Evaluation of normalization methods on GeLC-MS/MS label-free spectral counting data to correct for variation during proteomic workflows

Abstract

Normalization of spectral counts (SpCs) in label-free shotgun proteomic approaches is important to achieve reliable relative quantification. Three different SpC normalization methods, total spectral count (TSpC) normalization, normalized spectral abundance factor (NSAF) normalization, and normalization to selected proteins (NSP) were evaluated based on their ability to correct for day-to-day variation between gel-based sample preparation and chromatographic performance. Three spectral counting data sets obtained from the same biological conidia sample of the rice blast fungus Magnaporthe oryzae were analyzed by 1D gel and liquid chromatography-tandem mass spectrometry (GeLC-MS/MS). Equine myoglobin and chicken ovalbumin were spiked into the protein extracts prior to 1D-SDS- PAGE as internal protein standards for NSP. The correlation between SpCs of the same proteins across the different data sets was investigated. We report that TSpC normalization and NSAF normalization yielded almost ideal slopes of unity for normalized SpC versus average normalized SpC plots, while NSP did not afford effective corrections of the unnormalized data. Furthermore, when utilizing TSpC normalization prior to relative protein quantification, t-testing and fold-change revealed the cutoff limits for determining real biological change to be a function of the absolute number of SpCs. For instance, we observed the variance decreased as the number of SpCs increased, which resulted in a higher propensity for detecting statistically significant, yet artificial, change for highly abundant proteins. Thus, we suggest applying higher confidence level and lower fold-change cutoffs for proteins with higher SpCs, rather than using a single criterion for the entire data set. By choosing appropriate cutoff values to maintain a constant false positive rate across different protein levels (i.e., SpC levels), it is expected this will reduce the overall false negative rate, particularly for proteins with higher SpCs.