Award details

Prediction of protein function from structure - an integrated bioinformatics and chemoinformatics approach

Reference	BB/C516052/1
Principal Investigator / Supervisor	Professor Michael Sternberg
Co-Investigators / Co-Supervisors
Institution	Imperial College London
Department	Biological Sciences
Funding type	Research
Value (£)	216,031
Status	Completed
Type	Research Grant
Start date	15/04/2005
End date	14/04/2008
Duration	36 months

Abstract

Current structural genomics projects are yielding structures for proteins whose functions are unknown. Indeed, in the current protein data bank there are 500 proteins entries that lack an assigned function (called hypothethical proteins). Given so many proteins without a functional annotation, current bioinformatics methods to assign function must be inadequate. This proposal is to develop a novel approach to predict function from structure using an integrated bioinformatics and chemoinformatics approach. The bioinformatics approach will be based on a novel and powerful approach (3D-PROACTIVE) recently developed in the applicant¿s group. 3D-PROACTIVE assigns protein function via the Gene Ontology (GO) classification to a protein structure. In the difficult area of low sequence identity between the test query protein and the database protein with known function, 3D-PROACTIVE was far more accurate in function assignment than assuming the function is similar to that of the closest homologue (functional inheritance). In addition to predicting function, 3D-PROACTIVE predicts residues directly involved in the activity of the protein. The central concept of the proposal is to validate the bioinformatics prediction with the results from in silico docking of the appropriate putative ligand into the protein. If the docking leads to a cluster of low energy solutions within the predicted binding site, then one has far greater confidence in the bioinformatics prediction. Pilot studies confirm that this approach is viable. They key steps of the proposal are: 1. To enhance the bioinformatics prediction of protein function and the residues involved in activity. Developments will include spatial clustering of residues and the benchmarking of enhanced methods using the recently available database of function sites in the protein data bank. 2. To develop and benchmark a strategy for in silico docking of the ligands against the protein to confirm or refute the bioinformatics predictions. A library of ligands and of small molecule fragments corresponding to the predicted functions will be developed. Docking will use the program Autodock3. The clustering of the best solutions together with a consideration of the relative energy of the solution will be used as a guide to decide if a ligand (or fragment) is predicted to dock into the protein. We will check that the site into which the ligand is docked corresponds to the site predicted by bioinformatics. 3. To integrate the bioinformatics and chemoinformatics approaches, in particular in the estimation of confidence indicators for end users. A support vector machine will be trained to integrate the different predictions and return a confidence measure. 4. To develop a web server to implement the strategy. 5. To apply the function prediction in the blind trial of protein structure prediction (CASP7) to be held in 2006. CASP now includes a section on function prediction and the bioinformatics methods in particular can be tested at CASP. 6. To apply the method to proteins in the Protein Data Bank that lack functional assignment.

Summary

unavailable

Committee	Closed Committee - Biomolecular Sciences (BMS)
Research Topics	Structural Biology, Technology and Methods Development
Research Priority	X – Research Priority information not available
Research Initiative	X - not in an Initiative
Funding Scheme	X – not Funded via a specific Funding Scheme

I accept the terms and conditions of use (opens in new window)

export PDF file

back to list new search