to an adjusted p value < 0.05, which took into
account multiple testing (Benjamini & Hochberg, 1995), a fold-change
(FC) threshold was enforced in the volcano plots. The FC threshold was
1.85 as calculated by MSstats (Choi et al., 2014) using a statistical
power analysis based on the coefficient of variance (CV) of all
experimental DIA data (Figure 2D). Furthermore, the vast majority of
transition peaks for all 12 samples in this dataset had mProphet (Reiter
et al., 2011) peak scores of q < 0.01, which was the peak
quality threshold for inclusion in MSstats quantitative DIA data
analysis (Figure 2e). In this example, 21 kidney proteins were
up-regulated and 21 down-regulated during acclimation of fish from FW to
BW (Figure 2A).