Unbiased False Discovery Rate Estimation for Shotgun Proteomics Based on the Target-Decoy Approach

Abstract

Target-decoy approach (TDA) is the dominant strategy for false discovery rate (FDR) estimation in mass-spectrometry-based proteomics. One of its main applications is direct FDR estimation based on counting of decoy matches above a certain score threshold. The corresponding equations are widely employed for filtering of peptide or protein identifications. In this work we consider a probability model describing the filtering process and find that, when decoy counting is used for q value estimation and subsequent filtering, a correction has to be introduced into these common equations for TDA-based FDR estimation. We also discuss the scale of variance of false discovery proportion (FDP) and propose using confidence intervals for more conservative FDP estimation in shotgun proteomics. The necessity of both the correction and the use of confidence intervals is especially pronounced when filtering small sets (such as in proteogenomics experiments) and when using very low FDR thresholds.

Scavager: A Versatile Postsearch Validation Algorithm for Shotgun Proteomics Based on Gradient Boosting

“…Final FDR at PSM and peptide levels are calculated using the following equation: where Decoys all are all decoy identifications, Decoys V are validation decoys, and Decoys T V and Targets T are validation decoys and target identifications above threshold, respectively.…”

Validation of Peptide Identification Results in Proteomics Using Amino Acid Counting

“…Intervals that satisfy all above criteria are considered in the subsequent peptide analysis. Nonredundant peptide identifications in each interval are filtered to 2% group‐specific FDR using TDA with correction according to the following equation: where i is the particular mass shift; d and t are the numbers of decoy and target peptides at the i ‐th mass shift. Using 2% FDR allows to discover low abundance mass shifts and increase the amount of peptides used for amino acid frequency calculation as compared to the traditional 1% threshold, which improves accuracy.…”

“…Intervals that satisfy all above criteria are considered in the subsequent peptide analysis. Nonredundant peptide identifications in each interval are filtered to 2% group-specific [17] FDR using TDA [16] with correction [20] according to the following equation:…”

Beyond target-decoy competition: stable validation of peptide and protein identifications in mass spectrometry-based discovery proteomics

“…This is the very reason why decoy database construction and conditions of application have been extensively studied. Among them, (i) the search engine must be compliant with TDC [4]; (ii) In theory, the longer the decoy database, the more precise the mismatch score distribution [5,6] (similarly to political polls which accuracy depends on the number of surveyed citizens); (iii) The decoys must respect the cleavage sites [7] to avoid systematic target matching regardless the spectrum quality; (iv) The inuence of randomness in the construction of the decoy database can be counter-balanced by boosting strategies, leading to less volatile FDRs [8]; (v) The way the decoys are counted has an inuence [6]. Beyond, numerous parameters have been reported for their relative importance and have been discussed [9].…”

“…However, it is possible to reconcile Cooper's observations and statistical theory. In fact, the correctness of any statistical estimate is only asymptotic: if the quality of the empirical model depicting the mismatches is improved (for instance, by increasing the size of the decoy database [5,6] or by averaging a growing number of TDC-FDRs resulting from randomly generated decoy databases, in a boosting-like strategy [8]), we should end-up with a series of estimates that converges theoretically towards the FDP.…”