I am a Sir Henry Wellcome Fellow at the Institute of Biomedical Engineering and Big Data Institute at the University of Oxford. I am interested in modelling multi-scale protein functions to understand how functions at the molecular level propagate to cellular, tissue and organ levels. The activity of proteins is highly dependent on cell context. Therefore, it is crucial to incorporate cell microenvironment and morphology through the use of multi-modal imaging techniques. This requires developing innovative quantitative imaging and analysis methods for modelling biological behaviour at cellular, tissue and organ levels as well as integration of heterogeneous datasets.
While at Oxford I pioneered the concept of knowledge-driven machine learning. The only way we can harness the power of big biomedical data is through the development of intelligent systems that systematically incorporate existing biological knowledge. Such systems can guide machine learning approaches to discover new patterns in such rich datasets. I already demonstrated that such an approach can facilitate the discovery of context-dependent gene functions when applied to large scale genetic screens. Compared to previous methods, knowledge-driven machine learning enabled the extraction of far more information from each dataset. I also developed multiple deep learning and image analysis algorithms for classifying microscopy imaging data and profiling biological phenotypes at subcellular, cellular and tissue levels.
I did my PhD at the Institute of Cancer Research in London with Prof Chris Bakal. While at the ICR I developed methods for integrating phenotypic data with gene expression, modelling of the relationship between cell signalling and its context, and modelling the dynamics of cell morphogenesis. In these studies, I discovered new links between cell shape and breast cancer progression.
I am also passionate about data visualisation and science communication. I devised PhenoPlot, one of the first tools that is specifically designed for visualising phenotypic data. This method facilitates the interpretation of high dimensional data by generating pictorial representations of cells based on hundreds to thousands of measurements.
We show for the first time that metrics of tissue topology in scratch assays allow better characterisation of cellular mechanisms underlying perturbation effects on collective cell migration.
We developed a computer vision algorithm for profiling different features of endothelial networks including connectivity, symmetry and complexity. The effects of 1280 drugs from the LOPAC library on endothelial network formation were investigated. This lead to better characterisation of drugs mechanism of actions and their roles in angiogenesis
KCML is an intelligent system that allow inference of systematic Inference of multi-scale gene funcitons from genetic screens. It annotate genes with GO terms based on phenotypic similarity. As these annotations are data driven they provide more tissue type-specific gene funcitons.
Development of an image-omic pipeline for inference of signaling networks linking the shape of breast cells to their transcriptional activities. Through this pipeline we identified genes that are predictive of the outcome of breast cancer patients.