URL do Projecto:
=> Aderir à Portugal@Home - Predictor@Home <=
What is Predictor@Home?
Our short term plan for Predictor@Home is to test and evaluating new algorithms and methods of protein structure prediction. In the near term we will be calibrating the P@H methods against a set of known structures. In the longer term we hope to open Predictor@Home up to the community as a resource to assist in protein structure prediction.
Why predict protein structure?
The connection between protein structure [the three-dimensional disposition of chemical functionalities that comprise Nature's palette of 20 natural amino acids which form the basis for all chemical processing in living organisms] and protein sequence [the one-dimensional expression of the chemical diversity of molecular organization that Nature expresses in individual genes composing the genome] remains as one of the premier challenges to physicists, chemists, biologists and information and computer scientists today. This challenge is particularly critical as a result of our recent advances in the methodologies to elucidate all the genes of entire organisms, including the humane genome, to identify the partnering of these genes in controlling cellular processes, as in cellular networks, and the well-established link between a protein's three dimensional structure and its biochemical function.
Molecular scientists have made significant progress in addressing this challenge through the development of fundamental theories that describe the relationship between the chemical diversity of protein sequences and the energy landscape dictated by this diversity. The energy landscape theory provides a framework not only for rationalizing and predicting/suggesting existing and new experiments but for the development of computationally based algorithms to predict the structure of unknown proteins based on their sequence alone. This activity, known as protein structure prediction, is now an active area of research that brings together scientists with diverse training and expertise ranging from physics to computer science and biology. The objective of this activity is to develop, test and apply methods to directly link protein sequences to their three-dimensional structure.
What protein structure prediction efforts are ongoing?
The enterprise of protein structure prediction is now wide-spread throughout the community of biophysical scientists; however, work in this area is both highly complex and computationally intensive. In an effort to assist the development, assessment of progress, and critical review of this field an effort known as the Critical Assessment of Techniques for Protein Structure Prediction (CASP) [you can learn more about who predicts protein structures, the CASP organization and history, and recent progress in the field by visiting the CASP web site at http://predictioncenter.llnl.gov/casp6] was initiated about twelve years ago. The function of this effort was to provide target sequences for the blind prediction of protein structure from sequence to the community of protein structure predictors on a biannual basis, to serve as a platform for community review and discussion of advances in structure prediction methods. We are now entering the 6th biannual CASP "exercise", many of us working in this area view this as a competition to put forward our best prediction methods and efforts. This competition typically involves 3-4 months of dizzying mental (and electronic) exertion during the summer (May to September, typically) to hone predictions for 50-70 unknown protein structures. The results of the predictors are analyzed in the fall and the structures are released and the predictions are assessed at the CASP meeting following the "prediction season".
What is the nature of our prediction effort?
We have assembled a team of scientists to explore different aspects of protein structure prediction for the previous two CASP exercises and again for the ongoing experiment. In the past we have focused our efforts on addressing basic algorithmic and/or scientific questions related to protein structure prediction, and have directed our objectives during those prediction periods to testing hypotheses regarding the nature of the prediction problem. One of the themes that has been present in each of these trials has been the importance of computational sampling of protein configurations in searching for the correct, functionally relevant, structure.
During the current CASP "season" we plan to focus directly on the question of conformational sampling, and whether, with augmentation of our earlier methods and algorithms by orders of magnitude more computational power, we can significantly improve our ability to predict protein structure. To achieve this objective we have assembled a "structure prediction supercomputer" based on volunteered resources (the unused computer cycles on your home computers) and distributed computing using the world-wide-web. Our world-community effort to address fundamental problems of protein structure prediction (Predictor@home) is similar to other efforts to discover new drugs to treat diseases like aids (FightAIDS@home), search for extra-terrestrial intelligence (SETI@home) and explore the physical processes of protein folding (Folding@home). To achieve these important objectives, predicting protein structure to provide links and targets to new disease treatments, we need your participation!
Why/How is Predictor@Home different from Folding@Home, both seem to be addressing the same objectives?
Protein structure prediction starts from a sequence of amino acids and attempts to predict the folded, functioning, form of the protein either a priori, i.e., in the absence of detailed structural knowledge, or by homology with other known, but not identical, proteins. In the case of the a priori folding, it is a blind search based on the sequence alone. Homology modeling first identifies other proteins of known structure with some level of sequence identity to the unknown structure and then constructs a prediction for the unknown protein by homology. Both approaches utilize multi-scale optimization techniques to identify the most favorable structural models and are highly amenable to distributed computing. Predictor@Home is the first project of this type to utilize distributed computing for structure prediction. Predicting the structure of an unknown protein is a critical problem in enabling structure-based drug design to treat new and existing diseases.
Protein folding studies and the characterization of the protein folding process are based on knowledge of the final folded protein structure (in Nature) and aims to understand the process of folding, beginning from an unfolded protein chain. The endpoint of these studies is a comparison between native protein (in nature). Analysis of the folding process too is a critical outcome allowing theories for protein folding to make direct connections to experimental measurements of this process. The Folding@Home project pioneered the use of distributed computing to study the folding process. Understanding the folding process is of significance in understanding the origin of diseases that arise from protein mis-folding, such as Alzheimer's disease and Mad-Cow disease.
Both approaches explore protein structure and folding, but with complementary aims.