Lung cancer remains the leading cause of cancer death globally and is the primary cause of death among all Australian aged 65-74.¹ The disease is broadly categorized into non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC), with adenocarcinoma, a subtype of NSCLC, accounting for 40% of cases and showing an increasing trend.² Early-onset lung cancer is a devastating diagnosis that strikes individuals during their most productive years, typically resulting in premature death due to late-stage detection when treatments are largely ineffective.

In response to the high mortality rate, Australia has recently commenced the National Lung Cancer Screening Program, offering targeted screening to high-risk individuals aged 50-70 who undertake a risk assessment using the PLCOm2012 model.³ However, due to this arbitrary cut-off at 50 years, younger individuals at high-risk for early-onset lung cancer are not captured.

Genome-wide association studies (GWAS) have identified 46 susceptibility loci, with 23 of these occurring across all lung cancer subtypes.⁴  Polygenic risk scores (PRS) may enhance lung cancer (LC) risk assessment beyond clinical risk factors,^5,6 with the best published LC-PRS derived using UK Biobank (UKB) data.⁶ A recent meta-PRS for incident LC using UKB data similarly showed independent discriminatory ability (C-index 0.58) ⁷ with PRSs of non-genetic LC-risk factors incorporating sentinel variants from the largest GWASs in Europeans to date.⁸ Functional post-GWAS studies have provided insights into the biological implications of several LC-associated loci, from smoking behaviour to DNA repair.⁴ An exploratory GWAS in Chinese adults identified 4 SNPs, associated with early-onset LC.⁹ However, a comprehensive risk model integrating genetic and clinical factors for predicting early-onset lung cancer is still lacking.

This proposal will leverage the familial Tasmanian Longitudinal Health Study (TAHS) initiated in 1968, comprising ~8500 probands, ~17,000 siblings, and ~22,000 parents of northern European ancestry; >55% being smokers. This proposal will apply the best published LC-PRS ⁶ to the whole exome sequencing and genome-wide SNP array data of TAHS probands. The new risk model for predicting early-onset LC will be developed using UKB data, tested in TAHS, to investigate genetics/familial history, host and environmental predictors and their interactions.

Whilst many prostate tumours progress slowly and are treatable using procedures that are usually minimally invasive, identifying highly aggressive forms of prostate cancers that proliferates rapidly, metastasises and lead to death remains challenging. It is critical to identify the aggressive cases associated with high rates of biochemical recurrence and mortality at the time of diagnosis in order to provide timely and targeted treatment. The Lifestyle and Genetic Risk Factors for Prostate Cancer Study (LGRFPCS) was established to address this need. We have done extensive work to molecularly characterise the germline DNA and the tumours from men participating in this study.

The Lifestyle and Genetic Risk Factor Prostate Cancer Study
The LGRFPCS includes 1,461 incident cases recruited through the Victorian Cancer Registry and urology clinics in Victoria. 974, eligible matched controls were selected from men that underwent a prostate biopsy at the same practice and had a negative biopsy result. Of the 1,461 cases, 360 men in the LGRFPCS were linked to the Victorian Prostate Cancer Outcomes Registry to obtain treatment response and clinical outcome data. These 360 men became the focus for collection of additional data for multi-omic analysis.

Germline molecular analyses
Men in these studies have been included in genome-wide studies of genomic variation using the OncoArray (1), in panel testing studies of putative cancer susceptibility genes (2,3) and many were included in the 1R01CA196931 exome sequencing initiative led by Chris Haiman (4).

Molecular studies of the prostate cancers
The tumours arising in the 360 men have been systematically reviewed by a pathologist which included marking up of the relevant tumours areas for macrodissection. DNA extracted from tissue macrodissected from these marked up tumour areas has been measured for genome-wide methylation using the Infinium methylationEPICv1.0 array (5, 6). Via a collaboration with Prof Renea Taylor (BDI, Monash University), these tumours have undergone staining for several proteins important to prostate cancer characterisation via immunohistochemistry (IHC). We have conducted whole exome sequencing (WES) for a proportion of these tumours with funding from the NHMRC (Southey) and VCA (Dugué). Of the 360 tumours, 160 still require WES data generation.

Epigenetic scores
We have already made great strides in terms of the epigenetic characterisation of prostate cancer aggressiveness using the LGRFPCS resource by assessing a range of epigenetic scores including

Research Approach
Aim 1. Conduct whole-exome sequencing for 160 tumours arising in men participating in the LGRFPCS.
Aim 2. Develop integrative models of prostate cancer outcomes using multi-omic data.
Aim 3. Analyse the somatic genetic and epigenetic features underlying prostate cancer risk factors and its subtypes.

Using family data from the Tasmanian Longitudinal Health Study (TAHS), this project aims to: 1) identify causal risk factors for asthma and chronic obstructive pulmonary disease (COPD); and 2) investigate the causal relationships between asthma and middle-age lung function and COPD. The findings will be validated using UK Biobank data.

