Interrogating genome and transcriptome data to identify novel cancer biomarkers and create new diagnostic tools.
Timely diagnosis and treatment of cancer can significantly impact a patient’s chances of recovery. Advances in genomics is enabling more rapid approaches, but swiftly managing the vast amounts of data generated can be challenging.
The Cancer Bioinformatics group led by Professor Lachlan Coin is developing and applying tools that analyse genomic and transcriptomic data to quickly identify significant cancer biomarkers, characterise disease state, and predict how well a patient might respond to treatment.
The group is working on streaming algorithms, utilising approaches from high-dimensional statistics, information theory and machine learning, including deep neural networks. They investigate collaborative learning from diverse, linked datasets.
These algorithms process data as soon as it is generated, enabling real-time analysis and visualisation of the most likely disease state and clinical outcomes, along with an understanding of any uncertainties that may decrease as more information is gathered.
Nanopore sequencing for cancer pathology
The Cancer Bioinformatics group are interested in the utility of real-time, streaming analysis of nanopore DNA and RNA sequencing data. Their investigations aim to identify repeat expansions, detect copy number alterations, discover low-frequency nucleotide variants, characterise novel isoforms, and analyse variations in polyA tail length.
Developing minimal multi-omic signatures for diagnosis and prognosis
The Cancer Bioinformatics group has created a comprehensive suite of tools designed to identify minimal transcriptomic, proteomic, and metabolomic signatures for predicting disease states. They aim to apply these signatures to predict drug susceptibility, survival time, and other clinical outcomes. To date, they have successfully utilised the approach to discover biomarkers that differentiate various types of blood cancers.
Developing solutions for linking multi-omic data with clinical outcome and epidemiological data
The group are actively developing methodologies to link multi-omic data, such as genomic and transcriptomic information, with clinical and epidemiological outcomes. They are developing these methods in the HIV/HCV genomics domain, and will explore applications of these approaches in the cancer genomics space.
Using AI to learn from health data
The Cancer Bioinformatics group is investigating federated artificial intelligence (AI) methods to facilitate collaborative learning from linked omic and clinical data.
This enables researchers to extract insights from diverse datasets across institutions, improving overall understanding of disease mechanisms and treatment outcomes.