(P 421) USING TRANSCRIPTOMICS TO IDENTIFY LEIOMYOSARCOMA PATIENTS AT RISK OF METASTASIS

Objective:

Objective: Leiomyosarcoma (LMS) is a rare, aggressive mesenchymal malignancy and represents one of the more common forms of soft tissue sarcoma (STS). Although distant disease failure constitutes the primary feature governing patient outcome, there are no universally accepted biomarkers which can identify patients at high risk for metastatic progression. As the role for adjuvant systemic therapy remains controversial in this disease, we hypothesize that enhanced characterization of LMS molecular subgroups has the potential to better define clinically important biomarkers to tailor more biologically-informed treatments. As such our objective for this study is to robustly identify and characterize molecular subgroups of LMS and translate their utility to the clinical setting.

Methods:

Methods: After Research Ethics Board approval and appropriate Materials Transfer Agreement, we recruited a large cohort of n = 195 LMS tumors and n = 17 leiomyomas with transcriptome data from the Toronto Sarcoma Program, The Cancer Genome Atlas (TCGA), Clinical Sequencing Evidence-Generating Research (CSER) and collaboration with the European Union. We complemented this with clinical data from the respective sources as well as under the Canadian Sarcoma Research and Clinical Collaboration (CanSARCC) consortium agreement. We performed transcriptome analysis using an established locally-developed, previously validated bioinformatic pipeline called RACCOON (Resolution-Adaptive Coarse-to-fine Clusters OptimizatiON), a multilevel RNA-seq unsupervised hierarchical clustering approach to identify molecular subgroups with all LMS tumors, leiomyomas as well as previously published large-scale transcriptome data (n = 13,313 transcriptomes, Comitani et al, Nature Medicine 2023). Translational implications of these molecular subgroups were analyzed through annotation with demographic and clinical variables for survival analysis utilizing Kaplan-Meier methodology and analyzed using GraphPad Prism software (GraphPad Prism 8, Boston USA).

Results:

Results: Transcriptome analysis of the dataset utilizing RACCOON robustly identified three major subgroups of LMS with further multilevel coarse-to-fine clustering delineating further heterogeneity notably of subtype 2 into finer clusters (n = 7). Kaplan-Meier survival analysis comparing clusters identified statistically significantly worse event-free survival associated with subtype 2a (P < 0.05, Gehan-Breslow-Wilcoxon test), but no differences in overall survival. The proportion of patients who developed metastatic disease differed between subgroups, with approximately 75% (12/16) subtype 2a patients developing metastases.

Conclusion:

Conclusions: Large-scale transcriptomics can delineate biologically distinct molecular subgroups of LMS and in particular can help to identify high-risk subgroups with distinct propensity for the development of metastases. Collaborative efforts in the LMS community will aid in further refinement and validation, which is necessary prior to clinical usage.