Biomarker selection system, employing an iterative peak selection method, for identifying biomarkers related to prostate cancer

Authors:
Panagiotis Bougioukos;Dionisis Cavouras;Antonis Daskalakis;Ioannis Kalatzis;Spiros Kostopoulos;Pantelis Georgiadis;George Nikiforidis;Anastasios Bezerianos
Affiliations:
Department of Medical Physics, School of Medicine, University of Patras, Rio , Greece;Medical Signal and Image Processing Lab, Department of Medical Instrumentation, Technology, Technological Education Institution of Athens, Athens, Greece;Department of Medical Physics, School of Medicine, University of Patras, Rio , Greece;Medical Signal and Image Processing Lab, Department of Medical Instrumentation, Technology, Technological Education Institution of Athens, Athens, Greece;Department of Medical Physics, School of Medicine, University of Patras, Rio , Greece;Department of Medical Physics, School of Medicine, University of Patras, Rio , Greece;Department of Medical Physics, School of Medicine, University of Patras, Rio , Greece;Department of Medical Physics, School of Medicine, University of Patras, Rio , Greece
Venue:
CAIP'07 Proceedings of the 12th international conference on Computer analysis of images and patterns
Year:
2007

Citing 3
Cited 0

Probabilistic neural networks

Neural Networks
Reproducibility of SELDI-TOF protein patterns in serum: comparing datasets from different experiments

Bioinformatics
Analysis of mass spectral serum profiles for biomarker selection

Bioinformatics

Quantified Score

Hi-index	0.00

Visualization

Abstract

A biomarker selection system is proposed for identifying biomarkers related to prostate cancer. MS-spectra were obtained from the National Cancer Institute Clinical Proteomics Database. The system comprised two stages, a pre-processing stage, which is a sequence of MS-processing steps consisting of MSspectrum smoothing, novel iterative peak selection, peak alignment, and a classification stage employing the PNN classifier. The proposed iterative peak selection method was based on first applying local thresholding, for determining the MS-spectrum noise level, and second applying an iterative global threshold estimation algorithm, for selecting peaks at different intensity ranges. At each global threshold, an optimum sub-set of these peaks was used to design the PNN classifier for highest performance, in discriminating normal cases from cases with prostate cancer, and thus indicate the best m/z values. Among these values, the information rich biomarkers 1160.8, 2082.2, 3595.9, 4275.3, 5817.3, 7653.2, that have been associated with the prostate gland, are proposed for further investigation.