Abstract:
|
Predicting the survival time of a cancer patient based on their genome-wide gene expression remains a challenging problem. In particular, for certain types of cancer, the effects of gene-expression are both weak and abundant, so identifying nonzero effects with reasonable accuracy is difficult. As an alternative to the existing methods, we propose a Gaussian process accelerated failure time model. Using a Monte-Carlo EM algorithm, we impute censored failure times and estimate model parameters jointly via maximum likelihood. We demonstrate our method's accuracy in predicting the survival times of patients with kidney renal clear cell carcinoma based on the expression of more than 20,000 genes. The proposed method is broadly applicable as it can accommodate right, left, and interval censoring; and provides a simple way to integrate multiple types of omics data.
|