Rescuing error control in crosslinking mass spectrometry

Crosslinking mass spectrometry is a powerful tool for studying protein-protein interactions under native or near native conditions in complex mixtures. By help of novel search controls, we show that measures that aim to improve the number of identifications based on heuristic considerations can undermine error estimation in non-obvious ways. The relationship between decoys and false positives is very sensitive to the injection of information that favour likely correct matches. We identify a wider challenge in crosslinking data analysis tools with maintaining the decoy-false positive relationship, and exemplify this with the filtering of results based on the information of which proteins can be observed as having reacted with the crosslinker. Without correcting for this problem, we could identify an average of 260 interspecies protein-protein interactions in 16 analyses, "solidly" suggesting groundbreaking biological connections that dont actually exist. We also show how data analysis procedures can be tested and modified to rescue the decoy-false positive relationship. The importance of this relationship for reliably identifying protein-protein interactions cannot be overstated.