INTEGRATED GENOMIC MARKERS FOR CHEMOTHERAPEUTICS

Abstract

As translational research has created opportunity for an increasing number of anticancer agents, the need to develop computational methods to identify and understand predictive biomarkers has become emergent. This dissertation introduces a generic and systematic bioinformatics method to develop biomarker(s) for cancer therapeutics. The overall methodology includes the conceptualization of general types of biomarkers, implementation of algorithms, a uniqueness test of the signature markers in the test data using a novel computational algorithm and innovative bioinformatics algorithms to detect the presence of the signature with the pattern remained in the test data. An integrated genomic analysis to model gene expression and genomic aberrations is proposed to identify the minimal marker sets for clinical translation. We then study a novel biological phenomenon in cancer therapeutics, that cancer cells may show concordant chemo-response to multiple anticancer agents. The representative preclinical models (both cell lines and primary tumor derived explants) are selected to reflect concordant sensitive and concordant resistant tumor cells. Moreover, we developed the gene expression signature of concordant chemotherapeutics using NCI60 data to characterize the concordance of chemotherapeutics. A high predictive value (AUC = 0.88±0.10) is observed in an independent validation using Oncotest tumor clonogenic assay and gene expression data from primary xenograft tumor models. When the signature is applied to expression data from tumors of breast cancer patients treated with (TFAC) combination chemotherapy, the signature predictor predicts treatment outcome (pCR vs RD) with a p-value=0.017. We also find that the signature predictor is able to predict the survival of patients in breast cancer and lung cancer. Meta-analysis using Oncomine tools shows that more than 20 unique drug sensitivity concepts are significantly associated with the developed signature of concordant chemotherapeutics. These results demonstrate that concordance of chemotherapeutics is present in both preclinical models and clinical patients; the developed signature may have clinical utility for patients treated with standard of care chemotherapeutic agents in solid tumors. In summary, we present innovative bioinformatics methods to develop genomic markers for cancer therapeutics and we identify a novel biological problem in cancer therapeutics using translational research methods.