Machine learning and graph computing for discovering biological mechanisms of cancer

Graphs (networks) have been used to model and analyse biological systems, which have shown promise for gaining insights into the underlying biological mechanisms of complex biological phenomena such as cancer. However, to unlock the full potential of graphs to discover reliable mechanisms for efficient cancer diagnosis and treatment, we need to address two challenges. The first challenge is how to extend simple binary interaction graphs to integrate diverse and complex prior knowledge as well as ‘omics’ datasets so that they can more accurately describe cancer biological systems. The second challenge is how to extract the most important information (features) from the integrated graph that is relevant to the underlying mechanisms.

To discover reliable biological mechanism of cancer, this PhD project will develop novel computational methods to address the above two challenges. The principle idea is to combine graph computing and graph feature (representation) learning, two fast growing fields in network science and machine learning. Collaborating with cancer biologists, you will apply your methods to investigate how the RTK (Receptor Tyrosine Kinases) pathway, a critical signalling pathway in the development of breast cancer, is activated, and more importantly, how it can be targeted and disrupted by drugs. Specifically, the overarching project objectives are:
1. Developing methods to construct property (attributed) graphs, a popular data model in graph computing, to integrate multi-omics data and prior knowledge
2. Design feature learning algorithms, e.g., graph embedding algorithms for property (attributed) graphs
3. Applying above methods and algorithms to discover the reactivation mechanisms of RTK pathway in breast cancer

The student project will be defined within this project scope; specific research questions will be refined in line with the student’s interests and prior experience.