Skip to navigation Skip to main content Skip to footer

Approved research

BioBank Behavioural BioMarkers (B4M)

Principal Investigator: Dr Aldo Faisal
Approved Research ID: 21770
Approval date: May 11th 2018

Lay summary

Ageing is an important risk factor in many common diseases and is also associated with observable differences in posture and behaviour. Behaviour provides a rich phenotype that can be collected at a largescale, and is affected by a wide range of underlying physiological changes, making it a useful integrative phenotype. However, identifying which aspects of this complex time varying phenotype are most useful remains a challenge. We aim to use machine learning to identify optimal combinations of parameters that will serve as behavioural biomarkers of ageing and disease by taking advantage of the crosssectional data available in the UK biobank. Our purpose is develop data-driven behavioural biomarkers of ageing and disease that will eventually enable us to provide behaviour-driven diagnostics, monitoring and stratification of diseases and ageing. Behavioural biomarkers are especially appealing as they can be collected using wearable sensors, such as existing smartphones and smartwatches and therefore offer a large-scale low-cost opportunity to promote health throughout society. Thus, our aims align perfectly with UK Biobank project?s goals of improving diagnosis of illness and promoting health throughout society. We will perform analysis of movement/behavioural data in the UK Biobank and employ pattern recognition algorithms to extract meaningful features of behaviour (behavioural fingerprinting). We will then correlate these features with age and disease related variables and use pattern classification techniques to develop polymarkers of age and disease based on behavioural motion patterns. We aim to to analyse the full cohort for which high-resolution accelerometer data are available (currently listed on the website as 103,711 participants). This is because we expect from our own experience to date that Biobank behaviour data is highly variable depending on context and individual differences. We aim to develop an objective marker that is defined