BACKGROUND

Immunotherapy has revolutionized cancer treatment. However, its clinical application remains limited. There is an urgent need for new predictive and prognostic biomarkers that can identify more patients with objective and durable responses and thus, improve the accuracy of prognosis.

METHODS

A predictive model for immunotherapy was developed using 34 single-cell RNA sequencing (scRNA-Seq) datasets from various cancer types and eight bulk RNA-Seq datasets from immune checkpoint inhibitor (ICI) cohorts. Seven machine learning (ML) methods were applied to identify vital genes associated with both cancer and immune characteristics. Differentially expressed genes (DEGs) were validated using RT-PCR and immunohistochemical (IHC) analyses of clinical samples.

RESULTS

Analysis of scRNA-seq datasets and autonomic nervous system development (ANSD) scores revealed 20 genes comprising a novel ANSD-related differential signature (ANSDR.Sig). A pan-cancer predictive model for ICI prognosis based on ANSDR.Sig was constructed, with the random forest algorithm yielding the most robust performance. Further screening using five ML methods on the ICI RNA-seq datasets identified 18 key genes, forming the Hub-ANSDR.Sig. Regulatory network analysis revealed diversified molecular interactions between Hub-ANSDR.Sig genes, transcription factors, and miRNAs. Hub-ANSDR.Sig was strongly associated with immune cell infiltration, microsatellite instability (MSI), and overall survival (OS) across various cancer types. In gastric cancer (GC), its role in immune dysfunction, tumor mutational burden (TMB), MSI, mutation frequency, immune infiltration, cell-cell communication, and developmental trajectories was confirmed. Moreover, several Hub-ANSDR.Sig genes were differentially expressed in GC compared to normal tissue and were enriched in immunotherapy-sensitive GC samples relative to resistant ones.

CONCLUSION

Our results offer novel insights into predicting immunotherapy efficacy using ANSD-related signature, with the goal of improving clinical strategies and expanding potential indications. This approach also aims to develop more accurate prediction models and therapeutic interventions, thereby helping more patients benefit from immunotherapy.