91̽»¨

Aims and objectives

Our aim is to use artificial intelligence approaches to identify and follow clusters of multimorbidity to improve the design, testing and selection of more effective preventive interventions. The main implication and novelty of a cluster medicine approach is that instead of considering one cause – one treatment for one disease, it considers causes, and thus possible prevention and treatments, for the entire cluster of long term chronic conditions. 

Using state-of-the-art machine learning approaches (e.g. agglomerative clustering and language modelling) we aim to identify clusters of multimorbidity, their trajectories as well as possible predictive biomarkers and risk factors using health data from a variety of sources. We aim to apply new natural language processing approaches we have developed for sequence discovery and representation learning to validate the clusters and augment the data allowing the identification of new rare, indirect clusters. We aim to establish causal links between patient characteristics (e.g. age, comorbidities, gender, ethnicity), potential drivers of the clusters (e.g. inflammation) and the evolution of multimorbidity.

---

Our expertise

Machine learning

91̽»¨ has an excellent track record in machine learning, in the development of unsupervised and reinforcement learning models, and their application to neurodegenerative and cardiovascular diseases. It also has a strong international reputation in Natural Language Processing, hosting one of the largest NLP groups in Europe and one of the most successful in the UK. Its internationally-recognised natural language processing and speech technology research has a track record spanning more than 30 years and supported by over £45m of UK and European funding since 1990 and includes most aspects of NLP research, such as language learning, machine translation, text simplification, social media analysis and information extraction. 

This group is supported by biologists experts in ageing and clinicians and has an international reputation in social gerontology, in the mining of data on deprivation and health, and the impact of social and behavioural factors on the prevalence of disease and tools which allow statistical analysis and modelling of the onset of multimorbidity. The group also has expertise in information governance and management of large national datasets, including NHS Digital, to address multimorbidity questions.

We work closely with 91̽»¨ Teaching Hospital NHS Trust, Rotherham Foundation NHS Trust and .

---

Our capabilities

High performance compute clusters and GPU clusters

Our computational infra-structure includes 

• Two High Performance Compute (HPC) clusters at the University of 91̽»¨ ( and )
• 91 individual rackmount servers, of which at least 11 are GPU servers
• One major HPC grid of 25 nodes with up to 32 CPU cores, 256GB RAM per node, and around 28 GPU cards spread across various nodes
• One smaller 4-node grid and a 5-node Cloud cluster used to provide IaaS
• 6 DAS and NAS RAIDs of between 10TB and 40TB each, a Lustre file system running on an Infinband backbone of 160TB for the cluster, and a Ceph filesystem of around 300TB.

The University also has 1000+ managed desktops, some of these are high performance machines and some are also equipped with GPUs. In addition, we have access to external Tier 2 GPU clusters

(based in Durham)

(University of Oxford)

These clusters are available for specific machine/deep learning projects that require extensive GPU compute.

---

People

Name	Speciality
Professor Peter Bath	Information Science
Professor Ilaria Bellantuono	Musculoskeletal Ageing
Professor Simon Heller	Diabetes
Dr Nafise Sadat Moosavi	Computer Science
Dr Alvaro Martinez-Perez	Social Sciences
Professor George Panoutsos	Electrical and Electronic Engineering
Dr Dan Holman	Social Sciences
Dr Chenghua Lin	Natural Language Processing
Dr Carolina Scarton	Natural Language Processing
Dr Mark Stevenson	Natural Language Processing
Dr Mari-Cruz Villa-Uriol	Clinical data analytics
Professor Aline Villavicencio	Natural Language Processing
External collaborators
- Karolinska Institutet	Geriatric Epidemiology
- Karolinska Institutet	Geriatric research
- Karolinska Institutet	Getiatric research
- Queen Mary University of London	Clinical AI and Machine Learning

The group is also supported by biologists, experts in ageing and clinicians.