Statistics and Its Interface

Volume 5 (2012)

Number 1

Bayesian false discovery rates for post-translational modification proteomics

Pages: 47 – 59

DOI: http://dx.doi.org/10.4310/SII.2012.v5.n1.a5

Author

Yan Fu (Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing, China)

Abstract

Tandem mass spectrometry-based proteomics enables high throughput analysis of post-translational modifications (PTMs) on proteins. In current researches of shotgun proteomics, peptides with various PTMs and those without PTMs are often identified together and an overall false discovery rate (FDR) is estimated. However, it is often the case that only a subset of identifications, e.g. those with specific PTMs, are emphasized or reported. In doing so, the risk arises that the FDR of reported results is seriously underor overestimated, based on which unreliable conclusions may be drawn. But unfortunately, this has not been widely realized in the field, and there is still no agreement on the right way to control the FDR of PTM identifications. As a result, the ostrich policy is commonly adopted wittingly or unwittingly, i.e., a simplistic overall estimate is assumed. This paper, for the first time, proves that the FDRs of various PTM identifications are in theory not equivalent to the overall FDR and quantifies several major factors influencing their relationships. Elaborate simulation experiments are carried out to empirically verify the theoretical conclusions. Strategies are suggested for better control of PTM FDRs.

Keywords

false discovery rate, group structure, protein identification, post-translational modification, proteomics

Full Text (PDF format)