Title: Classification of proteomic data with multiclass Logistic Partial Least Squares algorithm

Authors: Zhenqiu Liu, Dechang Chen, Jianjun Paul Tian

Addresses: Division of Biostatistics, Greenebaum Cancer Center, University of Maryland Medicine, 22 South Greene Street, Baltimore, 21201 MD, USA. ' Department of Preventive Medicine and Biometrics, Uniformed Services, University of the Health Sciences, 4301 Jones Bridge Road, Bethesda, 20814 MD, USA. ' Mathematics Department, College of William and Mary, Williamsburg, 23187, VA, USA

Abstract: Early detection of cancer is crucial for successful treatments. In this paper, we propose a multiclass Logistic Partial Least Squares (LPLS) algorithm for classification of normal vs. cancer using Mass Spectrometry (MS). LPLS combines the multiclass logistic regression with Partial Least Squares (PLS) algorithm. Wavelet decomposition is also proposed for pre-processing of original data. Wavelet decomposition and the proposed LPLS are applied to real life cancer data. Experimental comparisons show that LPLS with wavelet decomposition outperforms other methods in the analysis of MS data.

Keywords: mass spectrometry; wavelet decomposition; logistic regression; partial least squares; PLS; bioinformatics; classification; proteomic data; cancer detection; early detection.

DOI: 10.1504/IJBRA.2008.017160

International Journal of Bioinformatics Research and Applications, 2008 Vol.4 No.1, pp.1 - 10

Published online: 17 Feb 2008 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article