DREAMweb: An online tool for graph-based modeling of NMR protein structure

Area of Science:

• Biophysics
• Computational Biology
• Structural Biology

Background:

• Accurate protein structural models are crucial for understanding biological function.
• Existing algorithms may produce artifacts or struggle with sparse experimental data.
• The Distance Geometry Problem (DGP) is central to protein structure determination.

Purpose of the Study:

• To introduce DREAMv2, an enhanced algorithm for protein structure modeling.
• To improve accuracy, reduce artifacts, and enhance secondary structure conformance in protein models.
• To provide an efficient web resource (DREAMweb) for protein structure determination.

Main Methods:

• DREAMv2 incorporates a tighter bound in the optimization constraint set.
• Employs a bottom-up strategy, building and consolidating smaller substructures.
• Models regions with sparse experimental data efficiently, reducing artifacts.
• Balances performance and accuracy by solving smaller substructures ($$ atoms) under specific conditions.

Main Results:

• DREAMv2 significantly reduces artifacts in protein structure modeling compared to DREAM.
• Achieves improved secondary structure conformance consistent with experimental data.
• Benchmarks on 10 structures demonstrate the method's effectiveness.
• Efficiently models regions with sparse experimental data.

Conclusions:

• DREAMv2 offers a more accurate and reliable approach to protein structure modeling.
• The DREAMweb resource facilitates protein structure determination using NMR data.
• DREAMv2 is compatible with various NMR-based protein structure determination workflows, including iterative approaches with incomplete or ambiguous assignments.