Month: July 2018

Greenhouse gases – a major international research priority 

Greenhouse gases are accumulating in the atmosphere but our current understanding of their sources and sinks is poor. Robust, transparent evaluation of global greenhouse gas emissions is now recognised as a major international research priority. 

Unprecedented data volumes require new statistical techniques 

The Earth’s surface is far too large to measure its gas emissions directly. The Atmospheric Chemistry Research Group has therefore pioneered the use of statistical methods to indirectly infer spatio-temporal emissions based on a limited number of concentration measurements and atmospheric models. 

The successful launch of the TROPOMI instrument on board the Sentinel 5 Precursor satellite in October 2017 will provide unprecedented data volumes (around 105 soundings per hour) on atmospheric methane in the atmosphere. The new measurements have the potential to provide us with a wealth of information to improve emission estimates but the currently used statistical techniques will become computationally prohibitive with such large data volumes. 

Project goals 

The project is a proof of concept application of an integrated nested Laplacian approximation (INLA) to infer greenhouse gas emissions, firstly to a synthetic dataset and then to the well-understood problem of estimating methane emissions in the UK. 

This study paves the way for a much larger research proposal, being the first to use advanced spatio-temporal statistics to exploit the new abundance of Earth observation data for regional and global greenhouse gas emission estimation. 

Policy makers, industry and scientists all have a role in tackling climate change 

• Policy makers need to understand emissions at national scales so that they can enact appropriate regulation and ensure that protocols are being followed.
• Industry representatives need to know the impact that their activities have on the global climate.
• Scientists hope to improve their understanding of the global carbon cycle, which will allow for more accurate predictions of future climate change.

Leading to bigger things 

The project was successful in demonstrating that we can use large data sets to infer methane emissions using INLA. This means that we will be able to take advantage of new large data volumes offered by the TROPOMI instrument. 

The work also led to a result that improves existing top down estimates of regional greenhouse gas emissions using novel spatio-temporal statistics. 

With the ability to handle these large data volumes, we aim to move up to a wider project estimating global methane emissions with the detail currently only possible on a regional scale. 

Blog written by Luke Western, Research Associate, School of Chemistry.  

Acknowledgements

Zhe Sha, Research Associate in Sea Level Research (Computational Statistics), School of Geographical Sciences. 

This project was funded by the Jean Golding Institute Seed Corn Funding Scheme 2018. To find out about other projects supported by this scheme, take a look at the Jean Golding Institute Projects. 

Psychosis 

Psychosis defines a category of serious and long-term mental disorders, like schizophrenia, characterised by loss of contact with reality and symptoms such as hallucinations, delusions and thought disorder. Psychosis is extremely expensive to treat and one of the leading causes of disability worldwide. 

A data-driven method 

We are working on a data-driven method to assess non-verbal synchrony, which is a core component of social functioning and social cognition. Our approach is based on a set of socio-motor biomarkers measured during a joint-action task, where the participant mirrors the movement of a computer avatar. Work carried out in our group reports that such biomarkers can accurately discriminate between patients with chronic schizophrenia and controls. Our proof of concept work demonstrates that data-driven approaches developed in interdisciplinary collaborations between psychiatric epidemiologists and mathematicians have the potential to address healthcare challenges. We are now using our movement-based test in a group at risk of or with newly diagnosed psychosis because they will be largely free of side effects due to chronic illness and/or medication. By further refining our methods and algorithms for detecting subtle differences in movement and coordination we hope to create an individual cognitive deficit-signature which could be used for designing individually tailored therapies. 

Pilot study 

This project is a pilot study to investigate the use of social movement differences as a diagnostic aid for psychosis, using an interactive virtual reality avatar. Participants are asked to take part in a mirror-movement task which assesses interactional synchrony. This project involves participants from the Bristol Early Intervention for Psychosis team. 

Virtual Reality Avatar 

We are using a virtual reality avatar to find out if people with first episode psychosis or at risk of psychosis perform differently at a mirror movement task, compared to people not at risk of psychosis. 

Looking ahead 

A new approach to psychosis prediction using data driven tools would add useful information to current methods (psychometric interviews and clinical judgement). More accurate prediction tools would improve outcomes for patients but would also help to allocate NHS resources more effectively. Furthermore, quantitative approaches may reveal new aspects of psychotic aetiology. 

Blog written by Sarah Sullivan, Research Fellow, Centre for Academic Primary Care and Centre for Academic Mental Health 

Also working on this project: Piotr Slowinski, Research Fellow, College of Engineering, Mathematics and Physical Sciences, University of Exeter. 

This project was funded by the Jean Golding Institute Seed Corn Funding Scheme 2018. To find out about other projects supported by this scheme, take a look at the Jean Golding Institute Projects. 

