Bioinformatics Seminar
Time: 11AM
Venue: Davis Auditorium and Online
12 August 2025
Tracking 'Biometric Fingerprints' for Paediatric Cancer with Deep-Learning
Kelvin TuongUniversity of Queensland
Paediatric cancers are remarkably heterogeneous, both clinically and genomically. Traditional minimal residual disease (MRD) detection methods, such as multi-parametric flow cytometry, are limited by sensitivity, often missing low-burden disease that ultimately relapse. In this talk, I will present our preliminary/unpublished proof-of-concept work on training deep-learning models to learn and track these individualised cancer and immune signatures, akin to 'biometric fingerprints'. With single-cell gene expression data, we use deep-learning on individual patients' unique cancer expression profile. For immune tracking, we are training graph neural networks to perform classification of T-cell receptors in cancer versus healthy. Together, we hope that the models have the potential to better detect residual disease and forecast relapse by recognising the recurrence of a patient's molecular fingerprint in follow-up samples.