Using Artificial Neural Networks to inform the exploitation of ‘omic biomarkers in lung cancer diagnosis.

Supervisors: Prof Luis A. J. Mur ([Email Address Removed]) Dr Chuan Lu,: [Email Address Removed] (Computer Science Department); Dr. Keir Lewis; ([Email Address Removed]) Consultant in Respiratory and General Medicine, Hywel Dda Health Board.
External partner: Hwyel Dda University Health Board.

Lung cancer (LC) is the most prevalent cancer in the world, and responsible for 1.3 million deaths each year. There are several subtypes of LC with the main subdivisions being non-small-cell lung carcinoma (NSCLC) and small-cell lung carcinoma (SCLC). Over the last 30 years, the overall five year survival rate for LC has shown little improvement, with only 15% of patients living for five years or more after the date of their initial diagnosis. These relatively poor survival rates, compared to other cancers, are primarily as a result of the late detection of a malignancy, where the success rates of clinical interventions are significantly reduced. Clinicians currently rely on three main tools for LC diagnosis, namely X-ray, computerised tomography (CT) scans, and bronchoscopy. Although these methods have improved our ability to detect lung cancer, they have nevertheless failed to improve the rate of early diagnosis of LC.

The key to improving the five-year survival rate in LC is the development of screening and diagnostic methodologies which allow for patients with LC to be identified at an earlier stage in the disease. Luis Mur and Keir Lewis have been collaborating in the application of cutting edge omic technologies to identify biomarkers in sputum from LC linked to pathological status and stage. This has focused on the detection of metabolite and microbial (and now proteomic) biomarkers and has led to the filing of two patents and important publications (see below).

The strategy employed in these ‘omic approaches it to identify the main sources of variation in a dataset which are correlated with patient records. Such an approach although powerful, fails to detect complex relationships within and between dataset types and information. Definition of such patterns could greatly aid the successful application of biomarker based diagnoses of different types of LC. This would aid in clinical decision-making in NHS cancer treatment pathways to improve the treatment of lung cancer and therefore improve patient prognosis.

Artificial neural networks (ANN) are a form of artificial intelligence which uses a mathematical representation of the human neurological architecture to replicate its learning and generalisation abilities. ANNs are widely used throughout research as they have the uncanny ability to model immense non-linear systems in which the relation between the variables are either unknown, or acutely convoluted. This multi-disciplinary project bringing together biology, computer science and medical practice will use ANN reveal novel patterns of biomarkers and clinical histories that to better inform diagnoses and improve patient care.