The DART project aims to pioneer a ground-breaking capability to enhance the performance and energy efficiency of reconfigurable hardware accelerators for next-generation computing systems. This capability will be achieved by a novel foundation for a transformation engine based on heterogeneous graphs for design optimisation and diagnosis. While hardware designers are familiar with transformations by Boolean algebra, the proposed research promotes a design-by-transformation style by providing, for the first time, tools which facilitate experimentation with design transformations and their regulation by meta-programming. These tools will cover design space exploration based on machine learning, and end-to-end tool chains mapping designs captured in multiple source languages to heterogeneous reconfigurable devices targeting cloud computing, Internet-of-Things and supercomputing. The proposed approach will be evaluated through a variety of benchmarks involving hardware acceleration, and through codifying strategies for automating the search of neural architectures for hardware implementation with both high accuracy and high efficiency.

Bayesian inference is a process which allows us to extract information from data. The process uses prior knowledge articulated as statistical models for the data. We are focused on developing a transformational solution to Data Science problems that can be posed as such Bayesian inference tasks. An existing family of algorithms, called Markov chain Monte Carlo (MCMC) algorithms, offer a family of solutions that offer impressive accuracy but demand significant computational load. For a significant subset of the users of Data Science that we interact with, while the accuracy offered by MCMC is recognised as potentially transformational, the computational load is just too great for MCMC to be a practical alternative to existing approaches. These users include academics working in science (e.g., Physics, Chemistry, Biology and the social sciences) as well as government and industry (e.g., in the pharmaceutical, defence and manufacturing sectors). The problem is then how to make the accuracy offered by MCMC accessible at a fraction of the computational cost. The solution we propose is based on replacing MCMC with a more recently developed family of algorithms, Sequential Monte Carlo (SMC) samplers. While MCMC, at its heart, manipulates a single sampling process, SMC samplers are an inherently population-based algorithm that manipulates a population of samples. This makes SMC samplers well suited to the task of being implemented in a way that exploits parallel computational resources. It is therefore possible to use emerging hardware (e.g., Graphics Processor Units (GPUs), Field Programmable Gate Arrays (FPGAs) and Intel's Xeon Phis as well as High Performance Computing (HPC) clusters) to make SMC samplers run faster. Indeed, our recent work (which has had to remove some algorithmic bottlenecks before making the progress we have achieved) has shown that SMC samplers can offer accuracy similar to MCMC but with implementations that are better suited to such emerging hardware. The benefits of using an SMC sampler in place of MCMC go beyond those made possible by simply posing a (tough) parallel computing challenge. The parameters of an MCMC algorithm necessarily differ from those related to a SMC sampler. These differences offer opportunities for SMC samplers to be developed in directions that are not possible with MCMC. For example, SMC samplers, in contrast to MCMC algorithms, can be configured to exploit a memory of their historic behaviour and can be designed to smoothly transition between problems. It seems likely that by exploiting such opportunities, we will generate SMC samplers that can outperform MCMC even more than is possible by using parallelised implementations alone. Our interactions with users, our experience of parallelising SMC samplers and the preliminary results we have obtained when comparing SMC samplers and MCMC make us excited about the potential that SMC samplers offer as a "New Approach for Data Science". Our current work has only begun to explore the potential offered by SMC samplers. We perceive significant benefit could result from a larger programme of work that helps us understand the extent to which users will benefit from replacing MCMC with SMC samplers. We propose a programme of work that combines a focus on users' problems with a systematic investigation into the opportunities offered by SMC samplers. Our strategy for achieving impact comprises multiple tactics. Specifically, we will: use identified users to act as "evangelists" in each of their domains; work with our hardware-oriented partners to produce high-performance reference implementations; engage with the developer team for Stan (the most widely-used generic MCMC implementation); work with the Industrial Mathematics Knowledge Transfer Network and the Alan Turing Institute to engage with both users and other algorithmic developers.

Vascular disease is the most common precursor to ischaemic heart disease and stroke, which are two of the leading causes of death worldwide. Advances in endovascular intervention in recent years have transformed patient survival rates and post-surgical quality of life. Compared to open surgery, it has the advantages of faster recovery, reduced need for general anaesthesia, reduced blood loss and significantly lower mortality. However, endovascular intervention involves complex manoeuvring of pre-shaped catheters to reach target areas in the vasculature. Some endovascular tasks can be challenging for even highly-skilled operators. The use of robot assisted endovascular intervention aims to address some of these difficulties, with the added benefit of allowing the operator to remotely control and manipulate devices, thus avoiding exposure to X-ray radiation. The purpose of this work is to develop a new robot-assisted endovascular platform, incorporating novel device designs with improved human-robot control. It builds on our strong partnership with industry aiming to develop the next generation robots that are safe, effective, and accessible to general NHS populations.

