Small angle X‐ray scattering and cross‐linking for data assisted protein structure prediction in CASP 12 with prospects for improved accuracy

Abstract

Experimental data offers empowering constraints for structure prediction. These constraints can be used to filter equivalently scored models or more powerfully within optimization functions toward prediction. In CASP12, Small Angle X-ray Scattering (SAXS) and Cross-Linking Mass Spectrometry (CLMS) data, measured on an exemplary set of novel fold targets, were provided to the CASP community of protein structure predictors. As solution-based techniques, SAXS and CLMS can efficiently measure states of the full-length sequence in its native solution conformation and assembly. However, this experimental data did not substantially improve prediction accuracy judged by fits to crystallographic models. One issue, beyond intrinsic limitations of the algorithms, was a disconnect between crystal structures and solution-based measurements. Our analyses show that many targets had substantial percentages of disordered regions (up to 40%) or were multimeric or both. Thus, solution measurements of flexibility and assembly support variations that may confound prediction algorithms trained on crystallographic data and expecting globular fully-folded monomeric proteins. Here, we consider the CLMS and SAXS data collected, the information in these solution measurements, and the challenges in incorporating them into computational prediction. As improvement opportunities were only partly realized in CASP12, we provide guidance on how data from the full-length biological unit and the solution state can better aid prediction of the folded monomer or subunit. We furthermore describe strategic integrations of solution measurements with computational prediction programs with the aim of substantially improving foundational knowledge and the accuracy of computational algorithms for biologically-relevant structure predictions for proteins in solution.

Assessment of chemical‐crosslink‐assisted protein structure modeling in CASP13

“…Crosslinking was carried out according to previously described procedures . Briefly, target proteins were crosslinked separately using sulfosuccinimidyl 4,4′‐azipentanoate (sulfo‐SDA) (Thermo Scientific Pierce, Rockford, IL) in a two‐stage reaction (using eight different crosslinker‐to‐protein ratios: 0.13:1, 0.19:1, 0.25:1, 0.38:1, 0.5:1, 0.75:1, 1:1 and 1.5:1 [w/w], a protein concentration of 0.5 mg/mL and using 20 μg protein aliquots), with reaction of the NHS‐ester firstly, subsequently followed by UV photoactivation at 365 nm, from a UVP CL‐1000 UV Crosslinker (UVP Inc.).…”

Small angle X‐ray scattering‐assisted protein structure prediction in CASP13 and emergence of solution structure differences

“…However, many CASP12 targets had intrinsically disordered regions and/or were multimeric, as we have found with most proteins that we have studied by SAXS . Comparing the sequence of the SAXS sample and what was modeled in the respective crystal structure, the average CASP12 crystal structure was missing 20% of the sequence with an extreme of 44% . These were generally terminal ends of the protein and were largely predicted from sequence to be intrinsically disordered.…”

“…Similar discrepancies resulted from CASP12 predictor's lack of awareness of why modeling the proper multimer to the SAXS data is essential. Over 50% of targets were multimers but many predictors fit the data against a monomeric structure. On the data side, there were issues when targets were stoichiometrically heterogeneous, as data were collected by HT‐SAXS.…”

“…Therefore, CASP13 included SEC‐SAXS, which can separate out stoichiometrically diverse populations and allow data collection on monodisperse sample. These strategies were suggested following the CASP12 assessment to increase accuracy …”

An integrated workflow for crosslinking mass spectrometry

“…UGGT was one of the data‐assisted de novo folding targets of CASP12 for which we contributed data in the form of 433 unique residue pairs obtained at a 5% FDR (http://predictioncenter.org/download_area/CASP12/extra_experiments/; Appendix Fig S1A) using SDA as crosslinker and 26 LC‐MS runs (Ogorzalek et al , ). Using sequential digestion, we now identified 1,523 unique residue pairs in only 12 runs (Appendix Fig S1B and C, Dataset EV1 and EV2).…”