Data-assisted protein structure modeling by global optimization in CASP12.

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Protein structure modeling is crucial for understanding biological function.
• Experimental data, such as SAXS and cross-linking, can guide computational modeling.
• Previous work demonstrated successful modeling using NOE and contact data.

Purpose of the Study:

• To evaluate the utility of SAXS and cross-linking data in protein structure modeling.
• To develop and integrate novel energy terms for experimental data into modeling workflows.
• To assess the impact of these data types on model accuracy and native-likeness.

Main Methods:

• Developed separate energy terms for SAXS and cross-linking data.
• Integrated these terms with existing template and de novo energy functions.
• Optimized the combined energy function for CASP12 targets.
• Compared model performance against experimental data and X-ray structures.

Main Results:

• Models generated using SAXS data showed better fit to SAXS data than X-ray structures.
• The improvement in model accuracy using SAXS data alone was not significant compared to models without it.
• Consistent structural improvements were observed when incorporating cross-linking data.
• Cross-linking data provided valuable low-resolution distance information.

Conclusions:

• SAXS and cross-linking data can assist protein structure modeling, but their sufficiency varies.
• Cross-linking data appears more consistently beneficial for structural improvement.
• Combined energy functions incorporating experimental data show promise for enhancing model accuracy.