MSACL 2022 Abstract

Introduction

The diagnosis and prognostication for Prostate cancer (PCa) remains challenging in this relatively common cancer. Diagnosis can involve a number of clinicopathological indicators, including Gleason score and prostate-specific antigen (PSA) for example. PSA which is the most common of the few blood-based protein biomarkers currently available in clinical practice, however, PSA by itself is not accurate especially since there is no reliable PSA range that explicitly signifies the presence of PCa. Studies on potential biomarkers and measurable signatures for the disease thus remain a key area of translational research. Appropriate assay accuracy, precision and sensitivity still need to be attained. Combining multiple levels of molecular information has the potential to improve biomarker panels and help gain a better understanding into the biology underpinning the disease.

Objectives

Within this study, we used a combination of lipid and protein measurements to identify biomarkers that are more beneficial in detecting the disease status of men who are most likely to develop PCa.

Methods

Data from newly diagnosed PCa patients at various stages of the disease, as well as age-matched controls, were used to generate proteomic and lipidomic profiles. Serum samples were collected from newly diagnosed prostate cancer patients and their age matched healthy individuals. Healthy control (n=126) samples satisfied both a normal digital rectal examination (DRE) and prostate-specific antigen (PSA) levels below 1 ng/mL (<1ng/mL). The inclusion criteria for newly diagnosed prostate cancer patients (n=205) were an abnormal prostate on DRE, symptomatic presentation with high PSA levels and abnormal biopsy; or alternatively, a diagnosis made solely on a steep rise in PSA associated with urinary symptoms.

Proteomic LC-MS data were collected using a high throughput, discovery workflow, which utilised an ion mobility-data independent analysis (IM-DIA) strategy. The corresponding lipidomic data were acquired using a high throughput, targeted MRM methodology, providing semi-quantitative information. A total of 262 lipids were monitored.

Subsequent data analysis involving Random Forest and Boruta algorithms, as well as the Mann-Whitney t-test, were utilised to build prediction models. Integration of the lipidomic and proteomic datasets, each log transformed lipidomic and proteomic matrix was Z-score transformed using the scale() function in R (version 4.0.2). Next the two matrices were combined by matching sample identifiers, whilst missing values in both datasets were omitted from the analysis by default.

Results

An integrative multi‘omic approach highlights the tightly regulated and inter-connected networks of lipids and proteins, associated with PCa phenotype, and regulation of acute phase response signaling pathways. This study demonstrates the added value of an integrated data analysis pipeline by identifying clinically relevant biomarkers, in addition to providing insight into the underlying biological pathways/mechanisms. Data resulting from this study have identified the differential expression of circulatory lipids and proteins that may preferentially be a consequence of localised PCa, providing a distinction between healthy controls and various stages of PCa. Blood-based multi‘omic strategies hold clinical promise; with the ability to distinguish between patients requiring active surveillance from those requiring immediate treatment, aiding the decision making in clinical practice.

Conclusions

We identified signatures for mild and advanced staged PCa, providing AUC values of 0.955 and 0.966, respectively. Combining lipidomic and proteomic data, provided a striking separation between cancer and non-cancer samples. Importantly, we found that based on the top five biomarkers (i.e., combination of lipids and proteins) provided cumulative AUCs of 0.940 and 0.955 for mild and advanced staged PCa respectively.