Integrated analysis of high-throughput molecular and metabolic profiles to develop pancreatic ductal adenocarcinoma  subtype-specific therapy

Overall survival of pancreatic ductal adenocarcinoma (PDA) patients is less than 6 months from the time of diagnosis. Currently, patients with advanced or metastatic diseases are treated with gemicitabine, and have only a modest increase in survival. These attributes may reflect the variable and often disappointing responses seen when deploying therapeutic agents in unselected PDAC populations, despite occasional significant responses. Studies in other solid tumours have shown that heterogeneity in therapeutic responses can be anticipated by molecular differences between tumours, and targeting drugs specific to tumour subtypes in which they are predicted to be selectively effective can indeed improve treatment. Seeking to extend this new paradigm, we recently reported three gene expression subtypes of PDA named as classical, quasi-mesenchymal; QM-PDA and exocrine-like PDA using a gene expression signature (62 genes; designated as PDAssigner; Collisson and Sadanandam, et al. Nature Medicine, 2011; co-first author). Interestingly, patients with classical tumours fared better than patients with QM-PDA tumours after resection. We also observed that QM-PDA subtype cell lines are, on average, more sensitive to gemcitabine than the classical subtype lines. The opposite relationship is observed with erlotinib. Along this line, we are interested in characterising the distinct metabolic, genetic and cellular phenotypes of PDA subtypes and their influence on drug responses (precision and personalised medicine) involving wet-lab and bioinformatics by integrating high-throughput molecular and metabolic profiles and correlating the mixed signatures to that of the therapeutic responses.

Characterising colorectal cancer subtypes and integrated analysis of molecular profiles to identify precise therapies

Colorectal cancer (CRC) is a heterogeneous disease that is traditionally classified based on genomic (microsatellite, MSI; or chromosomal instability, CIN) or epigenomic (CpG island methylator phenotype, CIMP) status. In order to achieve a robust and clinically useful means of classification, we performed a novel combination of consensus-based unsupervised clustering of gene expression profiles from patient tumours (n > 1000) to find subtypes within these samples. In total, we identified five integrated CRC subtypes with differential gene expression signatures and prognosis. Namely, we predicted and validated the cellular origin of our subtypes and associated this and the drug responses in order to guide cellular signalling pathway- and mechanism based therapeutic strategies that target subtype-specific tumours. In addition, we also associated our subtypes with (i) MSI status, (ii) Wnt signaling pathway activity, (iii) metastasis to distant organs and (iv) response to targeted and chemotherapy (Sadanandam, et. al., Nature Medicine, 2013). The personalised response of the subtypes to targeted- or chemo-therapy were validated using cell lines in vitro and mouse (xenograft and genetically engineered; cross-species analysis) models in vivo. We will use systems biology approach to extend the characterisation of CRC subtypes in order to facilitate personalised medicine for this devastating disease. In addition, we are interested in understanding cetuximab- and anti-angiogenic therapeutic agents-based adaptive drug resistance in colorectal cancer.

Developing assays using gene signatures that distinguish different subtypes in the clinic

Assigning individual patients to different molecular subtypes require assays that can be used in the clinic. We have developed an exploratory RT-PCR and immunohistochemistry assays that distinguish different subtypes of CRC. Currently, we are interested in further improving these assays and also, developing novel assays involving nCounter platform (Nanostrings Technologies).

Characterising consensus tissue-independent molecular subtypes from different epithelial cancers

We have recently identified subtypes using multiple epithelial type cancers that are independent of tissue specific genes. These subtypes were found to have differential drug responses. We are interested in further characterising these subtypes.