(P 410) PROTEOMICS ANALYSIS IDENTIFIES A SIGNATURE THAT IMPROVES PROGNOSTIC RISK STRATIFICATION OF THE SARCULATOR NOMOGRAM IN PATIENTS WITH SOFT TISSUE SARCOMAS OF THE EXTREMITIES AND TRUNK WALL

Objective: High-risk patients with soft tissue sarcomas of the extremities and trunk wall (eSTS), as defined by the Sarculator nomogram, are more likely to benefit from (neo)adjuvant anthracycline therapy compared to low/intermediate-risk patients. Nomogram predicted risk groups may have distinct biology that may explain their differential outcome after treatment. Here we used high resolution mass spectrometry (MS) to define the protein networks that are characteristic of nomogram-stratified risk groups and evaluate a protein signature based on DNA replication as a complementary prognostic tool to the Sarculator nomogram.

Methods: Proteomic data of primary surgical specimens from 123 eSTS patients across 8 histological types was acquired using MS. A Cox model for overall survival (OS) including the Sarculator nomogram was fitted to individual data to generate predicted probability of 5-year Overall Survival (pr-OS), which were used to define 4 quartile-based risk categories. This allowed us to recalibrate the baseline survival to the present case series. Proteins that were significantly enriched in different risk groups were identified by Wilcoxon-Mann-Witney test with multiple testing correction. Biological pathways that were upregulated in high-risk versus low-risk patients were defined using gene set enrichment analysis (GSEA). The expression pattern of the DNA replication signature - Sarcoma Proteomic Module 6 (SPM6) was evaluated for association with tumour grade (Wilcoxon test), histological type and nomogram risk groups (Kruskal-Wallis test). SPM6 was added as a covariate together with the Sarculator nomogram in a multivariable Cox model. Discriminative ability of the Sarculator only or Sarculator + SPM6 Cox models was evaluated by Harrell C index while Index of Prediction Accuracy (IPA) was used to evaluate discrimination and calibration. Agreement between predicted and observed OS was assessed using calibration plots.

Results: Quartiles were applied as cut-offs to pr-OS for identifying four categories corresponding to high risk (pr-OS≤39%), intermediate risk (39%< pr-OS≤52%), low risk (52%< pr-OS≤69%) and very low risk (pr-OS>69%) patients. Supervised clustering of the proteomic dataset of 3384 proteins of the 4 nomogram predicted risk groups is shown in Figure 1A. Analysis of proteins enriched in high risk versus very low risk patients identified components of the minichromosome maintenance (MCM) complex to be significantly upregulated. We previously defined 14 sarcoma proteomic modules which comprise co-expressed proteins across a range of biological functions. In particular, the SPM6 protein signature comprises 41 DNA replication proteins that includes the MCM complex (Figure 1B). SPM6 median protein expression levels are significantly associated with tumour grade (p = 1.6e-06), histological type (p = 1.4e-05) and nomogram predicted risk groups (p = 2.6e-06). Cox model analysis showed that SPM6 slightly improved the discriminative ability of the Sarculator nomogram (Harrell C index=0.69 for Sarculator nomogram alone vs 0.698 for Sarculator nomogram + SPM6). As regards to calibration, while the nomogram only predictions were quite accurate, despite the baseline survival recalibration there was some discrepancy between predictions and true outcomes particularly in the intermediate and very low risk groups (Figure 1C left) and addition of SPM6 as a covariate substantially contributed to a better calibration (IPA =0.109 vs 0.165; Figure 1C right).

Conclusion: We have performed the first molecular analysis of the biological pathways characteristic of nomogram predicted risk groups in eSTS. We further show that the SPM6 DNA replication signature improves prognostic risk stratification of the nomogram. While hypothesis generating, this study highlights the future potential of integrating protein signatures with Sarculator for the development of next-generation nomograms.