Research

We drive machine learning and pathology innovations for a more accurate prediction of cancer survival and treatment response. Our aim is to design impact-driven strategies and AI engines to accelerate the applications of AI and data science to pathology. This will enable accelerated, rigorous and actionable approaches for building data science capacity with translational impacts in pathology.

We focus on understanding tumors as complex, evolving ecosystems that may govern the evolutionary trajectory of cancer and its clinical progression. By drawing on the rich geospatial information in pathological tumor sections to complement integrated omics, our work has elucidated the complex, adaptive landscape of the tumor ecosystem and inferred immune escape strategies of cancer cells. Key insights from this research have provided the foundation for innovative clinical tests and treatment strategies that consider intra-tumor heterogeneity and incorporate cancer evolution dynamics.

Innovative integration of pathological images with next-generation sequencing can offer new insights into why some cancers are so difficult to treat. Darwinian evolution is fundamental to cancer biology and the main driver of drug resistance. However, it is often underexplored in the process of drug and biomarker development. By integrating the microenvironmental spatial context from pathology and cancer evolution inferred from omics, we direct novel data analysis to gain deeper insights into the complex biological processes that govern cancer evolution in unprecedented detail. 

As part of the Cancer Research UK-funded landmark program in lung cancer TRACERx, we develop deep learning systems for the analysis of spatial heterogeneity in the lung tumor microenvironment. TRACERx lung is a prospective, pioneering study of lung cancer evolution. Our recent progress on AI, histology, and cancer evolution in this study was published in Nature Medicine.