MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

MicroRNA-Based Survival and Relapse Prognosis for Oropharyngeal Cancer Treatment by Use of Cox Regression and Belief Function Theory

S He1*, Z Yi1 , S Ruan2 , M Anastasio1 , S Mutic1 , W Thorstad1 , H Gay1 ,X Wang1 , H Li4 , (1)Washington University in St. Louis, Saint Louis, USA (2) University of Rouen, Rouen, France

Presentations

(Wednesday, 7/17/2019) 10:15 AM - 12:15 PM

Room: 301

Purpose: MicroRNAs can be employed as prognostic biomarkers to identify whether patients are likely to fail standard therapy and to support clinical decision-making in individualized oropharyngeal cancer treatments. However, selecting optimal microRNA biomarkers and using them for accurate prognosis of treatment outcomes is very challenging due to the uncertainty and redundancy in microRNA features (biomarkers). In this method, we enhanced the developed Belief Function Theory-based method by incorporating a cox regression (CR)-based method to improve the selection of optimal prognostic biomarker candidates and the overall prognosis performance.

Methods: The proposed method includes: 1) microRNA biomarker pre-selection using the CR method, 2) biomarker refinement using the BFT-based method, and 3) outcome prediction using the selected biomarkers. CR-based method is first employed to evaluate the significance of each biomarker to prediction outcomes by analyzing their correlations, then the biomarkers with dominant significance are selected as biomarker candidates. FT-based method is subsequently employed to select the prognostic biomarkers from these candidates by minimizing a specific loss function that considers three requirements of prognostic biomarkers: 1) high prognosis accuracy, 2) low uncertainty, and 3) high sparsity to reduce over-fitting risk. Finally, an evidential K-NN (EK-NN) classifier is trained to predict the survival and replace of oropharyngeal cancer given as input the selected prognostic microRNA biomarker subset.

Results: 101 and 49 patient cases were employed as the training and validation sets, respectively. Other reported methods were compared with the proposed method. Experimental results show that the proposed method shows superior prognosis performance over other reported methods, in terms of Kaplan Meier survival curve, accuracy, and area under the curve (AUC).

Conclusion: A microRNA-based prognosis method is employed to predict oropharyngeal cancer therapy outcomes. By integrating CR-based biomarker pre-selection and BFT-based biomarker refinement, the method can provide more accuracy outcome prediction results.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email