Close
Reliable prognostic markers are not available for patients with conventional cartilaginous tumours. and there is limited understanding of the biological mechanisms by which such tumours develop. If it were possible to link prognostic markers to specific mechanisms this would create opportunities for enhanced therapies and further improvements to patient outcomes.
This project will use machine learning techniques and correlate patient genomic data (e.g., DNA-based features) with image-based phenotypical data (e.g., histopathology) to predict the treatments to which patients are most likely to respond best. We will correlate prognostic markers with computational models of chondrosarcoma development and tumour response to treatment, to understand the significance and robustness of each marker, and potentially identify better, i.e., more discriminative markers and opportunities for improved therapies.
Specifically, we will:
• Use machine learning (ML) to (1) identify histological features that predict clinical outcome using patient archived samples from the last 15 years correlated with clinical outcome, (2) correlate histological features with results from ctDNA, and (3) correlate mutation type, CNV and mutation burden of conventional cartilaginous tumours with clinical outcome.
• Build a computational model of chondrosarcoma growth to elicit a better understanding of the biological mechanisms associated with each prognostic marker. The model will associate specific genomic events with downstream phenotypic effects (cell fate), revealing the underlying mechanism of chondrosarcoma progression and treatment response. The computational model will be verified against the clinical data by using the correlations identified by ML as testing specification. The verified model would then be used to identify more discriminative markers and improved therapies.