|
Abstract:
|
In the context of precision medicine, data from gene expression has become an important source to be explored and utilized for patent selection in future clinical trials. Sometimes when the clinical data are also available, models incorporating these two data sources can be beneficial. With the large number of genes and relatively few subjects, instead of the least squares regressions, the methods such as Lasso or Elastic-net have often been used; however, these methods do not usually come up with the same model. Therefore, it can be very misleading without carefully examination of the results. In this case study, using a public data set (TARGET) from NCI’s website, we explore the modeling of prediction classifiers for overall survival of the study subjects. We will show the results from several machine learning methods and how the clinical insights can be usefully incorporated.
|
