Multi-RELIEF: a method to recognize specificity determining residues from multiple sequence alignments using a Machine Learning approach for feature weighting
Division of Medical Chemistry, LACDR, Leiden University, P.O. Box 9502, 2300 RA, Leiden, The Netherlands
Dept. of Computer Science, IBIVU, Vrije Universiteit, De Boelelaan 1081A, 1081 HV, Amsterdam, The Netherlands 数据挖掘研究院
*To whom correspondence should be addressed. Elena Marchiori, elena@cs.vu.nl 数据挖掘研究院
 |
Abstract |
|---|
Motivation:
Identification of residues that account for protein function specificity is crucial, not only for understanding the nature of functional specificity, but also for protein engineering experiments aimed at switching the specificity of an enzyme, regulator or transporter. Available algorithms generally use multiple sequence alignments to identify residue positions conserved within subfamilies but divergent in between. However, many biological examples show a much subtler picture than simple intra-group conservation versus intergroup divergence. 数据挖掘研究院
Results: We present multi-RELIEF, a novel approach for identifying specificity residues that is based on RELIEF, a state-of-the-art Machine Learning technique for feature weighting. It estimates the expected ‘local’ functional specificity of residues from an alignment divided in multiple classes. Optionally, 3D structure information is exploited by increasing the weight of residues that have high-weight neighbors. Using ROC curves over a large body of experimental reference data, we show that a) multi-RELIEF identifies specificity residues for the seven test-sets used, b) incorporating structural information improves prediction for specificity of interaction with small molecules, c) comparison of multi-RELIEF with four other state-of-the-art algorithms indicates its robustness and best overall performance.
Availability:A web-server implementation of multi-RELIEF is available at www.ibi.vu.nl/programs/multirelief. Matlab source code of the algorithm and data sets are available on request for academic use.
Associate Editor: Prof. Burkhard Rost 数据挖掘实验室