 

This project initiates a new collaboration between biology and applied mathematics to apply network analysis methods to a large and unique data set of immune measures of wild mice populations. 

Our aim is to understand the network relationship among the immune measures and to compare it between wild and lab mice. 

Immune systems of animals – wild versus lab 

Animals make immune responses to protect them from harmful infections, and in this way immune responses contribute to animals’ evolutionary fitness. Almost everything we know about mammalian immune responses has come from studies of laboratory animals. However, recent work has shown that wild mice immuno-regulate their responses differently than laboratory mice. How this is done is unknown. 

Key questions 

• What are the network relationships among these immune measures?
• How do the network properties differ between wild mice and lab mice?

A data set of immune function in wild mice in Bristol area 

In our previous study, we assembled a data set of 131 immune response measures of 450 wild mice from the local area. 

Network of immune measures 

We constructed networks of the 131 immune responses, where a node is an immune measure and edges are formed when two immune measures are similar across mice. 

This process was done for wild mice and lab mice separately. Then we carried out two types of analyses: 

1. Detection of network communities to know how nodes can be grouped into blocks.
2. Centrality, i.e. the importance of each node in the network.

Community structure 

We used two community detection algorithms: stochastic block model and modularity maximisation, to identify how the immune network can be divided into blocks of interrelated measures. 

Centrality 

To analyse the role of each immune measure, we calculated the degree and the eigenvector centrality measures. 

Outcomes 

• Similar structures are observed when community detection algorithms are used in different wild mice networks, showing robust results.
• Wild mice and lab mice networks show different community structure and different behaviours against network structure stress tests. 

Plans for extension 

We aim to study different network structures, such as core-periphery and different centrality measures, as well as how networks change among mice of different state (age, sex, etc.) and geographical location. 

Blog written by Naoki Masuda, Engineering Maths at the University of Bristol 

Acknowledgements: Mark Viney, School of Biological Sciences, Elohim Fonseca dos Reis, Engineering Maths 

This project was funded by the Jean Golding Institute Seed Corn Funding Scheme 2018. To find out about other projects supported by this scheme, take a look at the Jean Golding Institute Projects. 

 

The role of antibiotic use in animals, particularly those in the food chain, is highlighted as an area needing focus and research to drive policy change. It is therefore essential to understand the true nature of antibiotic use across animal health if we are to undertake informed actions to reduce unnecessary prescribing of antibiotics and reduce the pressure for bacteria to become resistant to these medicines. To achieve this goal, it is vital to develop a comprehensive and accessible databank of veterinary data that includes information on antimicrobial use (AMU) and antimicrobial resistance (AMR).

Motivation 

Our team is leading the UK veterinary science community by developing the first platform for livestock AMR research. We are drawing together information collected on cattle from many different sources – veterinary records, milk retailer data, findings from research studies and diagnostic laboratory results – into a database. We have access to surveillance data on AMU for over 50% of cattle in the UK.

Before the start of this project, following a series of successful AMR projects, we were already collecting data on AMU and AMR from > 100 UK dairy farms concentrated in South West England. However, we needed to extend this coverage to include cattle farms throughout the UK by collaborating with livestock diagnostic laboratories. Thus, the aim was to strengthen existing collaborations and foster new ones with a number of UK livestock diagnostic laboratories, to ultimately integrate their AMR data into our database.

Key objectives achieved 

• Data sharing agreement signed with a major UK diagnostic laboratory  

We signed a data sharing agreement with Farm Animal Diagnostics – Scotland’s Rural College (photo above) and integrated their AMR results for cattle into our growing database. Doing this threw up issues around data security, identifiability and ownership, especially in the wake of updates to the General Data Protection Regulations (GDPR). Because of these issues, we sought and were able to second a Senior Research Associate (Eleanor Walsh) to help us with the research governance of the database (including the ethical aspects). We have spent considerable time investigating GDPR and the impact it is likely to have on collaborators’ ability to process and share data for research purposes, which is still ongoing.

•  AMR coverage extended by strengthening our collaboration with Farmvet Systems 

Farmvet Systems (FVS) is the UK leading supplier of cloud-based data management technology for farm animal veterinarians and farmers. Thanks to this project, we were able to work closely with FVS to link their veterinary prescription data pulled out from several clinic management systems and their animal-specific indexed laboratory results captured by veterinarians/farmers participating through FVS’s smartphone apps. This work has allowed us to extend the coverage of our AMR data.

