Statistical Learning Approaches to Latent Variable Modelling to Accelerate Endotype Discovery in Asthma and Allergic Diseases
Dr Danielle Belgrave
Imperial College London, UK (SysGen visitor: Sep-Dec, 2016)
Tuesday 22nd November
Theatre 1, Alan Gilbert Building, The University of Melbourne
The grand challenge of identifying personalised treatment and management strategies for medical conditions relies on the advancement of statistical learning methods for discovery of the subtypes of complex diseases (which may indicate disease “endotypes”) by using ‘intelligent phenotypes’. Statistical learning methods can provide a flexible framework for endotype discovery through the application of latent variable modelling to disambiguate diseases where there are heterogeneous phenomena. The Bayesian framework allows us to upscale these models in order to integrate high-dimensional longitudinal data from immune responses, genetic data and clinical data. Probabilistic programming provides a powerful tool to express such statistical learning problems. The aim of this presentation is to provide an intuitive and flexible approach to endotypes discovery using probabilistic programming. This will be presented through a series of applications of latent trajectory models to disaggregate complex evolving endotypes which will enable the discovery of clinically meaningful subgroups of asthma and allergic diseases.
Danielle Belgrave is a Research Fellow in Biomedical Modelling at Imperial College London where she leads a statistical machine learning program within the Study Team for Early Life Asthma Research (STELAR). She is the recipient of an MRC Career Development Award in Biostatistics. Her research interests span developing Bayesian machine learning, latent variable and probabilistic graphical modelling strategies to accelerate endotype discovery to understand the progression of asthma and allergic diseases and comorbidities. This research is generalizable to profiling patients with greater accuracy to allow us to move towards more personalized disease management strategies through understanding the underlying latent manifestations of disease and their distinct genetic and environmental characteristics. Danielle is Area Chair of Women in Machine Learning (NIPs workshop) and is a faculty member of the International Congress of Pediatric Pulmonology.
Before joining Imperial College, she worked in both industry (GlaxoSmithKline) and academia (University of Manchester) and continues to have active collaborations with both groups. She was awarded a Microsoft PhD Scholarship to carry out her PhD at The University of Manchester (2010 – 2013) with Profs Iain Buchan, Christopher Bishop (Machine Learning and Perception Group in Microsoft Research Cambridge) and Adnan Custovic. Prior to that, she received her MSc in Statistics at University College London and a BSc in Business Mathematics and Statistics at The London School of Economics.
Enquiries: Andrew Siebel (firstname.lastname@example.org)