Prognostic biomarker detection, machine learning bias correction, and differential coexpression module detection

摘要: In this thesis, we present three projects on prognosis biomarker detection, machine learning bias correction and identification of differential coexpression modules in complex diseases. the first project, aimed to identify fusion transcripts that are predictive value prostate cancer prognosis, an important task avoid overtreatment patients. We discovered eight from 19 RNA-seq datasets validated its >200 patients sites (Pittsburgh, Stanford Wisconsin). The constructed prediction model showed consistently high accuracy predicting recurrence aggressiveness all cohorts. second consider a common practice apply many (up several hundred) classifiers dataset report best cross-validated accuracy. demonstrated downward using approach proposed inverse power law (IPL) method correct bias. was compared with existing methods simulation real superior performance. For third study, developed computational algorithm (MetaDiffNetwork) coexpressioin across disease conditions multiple transcriptomic studies. good performance simulated data applied it combine major depressive disorder studies (cases vs. controls) four breast (ER+ ER-). identified were by knowledge pathways. These can be used help generate new hypotheses regarding suspected genes. conclusion, areas research covered thesis critical bioinformatic elements for biomedical applications understand underlying mechanism ultimately improve patient treatment.