Assessing The Predictive Value Of Regulatory Molecules For Patient Outcome In Pancreatic Cancer: A Computational...

Pancreatic Cancer (PC) can be characterized as one of the most lethal cancers considering its poor diagnosis and symptoms in early stages. To assess the predictive value of regulatory molecules in terms of differentially expressed genes, we first performed a thorough search of gene expression profiling studies in pancreatic cohorts. We obtained the genes that have been identified and validated experimentally to be associated with patient outcome and also differentially expressed in tumors compared with adjacent non-tumor tissues. A two-step upstream analysis on the derived set of the genes under study was performed. The subsequent promoter and pathway analysis unveiled candidate transcription factors and regulatory molecules that potentially have regulated the detected differentially expressed genes. Predictive analysis was applied in the identified regulators and classification algorithms were implemented to model accurately patient outcome. In view of our findings, Gaussian Naïve Bayes model exhibited the highest classification accuracy and f-score concerning the predictive value of regulatory molecules in PC (accuracy =0.85, f-score =0.84).