Additional Predictive Value of Microarray Data Compared to Clinical Variables

logistic regression, boosting, generalized linear models, prediction, microarray data, clinical data, breast cancer

Microarrays as a promising technology for prediction of cancer diagnosis have attracted attention of many researchers in recent years. Researchers often neglect clinical data used for prediction of diagnosis compared to pre-microarray era. An important problem is determination of an additional predictive value of microarray data in relation to clinical variables. We propose a new two-step method combining logistic regression and BinomialBoosting models, to determine the additional predictive value of microarray data. This method is evaluated on two benchmark breast cancer datasets together with other already published method. The new method can combine clinical and microarray data more effectively and enables simple addition of various types of data into the combined prediction.