B3265 - Development of a widely applicable tool for phenome scans and data processing - 26/03/2019

B number: 
B3265
Principal applicant name: 
Louise AC Millard | MRC IEU, University of Bristol
Co-applicants: 
Dr Neil M Davies, Professor Kate Tilling, Professor Peter Flach
Title of project: 
Development of a widely applicable tool for phenome scans and data processing
Proposal summary: 

While studies examining the effect of an exposure of interest on specific health issues provides important understanding into the determinants of health and disease, it would be arguably more useful to determine, in a single study, the extent that the exposure affects a very large number of health-related traits and diseases. Phenome scans are a particular type of 'hypothesis-free' study, that test the association of a trait of interest with a comprehensive array of phenotypes – the “phenome”. They can be used with an approach called Mendelian randomization to search for the causal effects of a trait of interest with potentially many outcomes. Performing a phenome scan is typically non-trivial as the set of phenotypes available in a cohort tends to be highly heterogeneous. Recently, we developed a tool called PHESANT, that allows researchers to conduct phenome scans in UK Biobank. However, currently it is not easy to perform phenome scans in ALSPAC. In this project we will develop innovative software that enables researchers to easily perform their own comprehensive phenome scans in ALSPAC, searching across 'all' phenotypes is this cohort. We will demonstrate this software by searching for the causal effects of education, as an exemplar.

Impact of research: 
BENEFITS TO RESEARCHERS 1) Enabling researchers to easily conduct phenome scans in ALSPAC. For example, a researcher can use our software to systematically assess the causal effects of their exposure of interest on all phenotypes in ALSPAC, using MR-pheWAS. 2) Enabling researchers to perform other large scale data analytics using ALSPAC, e.g. machine learning. Our software can be used to automatically clean the ALSPAC data before learning machine learning models with these data. BENEFITS TO ALSPAC Setting up this system in ALSPAC will be of great benefit to the ALSPAC cohort, as it will enable researchers to perform phenome scans (and other large scale data analytics) hence help to maximise the impact of this world-leading cohort. We are aware that ALSPAC will move their data managment to new systems (e.g. using mongoDB) and we are keen to ensure that the software we develop is appropriate for this, to be of maximal value to ALSPAC and ALSPAC researchers.
Date proposal received: 
Tuesday, 5 March, 2019
Date proposal approved: 
Tuesday, 5 March, 2019
Keywords: 
Statistics/methodology, Hypothesis-free - we will search for effect across all phenotypes in ALSPAC., Statistical methods, Mendelian randomisation, Statistical methods