Statistical Prediction of the Local Folds of Protein Sequences
by
Jyothi S.
Indian Institute of Technology, Bombay, India.
Coauthors: Rajani R. Joshi (Indian Institute of Technology, Bombay)

Prediction of the secondary structure of proteins is an active area of research in Bioinformatics and Protein Engineering. Most of the current computational approaches rely on statistical and neural network based pattern recognition techniques. The reliability of some of the best methods seem to be around 70%. The quite high error rate is attributed to the lack of consideration of the global influences of the chain in the prediction. NMR studies further show that oligopeptides frequently adopt an ensemble of different conformations rather than a few well- defined discretized ones.

In this respect, we have used a nonparametric regression based method for the estimation of small and medium range inter-residual distances in proteins using only the features of the primary sequence as input. We have validated our model on a large and varied sample of proteins from the Protein Data Bank (PDB) and have found that the Root Mean Square Error (RMSE) of estimation for local conformations is 1.0 to 3.0 Angstroms only. Moreover, the local conformations obtained by us are unique and stable as compared to the ensemble of conformations obtained by the popular threading approaches. This may be due to the inclusion of the global and local features of the primary sequence during distance estimation.

The validation results and application of the method to specific areas of genomics will be highlighted in this paper.

Date received: October 11, 2000

Copyright © 2000 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cafr-22.