Statistical Models in Bioinformatics

Abstract

In recent years, biological research has been witness of a sea change mainly spearheaded by the advent of novel high throughput technologies that can provide unprecedented amounts of valuable data. This has given rise to novel field sharing the popular suffix ‘omics’. Genomics/transcriptomics, proteomics, metabolomics, interactomics/regulomics and numerous other terms have been coined to categorize this ever increasing number of new fields. Biomarkers comprise the most critical tools for the early detection, diagnosis, prognosis and prediction of diseases providing key clues for drug development processes. A significant challenge is to define appropriate levels of specificity and sensitivity of new biomarkers in detecting complex diseases. The establishment of new biomarkers is not only an issue of optimizing wet lab experiments but also of designing appropriate and robust data analysis methods.Various approaches, like multivariate analysis methods as well as standard statistical tests have been applied to search for the important features in ‘omics’ data. Likewise, several methods, e.g. FDA, SVM, CART, nonparametric kernels, kNN, boosted decision stump and genetic algorithms, have been reported. However, it still remains an unsolved challenge to analyze and interpret the enormous volumes of ‘omics’ data.