Design and Statistical Analysis of Mass-Spectrometry-Based Quantitative Proteomics Data

Proteomics as all high throughput omics data require extensive scrutiny of statistical analysis. An appropriate data analysis method should fit the characteristics of the proteomics studies and the experimental design, as well as provide an accurate answer to the question of interest. Analysis of binary experiments when one experimental condition is compared to control is quite straightforward and many statistical tests are routinely used. Analysis large datasets become much more complicated when two or more experimental, very often intertwining, manipulations of the biological system are used. To be able to decipher whether observed significant change is the sum or subtraction of two or more conditions, a set of controls need to be included and sample size calculated. Normalization is the subsequent step of data acquisition. Normalization will depend on several factors such as sample used for analytical analysis, sample processing, i.e. chemical or metabolic labeling etc. Because data can be inherently skewed, some form of mathematical transformation needs to be applied. In this chapter, we will first describe a couple of widely used MS-based quantitative proteomics experiment types. Following that, we will discuss the concepts and challenges for experimental design and statistical analysis of proteomics data for each type of quantitative MS-based proteomics study.