Background:
FDA approvals of drugs in clinical trials depends highly on the qualitative and  quantitative response of the patients in each test group. Choosing only those patients for clinical trials with a promising biomarker profile holds the promise of a much greater chance of getting high response rates and finally FDA approval. Bundling such a biomarker test assay helps a lot in getting orphan drug status approval by the regulatory affairs.  

Principle:
A compound is tested against 50-75 tumor xenografts growing sc in nude mice. These tumors should be chosen from tumor types that have shown at least some kind of chemo sensitivity  in the clonogenic assay. Using computational biology, a prediction model is build using 2/3 of the testes tumors data and correlated to their corresponding gene expression profile.
Validation of this prediction model is done by prediction of the remaining 1/3 of the tumors and comparison of the predictions to the real invivo outcome.

After successfull validation, this signature can be used for patient stratification for clinical trials.