Patient selection has always been a laudable goal of targeted therapy and early successes, such as trastuzumab for the treatment of human epidermal growth factor receptor 2¨Cpositive breast cancer and imatinib for c-KIT¨Cexpressing gastrointestinal stromal tumors, fostered the concept of tailored therapy to enhance response in clinical trials and reduce unnecessary patient exposure to inactive agents. Unfortunately, this goal remains elusive for almost all other targeted agents. The need to discover response markers has become more pressing as clinical trials started to show tantalizing but low overall clinical benefit rates for most biologically targeted agents. We previously suggested that it is possible to prospectively test candidate response markers in the clinic in a similar manner as we test new drugs in phase II studies to asses their potential clinical value as patient selection tools.

The current study describes the development of three conceptually different potential gene expression¨Cbased predictors of dasatinib response for clinical evaluation. For almost all drugs, one could rationally propose candidate predictors based on the known or presumed mechanism of action of the drug and based on data from preclinical models. We applied this strategy to dasatinib and used empirical data from cell line experiments in vitro, information from mechanism of action and biological insights into cancer biology to develop several different potential response predictors. Interestingly, marked differences were seen in the reproducibility and robustness of some predictors across cell lines and human samples. For example, a Src pathway activity predictor using cell line centroids yielded results that had low reproducibility in replicate experiments and the predictions were unstable in human breast cancer samples. Data from human tissues did not correlate highly with any of the cell line centroids and therefore small changes in the data due to noise could influence class assignment.

We also show that there was substantial discrepancy in prediction results generated by the same prediction method on matching FNA and core needle biopsies of the same tumor. This is likely due to the different cellular composition of these tissues and the confounding effects of tumor stroma in core biopsies. This is an important observation that suggests that clinical trials that assess biological markers should not use these sampling methods interchangeably without showing robustness of the predictor across tissue sampling methods. Hypothetically, both methods of sampling have their advantages. Biopsies obtained by FNA are usually less painful to patients and offer a higher percentage of cancer cells for response prediction. Although core needle biopsy can be more painful, it includes higher proportion of stroma; thus the interaction between tumor cell and nonmalignant surrounding tissue are both subject to analysis.

An interesting observation in this study was that the cell line¨Cderived predictor, the Src pathway activity score, and the target expression index each identified distinct and only minimally overlapping patient populations as possibly sensitive to dasatinib. Considering the very different conceptual underpinnings of these predictors, this may not be surprising from a technical point of view. The true predictive values of the three candidate markers described in this study are unknown. As mentioned in the introduction, we have initiated a study that uses these predictors to select patients for dasatinib therapy; only individuals whose cancer is predicted to respond by one of these methods will receive kinase inhibitor dasatinib and early stopping rules apply to each marker group to stop accrual if less than expected clinical benefit is observed in the molecularly selected patient subset

As one of the world leading suppliers of high-performance life-science products. We have over 8,000 products which consist of inhibitors, antibodies, RNAis, proteins and peptides those which focus on signaling pathways such as kinase inhibitor, Velcade, High Throughput Screening & so on.