CS23D

CS23D generates three-dimensional protein structures from assigned nuclear magnetic resonance (NMR) chemical shifts and amino acid sequence to enable structure determination without NOE (nuclear Overhauser effect) and/or J-coupling data.

Key Features:

• Input Formats: Accepts assigned NMR chemical shifts and sequence data, including chemical shift files in SHIFTY or BMRB formats.
• Rapid Processing: Produces a set of PDB coordinates for the protein typically within approximately 10–15 minutes.
• Methodological Approach: Integrates several preexisting programs and servers and applies maximal subfragment assembly (homology modeling) for high similarity, chemical shift threading for moderate similarity, or shift-aided de novo structure prediction using Rosetta followed by chemical shift refinement for low/no similarity.
• Success Rate: Demonstrated convergence for more than 95% of queries in tests on over 100 proteins from the BioMagResBank.
• Accuracy: Generated structures show RMSDs of approximately 0.2–2.8 Å when compared to NMR structures derived from NOE-based methods or to X-ray crystallography structures.
• Data Dependence: Performance and accuracy are contingent on the completeness of chemical shift assignments and the degree of similarity between the query sequence and known 3D folds.

Scientific Applications:

• Structural biology: Determination of protein 3D structures when NOE or J-coupling data are unavailable or difficult to obtain.
• Functional annotation: Providing structural models for proteins of unknown structure to support functional inference.
• Drug discovery: Generating rapid structural hypotheses to inform early-stage ligand design and screening.

Methodology:

Integrates multiple existing programs and servers and selects between maximal subfragment assembly (homology modeling), chemical shift threading, or shift-aided de novo prediction using Rosetta followed by chemical shift refinement.