1115P - Multiomics clustering of patients with cutaneous melanoma to reveal survival trends based on tumor immune evasion features

Background

Immunotherapy resistance due to complex tumor-immune interactions leading to immune evasion is a major hurdle in treating melanoma. To better understand this complexity, we combined genomic and transcriptomic data linked to immune escape in a multiomics approach to discern patients with distinct tumor-immune microenvironment (TiME) types that define their survival.

Methods

Cutaneous melanoma samples from TCGA (discovery cohort, standard of care, n=442) were clustered using a non-negative matrix factorization model based on transcriptomic (functional gene expression signature (FGES) scores and signaling pathway scores calculated by ssGSEA and PROGENy) and genomic (mutational signatures, copy number alterations, and pathogenic mutations) features. The best model parameters were defined by analyzing the clusters’ Kassandra deconvolution, H&E TilMaps, and TCR/BCR repertoires. The CatBoost classifier was then trained to assign TiME types to a validation cohort treated with aPD-1 drugs (n=250). Overall survival (OS) was compared using log-rank test and multivariate Cox regression.

Results

Our analysis defined five TiME clusters. The Checkpoint-high TiME with high tertiary lymphoid structure score, diverse TCR/BCR repertoires, and alterations in melanogenesis pathway genes, showed the best OS in both the TCGA and aPD-1 cohorts. Next was the Immune-depleted TiME, low in all features except lymphoid checkpoints and DNA-repair gene deletions. The Interferon-deleted TiME, immune-rich with high myeloid component and altered interferon genes, showed poor OS in both cohorts. The Proliferative-metastatic TiME showed high tumor proliferation and hypoxia-associated features and the worst OS in the aPD-1 cohort. The Fibrotic/Exclusionary TiME with upregulated WNT signaling, matrix remodeling, and T cell exclusion had the worst OS in the TCGA cohort.

Conclusions

Our analysis of tumor genomics, signaling pathways, and FGES revealed specific TiME patterns affecting patient survival and expanded the transcriptome-only clusterization concept published in Bagaev et al, 2021, highlighting the potential utility of a multiomics approach for development of new therapeutic strategies in precision oncology.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

BostonGene, Corp.

Funding

BostonGene, Corp.

Disclosure

A. Leleytner, N. Lukashevich, E. Ocheredko, S. Ambaryan, A. Nadiryan: Financial Interests, Personal, Full or part-time Employment: BostonGene, Corp. M. Savchenko, S.T. Yong, M. Goldberg: Financial Interests, Personal, Full or part-time Employment: BostonGene, Corp; Financial Interests, Personal, Stocks/Shares: BostonGene, Corp. A. Bagaev: Financial Interests, Personal, Full or part-time Employment: BostonGene, Corp; Financial Interests, Personal, Stocks/Shares: BostonGene, Corp; Financial Interests, Project Lead: BostonGene, Corp; Financial Interests, Leadership Role: BostonGene, Corp. S. Kust: Financial Interests, Personal, Full or part-time Employment: BostonGene, Corp; Financial Interests, Institutional, Project Lead: BostonGene, Corp.