|Publication Type||Journal Article|
|Year of Publication||2008|
|Authors||Gad Getz, Joshua Gold, SM|
|Keywords||Differential analysis, Marker selection, Multiple-hypotheses Testing|
In this paper, we present a general method for the efficient estimation of p-values based on permutation tests, and we apply it to the task of marker selection from gene expression data. The proposed algorithm is tailored to the problem of testing several thousands hypotheses of which only a small proportion is expected to be rejected. It works by iteratively removing from the computation those Hypotheses whose estimated p-value has reached a desired level of accuracy, thus investing most of the computational resources on the estimation of the p-values for those hypotheses most likely to be rejected. We present experimental results on simulated and real data and theoretical estimates of the boosting factor we may expect to achieve under several representative circumstances.