hpr.predict (His-phosphotransferase)
nm23h2.predict (Nucl-diphosphate kinase)
antibody_E5.2.pred
p450eryf.predict (cytochrome)
crabp.predict (retinoic acid binding)
COMPARATIVE MODELING PROCEDURE - ABSTRACT Five homology models were submitted for the first Critical Assessment of Techniques for Protein Structure Prediction Meeting in Asilomar, CA this December. The meeting has been organized around a blind prediction contest in which modelers from around the world have been provided with sequences and database information for molecules which were solved this year. Three categories were delineated: threading, ab initio and comparative (homology) modelling. This meeting represents the first organized assessment of the various methods for structure prediction.
METHOD- The global free energy optimization method used was the Biased Probability Monte Carlo (BPMC) conformational search using the ICM program (Abagyan and Totrov, J. Mol Biol., 1994; Abagyan and Totrov, Kuznetsov J. Comp. Chem., 1994). The steps in the procedure were as follows:
1) A multiple sequence alignment with solved structures was used as the starting point. The starting object consisted of the prediction sequence assigned standard geometry and regularized to the nearest related solved structure. The alignment and visual inspection revealed loop regions in which either insertions or deletions were present.
2) A model representing the starting structure with each of these loops deleted was constructed and optimized to provide a low energy environment against which loops or terminal extensions would be predicted. In the case of targets in which no insertions or deletions were present the modeling procedure ended here. Optimization at this step consisted of:
3) The loops were re-inserted into this optimized fragment and modeled by a local deformation BPMC procedure. Each loop was added individually and modelled separately in the absence of the others.
4) The best conformation from each loop modeling simulation was assigned resulting in a complete protein with each of the loops predicted. In the case of inter-acting loops (i.e. antibody combining site). Sequence conservation and visual inspection of related structures were used to select loops for fixation against which the least reasonable/conserved loop would be modelled. If the loops did not interact significantly, step 5) was immediately implemented.
5) A final optimization of the same nature as that in step 2 was performed using as the conformational space all the nonconserved side chains plus those in the loop regions. Energy was the discriminating function throughout the procedure, including terms for van der Waals interactions, hydrogen bonding, electrostatics, torsional energy and solvation energy (Abagyan and Totrov, J. Mol Biol., 1994).
SCREENING LIBRARY- Protein Data Bank
Submissions:
hpr.predict (His-phosphotransferase)
nm23h2.predict (Nucl-diphosphate kinase)
antibody_E5.2.pred
p450eryf.predict (cytochrome)
crabp.predict (retinoic acid binding)
Structural Biology home page Asilomar Conference home page