(P 325) CAN PROTEOMICS PREDICT METASTATIC RELAPSE IN LEIOMYOSARCOMA? DEVELOPMENT OF AN 8 PROTEIN SIGNATURE IN A >350 SAMPLE STUDY INCLUDING A VALIDATION COHORT

Author(s)

Elizabeth A. Connolly, MBCHB MRCP

Research fellow (medical oncology)
Chris O Brien Lifehouse
Sydney, New South Wales, Australia

Co-Author(s)

DB

Daniel Bucio-Noble, -

ProCan®, Children’s Medical Research Institute, Faculty of Medicine and Health, The University of Sydney, Westmead, NSW, Australia
Sydney, New South Wales, Australia
SC

Simon Cai, PhD

ProCan, Children’s Medical Research Institute
Sydney, New South Wales, Australia
AS

Ahilya Singh, PhD

ProCan, Children’s Medical Research Institute
Sydney, New South Wales, Australia
JK

Jennifer Koh, PhD

ProCan, Children’s Medical Research Institute
Sydney, New South Wales, Australia
RK

Rooshdiya Karim, FRCPA PhD

Royal Prince Alfred Hospital-Labs-Anatomical Pathology
Sydney, New South Wales, Australia
AL

Alexander Lazar, MD PhD

MD Anderson Cancer Center
Houston, Texas, United States
RL

Ratana Lim, -

WEHI - Walter and Eliza Hall Institute of Medical Research
Melbourne, Victoria, Australia
RM

Rohini Morey, n/a

The University of Texas MD Anderson Cancer Center -
Houston, Texas, United States
GR

Gayanie Ratnayake, FRACP

The Royal Children's Hospital Melbourne,
Melbourne,, Victoria, Australia
PH

Peter Hains, -

ProCan®, Children’s Medical Research Institute, Faculty of Medicine and Health, The University of Sydney, Westmead, NSW, Australia
Sydney, New South Wales, Australia
Lylee Ye, BSc (Adv) (Hons)

Biobank/Research Officer
Chris O'Brien Lifehouse
Sydney, New South Wales, Australia
DI

Davis Ingram, n/a

MD Anderson Cancer Center
Houston, Texas, United States
CS

Clare Scott, AM MBBS PhD FRACP

Walter and Eliza Hall Institute of Medical Research
Parkville, Victoria, Australia
PR

Phillip J. Robinson, -

ProCan®, Children’s Medical Research Institute, Faculty of Medicine and Health, The University of Sydney, Westmead, NSW, Australia
Sydney, New South Wales, Australia
HB

Holly Barker, PhD

WEHI - Walter and Eliza Hall Institute of Medical Research
Melbourne, Victoria, Australia
EN

Elise Nassif Haddad, MD

Assistant Professor
The University of Texas MD Anderson Cancer Center
Houston, Texas, United States
LH

Lisa Horvath, FRACP, PhD

Chris O'Brien Lifehouse
Sydney, New South Wales, Australia
Peter S. Grimison, BSc(Med) MBBS (Hons) MPH PhD

Staff Specialist in Medical Oncology and Director of Physician Training
Chris O'Brien Lifehouse
Sydney, New South Wales, Australia
RR

Roger Reddel, FRACP PhD

ProCan®, Children’s Medical Research Institute, Faculty of Medicine and Health, The University of Sydney, Westmead, NSW, Australia
Sydney, New South Wales, Australia

Objective: Biomarkers to predict metastatic relapse in LMS are needed to improve risk stratification and could inform adjuvant treatment trials. Recent advances in proteomic technologies may offer novel cancer insights. We aimed to identify a prognostic protein signature of metastatic relapse.

Methods: High throughput quantitative proteomic analysis of FFPE samples, using data independent acquisition mass spectrometry, was undertaken in two cohorts: discovery cohort LMS-AUS (n=225), validation cohort LMS-USA (n=127). For the protein signature, only primary samples of patients with localised disease at diagnosis (LDAD) were included. A novel bioinformatic and machine learning method was developed to identify prognostic proteins and construct a risk score-based signature for metastasis-free survival (MFS). Bioinformatic methods included feature selection with data partitioning, differential expression analysis, Cox and LASSO regression analysis, and cross validation.

Results: The table summarises the cohorts and results. An 8 protein signature (8-PS), developed in LMS-AUS, was highly significant in multivariate analysis (MFS HR 5.2, p< 0.0001) including LMS primary site, size, grade and margin. In LMS-USA, noting the small cohort size (24 patients, 14 MFS events), the 8-PS was significant in univariate analysis of MFS (MFS HR 3.0, p< 0.04) however was not independently significant for MFS in a restricted multivariate analysis with tumour grade/differentiation (HR 1.9, p value 0.26). These results are illustrated in figure 1A.

The 8-PS, when combined with clinicopathological variables (site, size, grade, margin), enhanced predictive performance for MFS in composite models in both LMS-AUS (ROC AUC 0.83) and LMS-USA (ROC AUC 0.80). These results are illustrated in figure 2.

The 8-PS was significant for overall survival (OS) in LMS-AUS (HR 2.1, p 0.03). In LMS-USA, again noting the small cohort size and events (24 patients, 16 deaths), the 8-PS showed a trend to OS significance (HR 2.0, p 0.18). These results are illustrated in figure 1B.

Conclusion: High throughput quantitative proteomics can define prognostic signatures of metastatic relapse in LMS that are significant in independent cohorts and multivariate analysis. An 8 protein signature showed good predictive performance when combined with clinical variables, paving the way for development of targeted proteomic panels that may offer enhanced precision and clinical utility. [Encore abstract: this work has been submitted to ESMO 2024 though updated data has been included in this abstract].