Knowledge Discovery in Spectral Data by Means of Complex Networks

Metabolites

Massimiliano Zanin¹  David Papo⁴  José Luis González Solís²  Juan Carlos Martínez Espinosa²  Claudio Frausto-Reyes³  Pascual Palomares Anda⁵  Ricardo Sevilla-Escoboza²  Stefano Boccaletti⁴  Ernestina Menasalvas⁴ 

[1] Faculdade de Ciências e Tecnologia, Departamento de Engenharia Electrotécnica, Universidade Nova de Lisboa, Portugal; E-Mail:;Biophysics and Biological Science Laboratory, Centro Universitario de los Lagos, Universidad de Guadalajara, 47460, Lagos de Moreno, Jalisco, Mexico; E-Mails:;Centro de Investigaciones en Óptica, A. C. 20200, Aguascalientes, Mexico; E-Mail:;Centre for Biomedical Technology, Polytechnic University of Madrid Pozuelo de Alarcón, 28223 Madrid, Spain; E-Mails:;Hospital Regional de Alta Especialización del Bajío 37660, León, Gto., Mexico; E-Mail:

关键词: complex networks;    data mining;    spectroscopy;    classification;   

DOI  :  10.3390/metabo3010155

来源: mdpi

【 摘 要 】

In the last decade, complex networks have widely been applied to the study of many natural and man-made systems, and to the extraction of meaningful information from the interaction structures created by genes and proteins. Nevertheless, less attention has been devoted to metabonomics, due to the lack of a natural network representation of spectral data. Here we define a technique for reconstructing networks from spectral data sets, where nodes represent spectral bins, and pairs of them are connected when their intensities follow a pattern associated with a disease. The structural analysis of the resulting network can then be used to feed standard data-mining algorithms, for instance for the classification of new (unlabeled) subjects. Furthermore, we show how the structure of the network is resilient to the presence of external additive noise, and how it can be used to extract relevant knowledge about the development of the disease.

【 授权许可】

CC BY   
© 2013 by the authors; licensee MDPI, Basel, Switzerland.