Towards low false discovery rate estimation for protein-protein interactions detected by chemical cross-linking

Abstract

Chemical cross-linking (CX) of proteins in vivo or in cell free extracts followed by mass spectrometric (MS) identification of linked peptide pairs (CXMS) can reveal protein-protein interactions (PPIs) both at a proteome wide scale and the level of cross-linked amino acid residues. However, error estimation at the level of PPI remains challenging in large scale datasets. Here we discuss recent advances in the recognition of spurious inter-protein peptide pairs and in diminishing the FDR for these PPI-signaling cross-links, such as the use of chromatographic retention time prediction, in order to come to a more reliable reporting of PPIs.

Cross-link scrambling in peptide pairs

“…In the remaining 997 precursor ions no mass overlap between the different possible charge states in each SCXC fraction occurs. This high predictability is both due to the extremely low FDR of self-links as identified by pLink 2 and inter-links after application of a composite filter [7,31] and to the large increase in retention time with an increase of one net charge. This large shift induced by an increase in net charge is illustrated by nine cross-linked peptide pairs from the used precursor collection.…”

Cross-Linking Mass Spectrometry for Investigating Protein Conformations and Protein–Protein Interactions─A Method for All Seasons

“…Interpeptide cross-links (see Scheme ) created by the covalent connection of two peptides by a cross-linker make an FDR estimation inherently more complicated compared with conventional proteomics experiments. Nevertheless, the commonly used target-decoy approach used for FDR calculation in proteomics experiments is widely applicable in XL-MS experiments as well. The decoy database is usually constructed from the reversed or shuffled amino acid sequences of the target database . In contrast to classical proteomics experiments, each peptide has an individual probability of being wrong in XL-MS data sets.…”

“…In a recent study, a stable fraction (∼80%) of identified cross-links, satisfying the maximum cross-linker distance, was found to be independent of the FDR stringency applied. This finding raises the important question of whether a structural validation of cross-links is indeed a suitable tool for a quality assessment of XL-MS data sets . Deriving a consensus on FDR estimation within the XL-MS community is highly desirable, particularly one that is applicable for a wide variety of sample types and software analysis tools.…”

“…The decoy database is usually constructed from the reversed or shuffled amino acid sequences of the target database. 214 In contrast to classical proteomics experiments, each peptide has an individual probability of being wrong in XL-MS data sets. In the targetdecoy approach, false-positives are generated and compared to cross-links containing two peptides from the target database (target−target, TT).…”

“…This finding raises the important question of whether a structural validation of cross-links is indeed a suitable tool for a quality assessment of XL-MS data sets. 214 Deriving a consensus on FDR estimation within the XL-MS community is highly desirable, particularly one that is applicable for a wide variety of sample types and software However, the simplest way to perform a computational analysis remains, so far, to map the cross-links obtained from XL-MS experiments on the 3D-structures available in the Protein Data Bank, 217 despite the concerns. 114 These structures have usually been determined by X-ray crystallography, cryo-EM, or NMR spectroscopy.…”