The use of social network analysis in disease biomarker detection

Especially, in recent years, the use of social network analysis has gained interest in biomarker discovery studies. Hybrid approaches involve social network analysis as a step of the feature selection process bring a different perspective to identify disease-specific biomarkers. In this thesis, dimension reduction, clustering and community detection methods used in the different steps of the hybrid approach called "SocialNetworkFeature Selection (SNFS)" were briefly reviewed; the different combinations of these methods in the steps of SNFS were compared by using open access genomic microarray data sets in terms of the effects on classification performance of Support Vector Machine (SVM) classifier. In addition, a simulation study was conducted to examine the changes in classification performance obtained from SVM classifier with the use of SNFS. In conclusion, it had been seen that SNFS approach applied in R improves the classification performance of SVM classifier tremendously and dimension reduction with SNFS has positive effects on classification performance in case of highdimensional data.