Graph Machine Learning to Solve Viral Jigsaw Puzzles

Project description

Viruses affect people, animals, and ecosystems, so detecting them can help us track diseases, protect health, and better understand the environment. But viruses are difficult to study because sequencing data is often generated as thousands of short, noisy DNA fragments. The challenge is a bit like solving a huge jigsaw puzzle without the picture on the box: which pieces belong together, and which ones are misleading? Existing software can miss rare or previously unseen viruses, especially when the data is incomplete. In this project, you will explore whether graph machine learning can do better. The idea is to represent the data as a graph, where each DNA fragment is a vertex and connections show evidence that two DNA fragments may be related. Graph Neural Networks (GNNs) can then learn from both the features of each DNA fragment and the structure of the wider graph, helping to group related DNA fragments more accurately. You will work on an applied AI problem involving graph construction, feature design, neural networks, clustering, evaluation, and software implementation. This project gives you the opportunity to apply machine/deep learning, graph algorithms, data science, and bioinformatics to a meaningful real-world problem. You do not need advanced biology knowledge. The outcome will be a prototype tool showing how modern AI can help solve problems in virus discovery, health and medical research, and environmental monitoring.

Co-supervisors

Prof Robert Edwards, Flinders Accelerator for Microbiome Exploration, College of Science & Engineering

Further information

For more information about our research, check out our GitHub profiles metagentools, Vini2, linsalrob, the Edward's lab website and the FAME group's website.

Assumed knowledge

You should have a background in Computer Science, Information Technology, or a related field. A basic understanding of Programming in Python, Graph theory, AI/ML model training would be awesome. We'll teach you bioinformatics and the related biology.

Note: You need to register interest in projects from different supervisors (not a number of projects with the one supervisor).
You must also contact each supervisor directly to discuss both the project details and your suitability to undertake the project.