DANGLE: A Bayesian inferential prediction method for protein backbone dihedral
DANGLE (Dihedral ANgles from Global Likelihood Estimates) predicts protein backbone φ and ψ angles and secondary structure assignments solely from amino acid sequence information, experimental chemical shifts and a database of known protein structures and their associated shifts. This new approach uses Bayesian inferential logic to analyse the likelihood of conformations throughout Ramachandran space, paying explicit attention to the population distributions expected for different amino acid residue types.
Simple filtering procedures can identify the most "predictable" residues, yielding 92% of all φ and ψ predictions accurate to within ±30°. In contrast to previous approaches, more than 80% of φ or ψ predictions for glycine and pre-proline are reliable. Furthermore, DANGLE provides meaningful upper and lower bounds for the predictions which are shown to represent the precision of the prediction. Over 90% of the experimental dihedral angles in the set of test proteins are within the boundary ranges suggested by DANGLE. At a lower resolution level, the program correctly assigns each residue to one of three secondary structure states (H, E or C) in 85% of cases.
DANGLE also provides an indication of the degeneracy in the relationship between shift measurements and conformation at each site. This could potentially be a useful new approach for studying the properties of denatured protein states.
Copyright (C) 2009 Nicole Cheung, Tim Stevens, Bill Broadhurst (University of Cambridge)