|
Abstract:
|
Developing models that accurately predict the risk of local recurrence, distant metastasis and survival is clinically important. While large scale integrative genomic profiling studies have been performed for many cancer types, the prediction models have poor performance and limited clinical value for most of cancer types. This raises a question that the poor performance is due to limited information, including small sample sizes or short follow-up time, or the fact that genomics has little contribution to the variation of clinical outcomes. To answer this question, we developed a statistical model for estimating the effect size distribution of high-dimensional genomic features for survival traits and the total contrition from these features. Markov chain Monte Carlo sampling is used to carry out Bayesian posterior computation. We applied our methods to The Cancer Genomics Atlas (TCGA) clinical data. The clinical implication will be discussed.
|
