Protein structure prediction guided by cross-linking restraints - A systematic evaluation of the impact of the cross-linking spacer length

摘要

Recent development of high-resolution mass spectrometry (MS) instrumentsenables chemical cross-linking (XL) to become a high-throughput method forobtaining structural information about proteins. Restraints derived from XL-MSexperiments have been used successfully for structure refinement andprotein-protein docking. However, one formidable question is under whichcircumstances XL-MS data might be sufficient to determine a protein's tertiarystructure de novo? Answering this question will not only include understandingthe impact of XL-MS data on sampling and scoring within a de novo proteinstructure prediction algorithm, it must also determine an optimal cross-linkertype and length for protein structure determination. While a longercross-linker will yield more restraints, the value of each restraint forprotein structure prediction decreases as the restraint is consistent with alarger conformational space. In this study, the number of cross-links and their discriminative power wassystematically analyzed in silico on a set of 2,055 non-redundant protein foldsconsidering Lys-Lys, Lys-Asp, Lys-Glu, Cys-Cys, and Arg-Arg reactivecross-linkers between 1 {\AA} and 60 {\AA}. Depending on the protein size aheuristic was developed that determines the optimal cross-linker length. Next,simulated restraints of variable length were used to de novo predict thetertiary structure of fifteen proteins using the BCL::Fold algorithm. Theresults demonstrate that a distinct cross-linker length exists for whichinformation content for de novo protein structure prediction is maximized. Thesampling accuracy improves on average by 1.0 {\AA} and up to 2.2 {\AA} in themost prominent example. XL-MS restraints enable consistently an improvedselection of native-like models with an average enrichment of 2.1.