The focus of the group is applying and developing new computational methodology with the purpose of understanding genetic influences in disease development, the role of genomic instability, cell state transitions and the contribution of the immune system to disease progression. Our ultimate goal is to develop new strategies for early cancer detection and/or treatment management.

We have previously shown that mutational footprints associated with extrinsic and intrinsic risk factors of cancer can be used to inform aetiology and treatment options in oesophageal adenocarcinoma (Secrier, Li et al, Nat Genet 2016). Specifically, we uncovered three distinct subtypes of this cancer with distinct therapeutic opportunities, including a subgroup with high mutational burden, which could benefit from immunotherapy, and one with prevalent DNA damage defects in the homologous recombination pathway, for which PARP inhibitors may be effective (Fig 1).

We are currently investigating further mechanisms of mutational signature development and their involvement in cancer progression.

Genomic instability is a major determinant of cancer development. We are interested in gaining insights into the timing and mechanisms of processes such as catastrophic events, whole-genome doubling and copy number variation leading to gene dosage-mediated regulation of pathway activity. Ongoing efforts in the group are focused on complex rearrangement events we have previously described in oesophageal adenocarcinoma (Fig 2).

We are interested in developing new methodology for evaluating cancer-immune cell interactions and immune activity in relation to cancer-driven signals (Fig 3). Some of the topics under current investigation include immune checkpoint activity and the identification of rare cell populations in cancer.

We employ deep learning and graph-based methodology to discover cell states and phenotypes from cancer histopathology images (Fig 4). We are developing multi-omics and imaging data integration approaches that enable us to explore tumour dormancy, metastasis and the role of the tumour microenvironment in cancer progression.


Methods

We apply unsupervised and supervised learning approaches for the purpose of patient stratification and prediction of disease outcome or response to therapy. At the cellular level, we apply similar methods to understand dependencies within protein interaction networks and the effect of signal propagation within key pathways regulating oncogenic activation, senescence, dormancy, metastasis etc.

We are developing new tools for multi-omics data integration. Methods employed include: causal reasoning, predictive models, network analysis, hotspot/module identification etc.