The Journal of Neurobehavioral Sciences
Yazarlar: Kaan Yilancioglu , Muhsin Konuk
Konular:-
DOI:10.5455/JNBS.1446109872
Anahtar Kelimeler:-
Özet: Genomic information obtained from robust analysis methods such as microarray and next generation sequencing reveals underlying disease mediating factors and potential diagnostic biomarkers. Data mining methods have been widely chosen for classification and regression studies of health sciences as well as other disciplines since the beginning. In the present study, public Gene Expression Omnibus (GEO) genome wide expression dataset (ID: GSE12679) consisting of mRNA transcripts of post-mortem brain tissues in schizophrenic and normal patients were analyzed by using Multilayer Perceptron Neural Network (MLP NN) algorithm. A set of most differentially expressed genetic features (p<0.001) were used for creating the classifier which can predict disease states in test set with ~82% accuracy. Differentially expressed genes used as classifying biomarkers gain utmost importance for revealing hidden underlying genetic factors associated with important psychiatric diseases. We could also suggest that such data mining tools might be applicable for developing genome-based diagnostic tools.