Focus on the spectra that matter by clustering of quantification data in shotgun proteomics

In shotgun proteomics, the information extractable from label-free quantification experiments is typically limited by the identification rate and the noise level in the quantitative data. This generally causes a low sensitivity in differential expression analysis on protein level. Here, we propose a quantification-first approach for peptides that reverses the classical identification-first workflow. This prevents valuable information from being discarded prematurely in the identification stage and allows us to spend more effort on the identification process. Specifically, we introduce a method, Quandenser, that applies unsupervised clustering on both MS1 and MS2 level to summarize all analytes of interest without assigning identities. Not only does this eliminate the need for redoing the quantification for each new set of search parameters and engines, but it also reduces search time due to the data reduction by MS2 clustering. For a dataset of partially known composition, we could now employ open modification and de novo searches to identify analytes of interest that would have gone unnoticed in traditional pipelines. Moreover, Quandenser reports error rates for feature matching, which we integrated into our probabilistic protein quantification method, Triqler. This propagates error probabilities from feature to protein level and appropriately deals with the noise in quantitative signals caused by false positives and missing values. Quandenser+Triqler outperformed the state-of-the-art method MaxQuant+Perseus, consistently reporting more differentially abundant proteins at 5% FDR: 123 vs. 117 true positives with 2 vs. 25 false positives in a dataset of partially known composition; 62 vs. 3 proteins in a bladder cancer set; 8 vs. 0 proteins in a hepatic fibrosis set; and 872 vs. 661 proteins in a nanoscale type 1 diabetes set. Compellingly, in all three clinical datasets investigated, the differentially abundant proteins showed enrichment for functional annotation terms. The source code and binary packages for all major operating systems are available from https://github.com/statisticalbiotechnology/quandenser, under Apache 2.0 license.