VITAL NMR: using chemical shift derived secondary structure information for a limited set of amino acids to assess homology model accuracy

Homology modeling is a powerful tool for predicting protein structures, whose success depends on obtaining a reasonable alignment between a given structural template and the protein sequence being analyzed. In order to leverage greater predictive power for proteins with few structural templates, we...

Full description

Saved in:
Bibliographic Details
Published in:Journal of biomolecular NMR 2012, Vol.52 (1), p.41-56
Main Authors: Brothers, Michael C., Nesbitt, Anna E., Hallock, Michael J., Rupasinghe, Sanjeewa G., Tang, Ming, Harris, Jason, Baudry, Jerome, Schuler, Mary A., Rienstra, Chad M.
Format: Article
Language:English
Subjects:
Amino Acid Sequence
Amino Acids - chemistry
Biochemistry
Biological and Medical Physics
Biophysics
Databases, Protein
Models, Molecular
Nuclear Magnetic Resonance, Biomolecular - methods
Physics
Physics and Astronomy
Protein Conformation
Protein Structure, Secondary
Proteins - chemistry
Spectroscopy/Spectrometry
Structural Homology, Protein
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Homology modeling is a powerful tool for predicting protein structures, whose success depends on obtaining a reasonable alignment between a given structural template and the protein sequence being analyzed. In order to leverage greater predictive power for proteins with few structural templates, we have developed a method to rank homology models based upon their compliance to secondary structure derived from experimental solid-state NMR (SSNMR) data. Such data is obtainable in a rapid manner by simple SSNMR experiments (e.g., 13 C– 13 C 2D correlation spectra). To test our homology model scoring procedure for various amino acid labeling schemes, we generated a library of 7,474 homology models for 22 protein targets culled from the TALOS+/SPARTA+ training set of protein structures. Using subsets of amino acids that are plausibly assigned by SSNMR, we discovered that pairs of the residues Val, Ile, Thr, Ala and Leu (VITAL) emulate an ideal dataset where all residues are site specifically assigned. Scoring the models with a predicted VITAL site-specific dataset and calculating secondary structure with the Chemical Shift Index resulted in a Pearson correlation coefficient (−0.75) commensurate to the control (−0.77), where secondary structure was scored site specifically for all amino acids (ALL 20) using STRIDE. This method promises to accelerate structure procurement by SSNMR for proteins with unknown folds through guiding the selection of remotely homologous protein templates and assessing model quality.
ISSN:0925-2738
1573-5001
DOI:10.1007/s10858-011-9576-3