Cross-linked peptides are indicative of protein conformations in samples due to their reactivity with multiple proximate residues. Since their cross-linked proteins often have alternative conformations, each with available structures, large-scale cross-link data sets require advanced capabilities to readily and accurately assess the likelihood of a cross-link originating from all available structures. Traditionally, the Euclidean or solvent-accessible surface distance (SASD) separating the residues has been used to assess the possibility that a cross-linked product arises from one of many structures. However, solvent accessibility and the potential for salt bridge formation of each reactive residue can also affect the likelihood of cross-linker attachment. Here, we describe the addition to the public cross-linking database XLinkDB of a confidence score that combines SASD, solvent accessibility, and salt bridge information to better assess the likelihood of a cross-link originating from residue pair sites among possible structures. We show how this score correctly predicts the state specificity of cross-links of ADP/ATP translocase, identifying cross-links that can serve as biomarkers of particular protein conformational configurations. These efforts also reveal that 12% of cross-links on XLinkDB have no PDB structures available that yield high confidence scores and thus may have value for additional modeling to predict alternate structures.
Keller et al. (Tue,) studied this question.