Comparative Analysis of Chemical Cross-Linking Mass Spectrometry Data Indicates That Protein STY Residues Rarely React with N-Hydroxysuccinimide Ester Cross-Linkers.

When it comes to mass spectrometry data analysis for identification of peptide pairs linked by N-hydroxysuccinimide (NHS) ester cross-linkers, search engines bifurcate in their setting of cross-linkable sites. Some restrict NHS ester cross-linkable sites to lysine (K) and protein N-terminus, referred to as K only for short, whereas others additionally include serine (S), threonine (T), and tyrosine (Y) by default. Here, by setting amino acids with chemically inert side chains such as glycine (G), valine (V), and leucine (L) as cross-linkable sites, which serves as a negative control, we show that software-identified STY-cross-links are only as reliable as GVL-cross-links. This is true across different NHS ester cross-linkers including DSS, DSSO, and DSBU, and across different search engines including MeroX, xiSearch, and pLink. Using a published data set originated from synthetic peptides, we demonstrate that STY-cross-links indeed have a high false discovery rate. Further analysis revealed that depending on the data and the search engine used to analyze the data, up to 65% of the STY-cross-links identified are actually K-K cross-links of the same peptide pairs, up to 61% are actually K-mono-links, and the rest tend to contain short peptides at high risk of false identification.

pDeepXL: MS/MS Spectrum Prediction for Cross-Linked Peptide Pairs by Deep Learning.

In cross-linking mass spectrometry, the identification of cross-linked peptide pairs heavily relies on the ability of a database search engine to measure the similarities between experimental and theoretical MS/MS spectra. However, the lack of accurate ion intensities in theoretical spectra impairs the performance of search engines, in particular, on proteome scales. Here we introduce pDeepXL, a deep neural network to predict MS/MS spectra of cross-linked peptide pairs. To train pDeepXL, we used the transfer-learning technique because it facilitated the training with limited benchmark data of cross-linked peptide pairs. Test results on more than ten data sets showed that pDeepXL accurately predicted the spectra of both noncleavable DSS/BS3/Leiker cross-linked peptide pairs (>80% of predicted spectra have Pearson's r values higher than 0.9) and cleavable DSSO/DSBU cross-linked peptide pairs (>75% of predicted spectra have Pearson's r values higher than 0.9). pDeepXL also achieved the accurate prediction on unseen data sets using an online fine-tuning technique. Lastly, integrating pDeepXL into a database search engine increased the number of identified cross-link spectra by 18% on average.