Modeling proteins using a super-secondary structure library and NMR chemical shift information

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• Predicting protein structures for sequences lacking similarity to known structures remains a significant challenge.
• The library of protein backbone supersecondary structure motifs (Smotifs) is considered saturated, suggesting potential for structure assembly.
• Limited experimental data can potentially bridge conformational differences between target proteins and known Smotif structures.

Purpose of the Study:

• To develop and validate a hybrid protein modeling algorithm.
• To predict protein structures using an exhaustive Smotif library and nuclear magnetic resonance (NMR) chemical shift patterns.
• To assess the algorithm's performance in cases with no sequence similarity to solved structures.

Main Methods:

• A hybrid modeling algorithm was developed, integrating an exhaustive Smotif library.
• Nuclear magnetic resonance (NMR) chemical shift patterns were utilized as input.
• No primary sequence information was required for the modeling process.

Main Results:

• The algorithm was tested on 102 proteins.
• 90 homology-model-quality protein models were generated.
• 24 high-quality models were achieved, with topological correctness in almost all cases.

Conclusions:

• The developed hybrid algorithm effectively models protein structures without sequence similarity.
• The approach leverages Smotifs and NMR chemical shifts for accurate structure prediction.
• This method offers a promising avenue for modeling larger proteins with available chemical shift data.