Exploring the key genes and signaling transduction pathways related - PowerPoint PPT Presentation

Exploring the key genes and signaling transduction pathways related to the survival time of glioblastoma multiforme patients by a novel survival analysis model Department of Computer Science of Southwest University Le Zhang

Scientific Significance 1. Exploring the key genes related to the survival time of GBM 2. Investigating the gene related signaling transduction pathways 3. Solving P>>N problem for survival model

Review The classical cox proportional hazards Hong et al., [Hong et al., 2015] model[Cox et al., 1972] can only P<<N proposed a conditional SIS method to data [Crichton et al., 2002]. increase the screening performance. Tibshirani et al.,[Tibshirani et al., 1996] Developing a systematic approach to integrated the Lasso algorithm into the identify the target generic drug for the classical Cox proportional hazards cancer treatment becomes a popular model to select key genes research field[Nelander et al., 2008].

Innovotion 1. Analyzing the relation between clinical GBM gene expression and survival time (The Georgetown Database of Cancer G-DOC https://gdoc.georgetown.edu/gdoc/. ) 2. Integrating the Lasso into classical Cox model to process P>>N type of data and using the SIS algorithm to improve the predictive accuracy. 3. Employing hypergeometric test to investigate the correlated GBM signaling transduction pathways regarding the explored survival time related key genes.

Methods start Pre-processed GBM data,denoted by Fig. 1.1 Cox+Lasso by Eq. 3 > Implementation the final selected predictors end > Combined Cox and Lasso (CoxLasso) strategy （ a ） A start start > Combined Cox and SIS (CoxSis) Pre-processed GBM Pre-processed GBM strategy (b) data,denoted by Fig. 1.1 data,denoted by Fig. 1.1 obtain the parameter estimate by a > Combined Cox, SIS and Lasso Selected covariates from A by marginal Cox regression model CoxLasso,denoted by C 0 (CoxSisLasso) strategy (c) Selected covariates from A-C 0 by Rank the magnitudes of the parameter Conditional SIS,denoted by C 1 estimates, retain the top ranked covariates Selected covariates from C 0 +C 1 by CoxLasso Implement Lasso with the selected covariates the final selected predictors the final selected predictors end end C B

Results > We use ROC and AUC to compare the performance of the three strategies

Results The key genes and signaling pathways by three strategies

Conclusion > Innovatively developing a CoxSisLasso strategy to interrogate the connections between GBM gene expression and GBM patients’ survival time > Employing the hypergeometric test to investigate the incoherent signaling transduction pathways and the survival time of GBM patient. > Lacking theoretically proof for the CoxSisLasso strategy, simulation study for the gene and pathway selection platform and so on.

THANKS