• Pilot data on AMR generated and used to support a grant application 

We used descriptive statistics to summarise the newly integrated AMR data imported because of this project. Interestingly, data contained almost 4000 potentially zoonotic Enterobacteriaceae, more than half of which are AMR. The vast majority of these are multi-drug resistant E. coli, the largest bacterial killer of humans in the UK. These and other results from the project have been used to support a full Wellcome Trust (WT) Biomedical Resource and Technology Development grant application, as well as the rebuttal to our WT reviews. The committee evaluating our proposal is meeting in July this year, so wish us luck!

Looking ahead 

We have plans to enlarge our current databank with broad and representative coverage of AMU and AMR across the UK cattle industry, with a view to expand to other animal species and human data to ultimately construct the world’s first comprehensive ‘research ready’ One Health resource for AMR.

Blog written by Fernando Sánchez-Vizcaíno, Kristen Reyher, Andrew Dowsey, Jon Massey, (Bristol Veterinary School) and Eleanor Walsh (Bristol Medical School), University of Bristol

This project was part-funded by the Jean Golding Institute Seed Corn Funding Scheme 2018. To find out about other projects supported by this scheme, take a look at the Jean Golding Institute Projects.

On 3rd July the Jean Golding Institute invited hundreds of guests from academia, public and industry sectors, as well as the community to come and join this inaugural showcase event hosted in the historic Wills Memorial Building at the University of Bristol. 

The event was a manifestation of the outputs of two years of work since the Institute inception in 2016. The showcase set out to celebrate the work and achievements of the JGI community who are working towards creating an environment where novel data-intensive research areas are explored and developed, and interdisciplinary exchanges promote a spirit of collaboration. 

We held a series of talks throughout the day as well as keynote speeches that reflected the advances of data science and the impact in society. We heard from current and previous directors of the Institute, project leads and were pleased to welcome a number of external collaborators to talk, including those from the Office for National Statistics and from Cardiff University.

At the Jean Golding Institute, we take inspiration from the pioneering work of Jean Golding, who broke down barriers between research domains and created integrated multidisciplinary teams with a shared vision. We were privileged to be joined by Jean who talked about her earlier experiences using a ‘punched card system’ for data processing. Jean said, “I have always loved, and been intrigued by data, and to have my name attached to an Institute that has the important remit of developing and illustrating ways in which data can be manipulated and analysed is a great honour.” Jean is the founder of ‘Children of the 90s’ (ALSPAC), one of the best characterised longitudinal cohorts in the world.

A wide variety of projects and data competitions which use data driven solutions to society challenges were displayed in our exhibition area. Many of these projects were supported by our annual JGI seed corn funding scheme and we felt immensely proud to have sponsored them and watch them as they mature and become independently sustainable. 

The showcase included interactive experiences like our first virtual reality project exploring the ‘Secrets of Brain Health’. This is a collaboration between researchers in the Clinical Research and Imaging Centre (CRIC) and BDH Immersive.  We also showcased a collaboration between MRC IEU and the company Free Ice Cream, who have produced a game engine that allows individuals to explore and interact with the complex network of human health conditions and disorders.

Two world-renowned data scientists Trevor Hastie, Professor of Statistics at Stanford University and Dan Crichton, Principal Investigator at NASA Jet Propulsion Labs were a focal part of our event.  

In his keynote talk, Dan described challenges that researchers at NASA are facing, which he considers apply universally: management of large, complex, distributed data sets and an effective exploration of such data to generate new knowledge. Dan proposed that these challenges “originate the need for a new scientific methodology”. 

Trevor’s talk focused on supervised learning and building models from data that predict an outcome using a collection of input features. The current challenge of big data is that the analysis starts with big datasets with many features and the experts use machine learning algorithms to determine the key features. Big data also varies in shape, and this calls for different approaches. Trevor advised researchers to think outside the box! How much accuracy is needed? How about timelines? And encouraged analysis on subsets of data if possible. Finally, he described new methodologies using programming languages such as R for statistical computing.  Both Trevor and Dan’s presentations will be available on the JGI website over the summer.

Prof Hugh Brady, Vice-Chancellor and President of the University of Bristol closed the showcase with “the Jean Golding Institute is at the centre of a momentous transformation at the University, since its inception the JGI has become a hub for data science, it has built new networks and extended existing ones, both within the University and between local and national organisations. The Jean Golding Institute will serve as a conduit between the ATI, the University and external partners and will play a key role in tackling some our most pressing global issues including urban analytics, security and health”. 

We would like to say a big thank you to everyone involved in our first showcase and those that sponsored our big day. The atmosphere was one of collaboration and optimism for the future, and since joining the Alan Turing Institute (ATI), Bristol has great opportunities to lead the advancements of data science and artificial intelligence in the UK as part of this prestigious consortium.

Blog written by Patty Holley, Manager and Liz Green, Coordinator, Jean Golding Institute