|Home | About | Journals | Submit | Contact Us | Français|
MicroRNAs (miRNAs) have demonstrated their potential as diagnostic and prognostic tools. While early studies mainly addressed tissue profiling, there is now a trend towards measuring peripheral profiles, including miRNA fingerprints in blood plasma, blood serum or urine in order to discover diagnostic biomarkers. We used a fully automated microarray platform to measure all human miRNAs as annotated in the most recent version of the miRBase (version 14) for almost 1000 samples representing a variety of human diseases. To evaluate the profiles we applied different hypothesis tests including t-test and Wilcoxon Mann-Whitney test and also computed the AUC values of all single biomarkers. Not surprisingly, single markers sometimes show fairly good results, however, they also allowed for a significant number of false positive and false negative classifications. By combining filter feature selection techniques together with radial basis function kernel, Support Vector Machines evaluated by 10-fold Cross Validation, we achieved very high classification accuracy rates. Our results thus demonstrate that complex and specific miRNA fingerprints show high diagnostic value and are well-suited to perform an non-invasive disease diagnosis.