The current 'gold standard' for diagnosis and grading of many diseases (including most solid tumours) is largely based on an expert histopathologist's visual microscopic assessment of an extremely thin (only a few micrometers thick) section of the suspicious tissue specimen glued to a glass slide. This practice has remained more or less the same for several decades, and results in subjective and variable diagnosis. However, the recent uptake of digital slide scanners by some diagnostic pathology laboratories in the UK marks a new revolution in pathology practice in the NHS trusts, with our local NHS trust being the first one in the country to use digitally scanned images of tissue slides for routine diagnostics. The digital slide scanner produces a multi-gigapixel whole-slide image (WSI) for each histology slide, with each image containing rich information about tens of thousands of different kinds of cells and their spatial relationships with each other. This project aims to introduce a novel paradigm for analytics and computerised profiling of tissue microenvironment. We will develop sophisticated tools for image analytics in order to reveal spatial trends and patterns associated with disease sub-groups (for example, patient groups whose cancer is likely to advance more aggressively) and deploy those tools for clinical validation at our local NHS trust. This will be made possible by further advancing recent developments made in our group, such as those allowing us to recognise individual cells of different kinds in the WSIs consequently enabling us to paint a colourful picture of the tissue microenvironment which we term as the 'histology landscape'. Understanding and analysing the tissue microenvironment is not only crucial to assessing the grade and aggressiveness of disease and for predicting its course, it can also help us better understand how genomic alterations manifest themselves as structural changes in the tissue microenvironment. We will develop tools and techniques to extract patterns and trends found in the spatial structure and the 'social' interplay of different cells or colonies of cells found in the complex histology landscapes. Our goal is to establish the effective use of image analytics for understanding the histology landscape in a quantitative and systematic manner, facilitating the discovery of image-based markers of disease progression and survival that are intuitive, biologically meaningful, and clinically relevant - eventually leading to optimal selection of treatment option(s) customised to individual patients. This project will analyse real image data and associated clinical and genomics data from patient cohorts for colorectal cancer as a case study. The research staff on this project will work closely with clinical collaborators to ensure the biological significance and clinical relevance of spatial trends and patterns found in the data. In collaboration with our industrial partner Intel, we will test and demonstrate the effectiveness of our methods in a clinical setting potentially leading to better healthcare provision for patients and potential cost savings for the NHS.

We will improve patient health and medical research by maximising the use of vast amounts of human data being generated in the NHS. But there are two obstacles: (i) inter-related clinical and research datasets are dispersed across numerous computer systems making them hard to integrate; (ii) there is a serious shortage of computational expertise as applied to clinical research. As part of the UK's healthcare strategy to overcome these limitations, we have assembled a world-class consortium of institutions and scientists, including UCL Partners (containing NHS Trusts treating >6 million patients), Francis Crick Institute, Sanger Institute and European Bioinformatics Institute. Close links with the NHS (through Farr and Genomics England) will allow information exchange for health and disease progression. We have also engaged leading companies like GSK and Intel. We will use the MRC funds for two purposes: 1. Create a powerful eMedLab data centre. We will build a computer cluster that allows us to store, integrate and analyse genetic, patient and electronic health records. By co-locating in a single centre, we eliminate delays and security risks that occur when information is transmitted. Research Technologists supplied by the partners will install and maintain the infrastructure and software environment. 2. Expand scientific and technical expertise in UK Medical Bioinformatics through a Research & Training Academy. Basic and clinical scientists, and bioinformaticians will be trained to perform world-leading computational biomedical science. We will train in the whole range of skills involved in medical bioinformatics research with taught courses, seminars, workshops and informal discussion. To coordinate research activities across partners, we will establish Academy Labs, which are flexible, semi-overlapping groupings of academic and industrial researchers to share insights and plan activities in areas of common analytical challenges. The Academy will provide a mechanism for information and skills exchange across the traditional boundaries of disease types. These will enable existing projects in 3 disease domains in which we have unique strengths: rare diseases, cardiovascular diseases and cancer. Rare: We house 31/70 Nationally Commissioned Highly Specialised Services; ~0.5M of the 6M of our patients have a rare disease, including >50% of those treated at Great Ormond Street Hospital. >200 research teams generate large quantities of genetic, imaging (eg, 3D facial reconstructions), and clinical information (eg, patient records). Cardiovascular: We also lead genomic, imaging, and health informatics programmes in cardiovascular disease with contributions to projects like UK10k project and host multiple national cardiovascular registries through the National Institute for Cardiovascular Outcomes Research. These are linked to primary and hospital clinical care records through Farr@UCLP with current cohort sizes of ~2M people. Cancer: We also have particular clinical expertise in some of the most difficult to treat cancer types and we host major international data resources. These include individuals recruited to the TRACERx study of lung cancer, 8,500 women with abnormal cervical smears in whom methylation patterns of the HPV16 genome predict progression to high-grade precursor disease, and one of the largest sarcoma biobanks in the world. Ultimately, this bid will allow us to use new computational approaches to (i) link patient records and research data in order to understand the pathogenesis of disease, (ii) use genomic, imaging and clinical data to identify diagnostic, prognostic and predictive biomarkers to guide therapy, predict outcome and increase recruitment to clinical trials based on stratified populations and (iii) translate new IP by engagement with the pharmaceutical industry.