Affiliations: [a] Institute of Biology, Medicinal Chemistry and Biotechnology, National Hellenic Research Foundation, Athens, Greece | [b] Department of Computer Science and Biomedical Informatics, University of Central Greece, Lamia, Greece | [x] University of Central Greece | [y] Democritus University of Thrace, Greece | [z] Frederick University, Cyprus
Correspondence:
[*]
Corresponding author: Aristotelis Chatziioannou, Institute of Biology, Medicinal Chemistry and Biotechnology, National Hellenic Research Foundation, 48 Vassileos Constantinou Ave., Athens 11635, Greece. E-mail: achatzi@eie.gr
Abstract: Obstructive Nephropathy (ON) is a renal disease, which pathological profile is the result of various, tightly coupled and co-regulated, molecular processes, pervading various layers of molecular dissection. In this context, an important goal is the integration of experimental data providing multi-faceted description regarding the interweaving of the cell's molecular circuitry (here transcriptomic and epigenomic) and how this confers to the variability of the disease phenotype. The exploitation of tools or methodologies from the field of artificial intelligence, decision support and data mining aspires to facilitate the interpretation procedure of such experimental data. In the current study, we apply an intelligent workflow for predictive analytical purposes, on an integrative ON dataset, encompassing human micro-RNA (miRNA) microarray data and mouse orthologous messenger RNA (mRNA) microarray data. The workflow is implemented in Rapidminer, a powerful open access data mining and predictive analysis platform. Our scope is i) the selection of the most reliable predictive biomarkers in the two aforementioned molecular information levels and ii) the assessment of their classification power for discriminating ON severity related classes. A forward feature selection method and an evolutionary feature selection method are initially applied. The selected features, which comprise ON biomarkers to be evaluated in future studies, are next fed to a series of classifiers and results show that high classification accuracies could be obtained.
