DERI Seminar with Jamie Timmons
When: Thursday, April 14, 2022, 11:00 AM - 12:00 PM
Where: remote
Speaker: Jamie Timmons, WHRI
Please join us on the 14th April when James A Timmons, from the WHRI, will present on Classification models applied to Age-related Diseases: The data label problem
Zoom Link: https://qmul-ac-uk.zoom.us/j/81148100921
Title: Classification models applied to Age-related Diseases: The data label problem
Abstract: Aging in humans is associated with an increased prevalence of a number of conditions, including cardiovascular, neurodegenerative and Type 2 Diabetes (T2DM, >95% of all diabetes). Deep phenotyping, involving extensive molecular analysis of blood/tissue and clinical and physiological profiling, promises to herald a new era in precision medicine. Statistical modelling can be used to develop biomarkers of disease status and progression, or study disease mechanisms directly in humans. In particular, transcriptomics is a powerful and reproducible laboratory method to study gene–environment interactions, a critical component of precision medicine that neither genetics nor non-invasive monitoring alone can fully address. I will provide some examples of our work applying transcriptomics to human phenotyping and age-related disease. I will discuss the relative ease of developing classification models for human physiological phenotypes that are components of aging. I will contrast this with the fact that disease labels e.g. Alzheimer’s Disease or T2DM, are increasingly recognised to be problematic for the purpose of supervised machine learning, and how this has practical implications for the development precision medicine technologies.
Short Bio: For twenty years James has driven biomedical projects in the fields of cardiovascular disease; diabetes and ageing. Most of his work is with complex age-related human disease where, as the title suggests, there is a question mark over the usefulness of traditional medical labels, when training models. This is predominantly because all older people have multiple diseases simultaneously. James has worked extensively in industrial research (drug discovery, cardio-metabolic disease and biomarkers); he has also worked as a senior academic for ten years where he published over 90 research articles and raised £10m in academic grants (including EU, MRC, BBSRC and NIH) to study the molecular aspects of neuromuscular aging and cardio-metabolic disease and the interaction with exercise. James current research relies on a set of novel clinical transcriptomic resources (>3,000 individuals, global transcriptomics human muscle, adipose and blood, treatment-responses) which enables the modelling of insulin resistance, human aging and cardiovascular disease; using machine learning approaches to develop biomarkers, drug-repurposing assays, study regulatory mechanisms, and develop predictive models of clinical outcome.