Gentle Introduction to the Statistical Foundations of False Discovery Rate in Quantitative Proteomics

The vocabulary of theoretical statistics can be difficult to embrace from the viewpoint of computational proteomics research, even though the notions it conveys are essential to publication guidelines. For example, "adjusted p-values", "q-values", and "false discovery rates" are essentially similar concepts, whereas "false discovery rate" and "false discovery proportion" must not be confused, even though "rate" and "proportion" are related in everyday language. In the interdisciplinary context of proteomics, such subtleties may cause misunderstandings. This article aims to provide an easy-to-understand explanation of these four notions (and a few other related ones). Their statistical foundations are dealt with from a perspective that largely relies on intuition, addressing mainly protein quantification but also, to some extent, peptide identification. In addition, a clear distinction is made between concepts that define an individual property (i.e., related to a peptide or a protein) and those that define a set property (i.e., related to a list of peptides or proteins).