The Turing-Roche partnership is excited to formally announce its first research projects.The partnership has awarded funding for two 18-month projects in the area of ‘Structured Missingness’ to Dr Robin Mitra and Professor Ginestra Bianconi and their teams.
The Turing-Roche strategic partnership, launched in June 2021, is a five-year collaboration between The Alan Turing Institute and Roche, aiming to develop and apply advanced analytics methodologies to generate insights from multiple types of healthcare data.
The partnership chose ‘Structured Missingness’ as its first research theme. Roughly defined as missing data that is structured in some way, this research challenge has not been systematically studied and yet is an increasingly common problem that is emerging in many fields that leverage ever-growing volumes of heterogeneous data. It can significantly complicate analysis and is particularly challenging in healthcare contexts where multiple datasets (and data modalities) are often combined at scale. Consequently, there is a need to develop rigorous ways to understand structured missingness and develop tools to handle it appropriately in order to enable the generation of novel insights. You can find out more in this blog written by the partnership community.
At the end of 2021 the partnership ran three collaborative workshops to convene a diverse group of researchers to discuss this theme and scope out approaches to structural missing data, whilst formulating substantive and collaborative research questions that merited further exploration. Read a full report on the workshops. An associated funding call was then opened and after a competitive process our two successful applicants were Dr Robin Mitra from UCL and Professor Ginestra Bianconi from Queen Mary University of London. Both projects will focus on concrete examples of structured missingness including utilising longitudinal patient level data for tens of thousands of patients
Dr Mitra and his co-researchers Professor Ana Basiri (University of Glasgow) and Assistant Professor Eleni-Roaslina Andrinopoulou (Erasmus Medical Center Rotterdam) will be developing a Bayesian Modelling framework to appropriately handle structured missingness- using the common imputation approach but only imputing missing values when it is possible to do so, and utilising relationships between missing data to inform imputation. They will also be investigating the suitability of existing software packages for dealing with missing data, as well as developing novel open-source software solutions.
Professor Bianconi and her co-investigator Dr Ruben Sanchez Garcia (University of Southampton) will be taking a Network Science approach- developing a data driven mathematical and computational platform to extract and analyse the geometrical organisation of patterns of missingness in complex, heterogeneous datasets. This approach will use network analysis tools to uncover hidden organisational structure, assess the degree to which patterns of missingness are structured, and the extent to which missingness depends on the available information about a patient.
The projects will be starting in September and taking a collaborative approach. They will be sharing regular updates through the partnership Slack Workspace and knowledge share events, as well as open publication of methods, algorithms and results.
Ryan Copping, Global Head of Data Science Acceleration, Product Development at Roche and lead for the partnership said ‘We’re thrilled to fund our first projects and kick off our research arm of the partnership. Structured missingness is a key methodological challenge and tools to address this will allow us to utilise our large-scale multi-modal data sources to unlock new scientific insights’.
Ben MacArthur, Deputy Programme Director for Health and Medical Sciences at The Alan Turing Institute and deputy lead for the partnership from the Turing side added ‘It’s exciting to see the this collaborative academic-industry partnership start to take shape – we’re delighted to fund leading academics who will be utilising Roche data in their projects and working collaboratively with The Alan Turing Institute’s health programme and academic network.’