Liverpool Centre for Mathematics in Healthcare

EPSRC Seminar - Discovery Data Analytics for Large Population Studies of Brain Diseases - Friday 5th May 2017

Speaker: Professor Paul Matthews, EPSRC Centre for Mathematics in Precision Healthcare, UK Dementia Institute, Centre for Neurotechnology, and the Division of Brain Sciences, Imperial College London

Venue: 12pm MATH-027 Ground Floor, Department of Mathematical Science Building

Abstract: A fundamental challenge for clinical neuroscience has been to understand brain diseases in terms of their underpinning neurobiological substrates. A first highly significant, practical objective of this would be for disease nosology- the ways in which diseases are classified that expresses their relations- to be based on their underlying neurobiological mechanisms, rather than symptoms. This would allow more rational re-purposing of medicines across diseases- deriving more value from the medicines that we have.  A second objective is to allow the preclinical phases of brain pathologies – those periods before symptoms are expressed, when slowing or stopping progression would fundamentally alter what a person experiences as disease- to be identified.   This could allow treatments to be started earlier, when they can have potentially greater healthcare impact. 

 Historically, our understanding of disease mechanisms has largely been derived from end stage pathology or clinical characterisation (in a care or research setting) of people identified because of their disease symptoms.  Recent large, longitudinal population studies are changing this.  By including detailed brain phenotyping using imaging and other tools, they promise to provide the “missing” link between the preclinical phases and clinically expressed brain disease.  Population based studies such as UK Biobank offer opportunities for joint analyses of data from large numbers of people concerning their genetics, heritage, development, occupations and exposures, nutrition, lifestyles and clinical histories.  They also pose major challenges for discovery data analytics! New approaches need to address data reduction multiple dimensions in ways that preserve key information.   Frameworks for parameterizing and comparing alternative strategies appropriate to the problem being addressed are needed.  Ways of accurately expressing the confidence with which relationships are defined need to be developed.  Part of this challenge lies in determining performance criteria (“success”) appropriate for discovery.  Addressing these challenges will have major impacts arising from their potential to extend understanding of brain mechanisms, better characterize disease risk factors and, ultimately, to enable diagnostic or treatment decisions.  In my discussion, I will highlight challenges and provide examples of some approaches being taken.  I will illustrate the opportunities with a few examples of novel insights already emerging from these large datasets.

Nie L, Matthews PM, Guo Y. 2016. Inferring Individual-Level Variations in the Functional Parcellation of the Cerebral Cortex. IEEE Trans Biomed Eng 63(12): 2505-2517.

Miller KL, Alfaro-Almagro F, Bangerter NK, Thomas DL, Yacoub E, Xu J, Bartsch AJ, Jbabdi S, Sotiropoulos SN, Andersson JL, Griffanti L, Douaud G, Okell TW, Weale P, Dragonu I, Garratt S, Hudson S, Collins R, Jenkinson M, Matthews PM, Smith SM. 2016. Multimodal population brain imaging in the UK Biobank prospective epidemiological study. Nat Neurosci 19(11): 1523-1536.

Matthews PM, Hampshire A. Clinical Concepts Emerging from fMRI Functional Connectomics. Neuron. 2016 Aug 3;91(3):511-28.