Protein Structure and Function: Application of Bioinformatics Methods - John Rigden 2014

Integrated Servers for Structure-Based Function Prediction
Introduction

A. Laskowski's Novel

No single method for predicting protein function from its 3D Structure is perfect in its own right; some perform well in certain scenarios, while others prove more effective in different contexts. Therefore, it makes sense to apply several distinct prediction Methods to a given Cell/13.html">Protein Structure to obtain either a consensus prediction or the most plausible one. In this chapter, we describe two web servers, ProKnow (http://proknow.mbi.ucla.edu) and ProFunc (http://www.ebi.ac.uk/profunc), which employ a combination of approaches for predicting protein function from Spatial Structure.

Predicting the function of a newly solved protein structure is somewhat akin to solving an engaging detective mystery. The 3D structure of a protein undoubtedly holds the clues to its function; the challenge lies in spotting these clues, assessing their reliability, identifying and discarding dead ends, and piecing the remaining clues together to reach the final solution of the puzzle.

In reality, this problem has emerged only recently as a direct consequence of various structural Genomics projects launched at the beginning of this decade. Previously, experimentalists already knew a great deal about their Proteins before diving into structure determination, and they could select proteins based on specific biological interests. The primary goal of solving a protein structure was to elucidate how the protein executes its biological function at the atomic level. Structural genomics projects, with their high-throughput structure Determination Methods, have entirely different objectives. Nowadays, a protein is targeted for structure determination if it belongs to a family with no previously known structures, is predicted to feature a novel fold, or holds significant medical importance for a particular disease. Determining the protein's function is no longer part of these immediate goals.

As a result, an increasing number of structures are being generated for proteins whose Functions remain unknown. Indeed, roughly one-third of the structures produced by structural genomics initiatives correspond to proteins with unknown or only partially known functions. This severely limits the utility of these structures, as they no longer explain how the protein performs its function since the function itself remains essentially undefined.

However, will a known structure truly provide Answers to all our questions? After all, The history of structural biology tells us that spatial structure dictates function. Virtually every previously solved structure has helped shed light on some biological or biochemical process. Thus, once we have the structure, voilà—we have the function.

Unfortunately, in life—as in bioinformatics—things are not quite that simple. A structure can explain a function, but only if you already know what that function is. Despite the availability of the many diverse methods discussed earlier in this book, determining function based solely on structure turns out to be surprisingly difficult.

Roman A. Laskowski European Bioinformatics Institute,

Wellcome Trust Genome Campus, Hinxton, Cambridge,

CB10 1SD, UK

e-mail: roman@ebi.ac.uk