Integrated multi-omics analysis and machine learning identify G protein-coupled receptor-related signatures for diagnosis and clinical benefits in soft tissue sarcoma.

Background: G protein-coupled receptors (GPRs) are associated with tumor development and prognosis. However, there were fewer reports of GPR-related signatures (GPRSs) in soft tissue sarcomas (STSs), and we aim to combine GPR-related genes with cellular landscape to construct diagnostic and prognostic models in STSs.

Methods: Based on AddModuleScore, single-sample gene set enrichment analysis (ssGSEA), differentially expressed genes (DEGs), and weighted gene co-expression network analysis (WGCNA), GPR-related genes (GPRs) were screened at both the single-cell and bulk RNA-seq levels based on The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. We developed a novel machine learning framework that incorporated 12 machine learning algorithms and their 127 combinations to construct a consensus GPRS to screen biomarkers with diagnostic significance and clinical translation, which was assessed by the internal and external validation datasets. Moreover, the GPR-TME classifier as the prognosis model was constructed and further performed for immune infiltration, functional enrichment, somatic mutation, immunotherapy response prediction, and scRNA-seq analyses.

Results: We identified 151 GPR-related genes at both the single-cell and bulk transcriptome levels, and identified a Stepglm[both]+Enet[alpha=0.6] model with seven GPR-related genes as the final diagnostic predictive model with high accuracy and translational relevance using a 127-combination machine learning computational framework, and the GPR-integrated diagnosis nomogram provided a quantitative tool in clinical practice. Moreover, we identified seven prognosis GPRs and five prognosis-good immune cells constructing the GPR score and TME score, respectively. The findings indicate that high expression of GPRs is associated with a poor prognosis in patients with STS, highlighting the significant role of GPRs and the tumor microenvironment (TME) in STS development. Building up a GPR-TME classifier, low GPR combined with high TME exhibited the most favorable prognosis and immunotherapeutic efficacy, which was further performed for immune infiltration, functional enrichment, somatic mutation, immunotherapy response prediction, and scRNA-seq analyses.

Conclusions: Our study constructed a GPRS that can serve as a promising tool for diagnosis and prognosis prediction, targeted prevention, and personalized medicine in STS.