Jianying Zhang, Charles L Sharpiro
Ohio State University, USA
Scientific Tracks Abstracts: Oncol Cancer Case Rep
Cox regression models have been used for prognostic prediction based on omics data for years. But due to model over-fitting and the improper way of model development and various other factors, very few published prognostic signature has found its clinical success in applications. This study illustrated how an improper model develop or testing procedure could mislead the result and applied the proper cross-validated (CV) Cox model building and testing procedure to identify a prognostic microRNA signature based on 125 triple negative breast cancer (TNBC) patients. In each CV procedure (K-fold), the full training data was split into train data and test data first. Then four steps were followed: feature selection, model selection, addition of significant clinical covariates, and model performance assessment. For each of the four steps, various methods were compared. For example, univariate cox model or progression vs progression-free comparison was applied to feature selection. Stepwise and penalized Cox model were used for model selection for the training data. Model performance was assessed on the test data by cross-validated AUC of 3-year or 5-year recurrence and ROC calculated using the time-dependent ROC method. The proposed optimal procedure is cross-validated stepwise selection based on the feature screening by progression vs progression-free. The final proposed cox model contained 5 miRs: miR-363, miR-155, miR-142-5p, kshv-miR-K12.11, and miR-1307. When the significant pathologic predictor NODES (positive vs negative) was added to the 5-miR model, the mean cross-validated AUC was increased by 0.06 on average. Permutation method was used to test if the cross-validated AUC of a model was significantly greater than 0.5. Enrichment analysis and supportive literature further verified the association between the five proposed oncomiRs and TNBC progression (recurrence/death). The five-miR model was validated in an independent group of N=34 TNBC patients with 2-year, 3-year and 5-year progression AUC 00.84, 0.72, 0.77, respectively. Survival analysis on the dichotomized risk groups based on the predictive risk score resulted in a P value of 0.065 (by median of validation risk score) or 0.0008 (by median of train risk score) for the log-rank test.
Jianying Zhang received her PhD degree in Statistics from Purdue University in 2008. She has worked at the Center for Biostatistics at the Wexner Medical Center of the Ohio State University, Columbus, OH for about 8 years. She has collaborated extensively with investigators in the Comprehensive Cancer Center on basic science and translational laboratory experiment and clinical trials and has been the Co-investigator of over 10 P01/U01/R01 and industrial grants. She has published over 40 collaborative papers in reputed journals.
Email: Jianying.Zhang@osumc.edu