Structure determination
Structural genomics: new routes from low-resolution structure to function
Structural genomics involves the systematic determination of structures for large numbers of proteins, with the ultimate goal of elucidating structural models for entire proteomes. The challenge of structural genomics is to characterise proteins having little or no structural similarity to proteins of known function, often resulting in low-resolution structures. Methods capable of accurately predicting protein function from these low-resolution structures will therefore be of increasing interest to groups seeking to exploit these growing databases of structural information.
In recent years, numerous large-scale predictions of protein structure based on comparative modeling, ab initio structure prediction, and threading-based approaches have generated a multitude of structural models, yet most of these are low-resolution structures with partially inaccurate atomic coordinates and missing atoms. Such structures are often limited to rough C-backbones showing the overall fold of the protein and the estimated position of each residue, but without the exact conformation of each side-chain. The challenge facing structural biologists today is therefore how to efficiently predict protein function from such limited and approximate protein structures.
Andras Szilagny and Jeffrey Skolnick, of the New York State Center of Excellence in Bioinformatics and Life Sciences, have just described an improved method for the efficient prediction of DNA-binding function on the basis of low-resolution protein structures, using only amino acid sequences and low-resolution, C-only protein models. Importantly, the method is tolerant of errors in atomic coordinates and can generate accurate predictions based even on inexact protein models. The Szilagny and Skolnick method compares favorably to a previously published approach that relies on higher resolution, all-atom structures that are not readily generated via structural genomics. This and similar methods may soon allow for the more efficient use of structural genomics data to predict functions for the many novel proteins still to be characterised.
Click here for useful links
