Physical Sciences for Health CDT
Thesis project - "Development of diagnostic technology for differentiating sterile and non-sterile inflammation"
Professor Peter Tino, School of Computer Science
Professor Paula Mendes, School of Chemical Engineering
Professor Janet Lord, Institute of Inflammation and Ageing
Traumatic injuries are a major cause of death and disability worldwide. It has been estimated that in 2012 injury resulted in the death of over 500,000 people across Europe. In England and Wales trauma is the number one cause of death amongst children and adults under 44 year of age. The category 'traumatic injury' covers damage to the body caused by physical means as opposed to infection which is caused by pathogens. Traumatic injuries trigger the same inflammatory response from a patient’s immune system as does infection. Inflammation caused by pathogens is called non-sterile inflammation whereas if pathogens are not present it is called sterile inflammation.
The current methods for detecting infection are based on culture techniques and can take days to provide a diagnosis. Trauma patients and any patient undergoing surgery are at a very high risk of developing an infection and this is a major cause of death, prolonged hospital stays and high healthcare costs. Because of these factors, when patients exhibit signs of inflammation they are prescribed prophylactic antibiotics. This has its own risks as the antibiotics can damage the patients’ gut bacteria leading to further health problems. The unnecessary use of antibiotics also increases the incidence of antibiotic resistance and is a needless expense. It is predicted that by the year 2050, antimicrobial resistance will cost the world $100tn (as a cumulative GDP loss from 2014) and be responsible for 10 million deaths per year. The ability to quickly distinguish between sterile and non-sterile inflammation could help save lives and will allow for a more economical and appropriate use of antibiotics, thus saving money and reducing the development of antibiotic resistance.
Patient metabolomics data (data about the types and amounts of chemicals patients are producing) could hold the key to identifying which patients are suffering from an infection. Statistical tests have been performed to identify chemical markers (biomarkers) indicative of infection but these approaches struggle with problems associated with metabolomics data (e.g. missing data). Machine learning techniques have the potential to overcome some of these problems.
Should promising biomarkers be found, sensors can be developed to detect them. Novel techniques have been developed which imprint binding sites on surfaces at a molecular level. These binding sites can be highly selective for specific biomarkers. Using a molecularly imprinted surface as a sensor has the advantages of it being easily washed, reused and the surface can have multiple (different) recognition sites which would allow groups of biomarkers to be targeted.
This project has the following aims:
- Develop a software system which analyses data to identify biomarkers which can be used to differentiate between sterile and non-sterile inflammation. The software will use machine learning algorithms where problems of missing data and small datasets have been tackled before but not for this problem domain. Understanding the biological problem will help in deciding upon techniques to use and in designing new solutions. Existing data will be used to test the developed algorithms.
- Design and test surfaces which will specifically bind single biomarkers.
- Design and test surfaces which will specifically bind a group of biomarkers. Spectroscopic, microscopic and other analytical techniques will be used to check the surfaces have been produced correctly and work as expected.
- Validate the surfaces created using clinical samples.
Tests using clinical samples will validate the technology used. The overall goal of the project is to create a sensor which can accurately diagnose trauma patients who have non-sterile inflammation.