The ability to dissect heterogeneity in colorectal cancer (CRC) is a critical step in developing predictive biomarkers. The goal of this study was to develop a gene expression based molecular subgrouping model, which predicts the likelihood that patients will respond to specific therapies.
Using microarray data compiled from 848 CRC patients, we developed a subgrouping model based on 23 activated oncogenic pathway expression signatures. Stability of the model was validated in two independent data sets. To test the capacity of our model to predict patient response to specific targeted therapies and to determine the sensitivity of the subgroups to chemotherapy in vivo, we treated patient-derived colorectal cancer explants (PDCCEs) predicted to have high mTOR pathway activity with everolimus or oxaliplatin. Results: A molecular profile of the discovery dataset revealed 8 molecular subgroups of colorectal cancer (MSCC). Both validation sets also exhibited 8 MSCC and demonstrated significant similarities to the discovery dataset. MSCC predicted to have active mTOR pathway signaling were more sensitive to everolimus compared to MSCC predicted to have low mTOR pathway activity (p=0.016). Furthermore, MSCCs displayed differential sensitivities to oxaliplatin (p<0.05).
The ability to predict oncogenic signaling pathway activation is a powerful tool that may be used to sub-classify CRCs into clinically relevant MSCCs. Patterns of pathway activation may be used to target specific therapies in CRC patients and identification of sensitivities of MSCC to specific drugs can be a powerful approach to guide therapy for CRC patients.
Overall design
133 (non-replicate) primary and metastatic colorectal cancer tissue samples were used in this analysis.
Less...
More...