False discovery rates in proteomics: a tale of two extremes
Lennart Martens is professor of Systems Biology at Ghent University, and Associate Director of the VIB-UGent Center for Medical Biotechnology. During his PhD at Ghent University, he built the Proteomics Identifications Database (PRIDE) at EMBL-EBI. His research concerns the application of bioinformatics to high-throughput data, and the development of novel data processing and interpretation algorithms. Dr. Martens serves on the Board of the Belgian Proteomics Association, and has been elected to the HUPO Council in 2016, and to the HUPO Executive Committee in 2017. Dr. Martens is a member of the Belgian Young Academy, has received the Ghent University Prometheus Award for Research Excellence in 2014, and the European Proteomics Association’s Juan Pablo Albar ‘Proteomics Pioneer’ Award in 2015. He is an author on more than 200 research papers and on two Wiley textbooks, and serves as Editor or Editorial Board member for several journals.
A lot of attention in proteomics (as in any high-throughput analytical field) is spent on controlling the false discovery rate (FDR). Over the past several years, we have investigated in detail many of the issues faced in the more extreme proteomics studies (e.g., multi-organism proteomics, metaproteomics, proteogenomics), and in doing so have discovered both the strengths and the limitations of current approaches. Here, two novel approaches will be presented that each tackle a different kind of problem in accurate FDR control: an approach to control the FDR for small sub-samples, as for instance encountered in host-pathogen studies, and the first ever approach that can verifiably control FDR accurately in open-ended modification searches while maintaining high sensitivity and high speed.