Receiving my PhD in 2009, I started my career as a data scientist in cancer research, where the ever-increasing volume and complexity in clinical and biological data calls for novel data analytical approaches to meet the unmet need for personalized precision therapy.

  As the lead statistician in National Biomarker Centre, CRUK, my research focused on tackling data science challenges in the development and deployment of cancer biomarkers, which, broadly speaking, comprises three areas. (1) Patients generate large quantify of longitudinal data via clinical assessments, imaging and biochemical tests during their treatment. These data have not been used in full to inform optimal treatment regime. I am developing novel Bayesian and machine learning models to integrate these data and use a joint modelling framework to identify novel biomarkers that predict treatment response and acquired resistance. (2) High throughput -omics technologies allow systematic investigation of tumorigenesis at multiple molecular and cellular levels. I am developing statistical and machine learning approaches to integrate multi-omics data, and strive to understand the cross-talking among these -omics levels. This knowledge will help understand the onset of tumour growth and can be used for the early detection of cancer occurrence. (3) Novel biomarkers need to be validated in clinical trials before being applied in clinical practice. Typically, biomarker validation trials involve development of clinical rules based on which the biomarkers will be applied, estimating clinical benefit in patients’ treatment outcomes, and understanding trade-off within biomarker-guided treatment triage. It needs more complicated designs than the conventional trials validating treatment efficacy. I am using the concept of Bayesian adaptive trial design to facilitate the translation of novel biomarkers from bench to bed.