AbstractThe medical field is getting flooded with high dimensional datasets, with the advent of high throughput technologies in the omics era. This calls for more sophisticated data analyses technology. Machine learning provides a generic solution to address all kinds of datasets, including high dimensional datasets.
Stratified healthcare involves stratifying patients into different risk groups or ‘endotypes’, which helps in optimising their disease management, whereas personalised medicine involves tailoring the diagnosis and treatment for the patients as per the individual’s biological makeup or use of ‘biomarkers’ in terms of genomic, epigenetic, transcriptomic, proteomic, or other omics profiles. An unsupervised machine learning technique helps in identification of endotypes, whereas a supervised machine learning technique helps in identification of biomarkers.
An unsupervised learning pipeline denoted as MulMorPip, was developed and applied to stratify patients with multimorbidity, in order to identify clusters based on disease diagnosis and interactions (Chapter 3). We have found evidence for five endotypes in patients with multimorbidity using this unsupervised approach. Further, two endotypes of RA were discovered using an unsupervised learning technique and a predictor denoted as ATRPred was developed for the prognosis of anti-TNF treatment response of rheumatoid arthritis patients using a supervised learning technique (Chapter 4). Furthermore, an algorithm muSignAl, was developed that can report multiple signatures with similar predictive power in case of high dimensionality data (Chapter 5).
The thesis has attempted to build computational methods/tools using various machine learning techniques for stratified healthcare and personalised medicine approaches in multimorbidity. These tools can be extended to similar applications in other disease conditions.
|Date of Award||Jun 2023|
|Supervisor||Priyank Shukla (Supervisor) & Tony Bjourson (Supervisor)|
- Treatment response
- Rheumatoid arthritis
- Multiple signature