Predicting Patient Survival from Proteomic Profile Using MALDI-TOF Mass Spectrometry Data

• Activity Number: 582
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Farida Mostajabi*+ and Susmita Datta
• Companies: University of Louisville and University of Louisville
• Keywords: MALDI-TOF ; LASSO ; PLS ; regularized and penalized methods

Abstract

In this project we consider predicting survival time from the proteomic profiles of patient serum peptide using Matrix-assisted Laser Desorption/Ionization Time-of-Flight (MALDI-TOF) data of non-small cell lung cancer (NSCLC) patients. Due to large dimension of features in a mass spectrum, traditional linear regression modeling failure times with high number of proteomic features is not feasible. We address this problem by fitting linear regression model of log-transformed failure times using regularized and penalized methods such as least absolute shrinkage and selection operator(LASSO), partial least squares (PLS), elastic net regularization and sparse partial least square(SPLS). In this work, we study the performances of all these models in terms of fit to the patient survival time and prediction of the future survival times using leave-out-one cross validation technique.