Asthma affects one in nine Australians, with an estimated economic burden of $28 billion, but its underlying aetiology remains incompletely understood, with current evidence primarily based on observational studies, limiting the development and implementation of effective asthma prevention strategies.

On the other hand, asthma is a potential risk factor of Chronic Obstructive Pulmonary Disease (COPD), the 4th leading cause of disease burden in Australia and worldwide. COPD was thought to be a smoker’s disease, but recent evidence from observational studies shows that multiple risk factors, including asthma, are also associated with lung function decline and the subsequent development of COPD. Further research is needed to decompose these associations and to inform COPD preventive strategies.

This project aims to address these critical knowledge gaps by employing advanced causal inference methods to explore potential causal associations between asthma, COPD, and their risk factors. By doing so, we seek to provide a more robust evidence base for targeted prevention and management strategies for both conditions

Prostate cancer has a high disease burden, and early detection is essential to improve survival. Current early detection relies on PSA testing, which leads to substantial over diagnosis. Risk based screening has the potential to improve the balance of benefits and harms at the population level. PGSs capture a substantial proportion of prostate cancer risk but are based on data influenced by PSA-driven diagnosis. As a result, current PGSs do not specifically identify individuals at high risk of aggressive disease, and improved approaches are needed.

This project aims to develop an improved PGS for prostate cancer that is independent of PSA levels, addressing a key limitation in current genetic risk prediction. It is hoped the new PGS may better capture true cancer susceptibility, improving specificity in risk prediction. This PGS could help identify men at increased risk of clinically significant prostate cancer, potentially reducing overdiagnosis and increasing screening efficiency.

The research aims are: (1) identify prostate cancer loci independent of PSA levels (1a) post GWAS analysis (2) construct and evaluate a PSA-independent PGS (3) data sharing (to support transparency and future research, CC-GWAS results and new PGS will be deposited in the GWAS Catalog and PGS Catalog, respectively.

Dr Ye’s project aims to investigate the epigenetic basis of age at menopause and menopausal status by identifying and validating differentially methylated CpGs and regions associated with these traits. The research involves integrating discovery and replication cohorts to strengthen evidence and use causal inference methods to evaluate underlying biological mechanisms. Functional annotation and association with epigenetic aging clocks will be used to interpret biological relevance and health implications.

Ultimately, it is hoped to inform preventive strategies and personalised health interventions targeting women at risk of early or late menopause and associated diseases. Identifying epigenetic marks associated with reproductive timing may also uncover broader aging mechanisms, contributing to public health and health span policy.

Despite the gradually wide use of polygenic risk scores (PRSs) in breast cancer risk prediction over the past years, the consistency of individual-level estimates remains an overlooked yet critical dimension for clinical translation. Existing PRSs often yield discordant risk predictions for the same individual, even when exhibiting comparable performance at the population level.

Previous studies have put major attention on that such inconsistencies may stem from methodological differences across PRS models, while potential impacts from individual-level characteristics—such as age, ancestry, and clinical profiles— might interact with these models in complex and poorly understood ways. This variability undermines the interpretability and reliability of PRSs, hindering their integration into clinical and public health practice.

This project aims to investigate the sources of individual-level inconsistency across widely used breast cancer PRSs and develop a more robust risk estimation framework. Utilizing extensive datasets such as the UK Biobank and the Breast Cancer Association Consortium (BCAC), the research will first calculate all publicly available PRSs for breast cancer listed in the PGS Catalog.

Subsequently, a comparative analysis will be conducted to elucidate the extent of discrepancies in the estimated values at the individual level. Unsupervised machine learning techniques will be applied to identify key individual-level factors associated with discordant risk predictions. In addition, ensemble learning methods will be explored to integrate multiple PRSs, with the goal of generating more stable and accurate individual-level risk assessments. By addressing an underexplored yet critical dimension of PRS, this research seeks to improve the clinical validity of PRS and inform the development of more reliable models for precision prevention in breast cancer. The findings are expected to have broader implications for the evaluation and refinement of PRSs across other complex diseases.

Colorectal cancer (CRC) is a leading cause of cancer-related morbidity and mortality worldwide. Polygenic risk scores (PRS) hold significant promise for identifying individuals at elevated genetic risk; however, their predictive performance remains limited for certain conditions such as CRC. A major challenge is that most PRS are derived using standard methods from European ancestry datasets, leading to reduced accuracy and generalizability across other ancestral groups.

This PhD project aims to address these limitations by developing a machine learning model for polygenic risk prediction. The model will be trained using large-scale genomic and phenotypic data from the UK Biobank, leveraging deep neural networks to capture complex, nonlinear relationships and interactions across the genome. A key focus will be on improving predictive performance across diverse ancestries, mitigating bias and enhancing clinical applicability in underrepresented populations. Although CRC is the primary focus, the model will be designed with flexibility to investigate its utility in predicting risk for other complex diseases, particularly other cancers such as breast cancer. External validation is planned using diverse datasets such as the All of Us Research Program, ensuring the model’s robustness and transferability.