# Integrative Single-Cell and Machine Learning Analysis Identifies a Nucleotide Metabolism-Related Signature Predicting Prognosis and Immunotherapy Response in LUAD

## TL;DR

This study uses single-cell RNA sequencing and machine learning to identify a nucleotide metabolism-related signature that predicts lung cancer prognosis and immunotherapy response.

## Contribution

The study introduces a novel nucleotide metabolism-related signature that predicts prognosis and immunotherapy response in lung adenocarcinoma.

## Key findings

- Tumors with high nucleotide metabolic activity are linked to worse survival and immunosuppressive tumor environments.
- The nucleotide metabolism-related signature accurately predicts patient outcomes and immunotherapy response across multiple cohorts.
- ENO1 is identified as a key gene associated with poor prognosis and immune suppression in lung adenocarcinoma.

## Abstract

Lung adenocarcinoma is the most common subtype of lung cancer and shows marked biological diversity among patients, which leads to different clinical outcomes and treatment responses. Cancer cells require large amounts of nucleotides to support rapid growth and survival, but how nucleotide metabolism varies at the single-cell level and influences the tumor immune environment remains unclear. In this study, we combined single-cell RNA sequencing with machine learning approaches to explore nucleotide metabolism in lung adenocarcinoma. We identified strong metabolic differences among tumor cells and found that tumors with high nucleotide metabolic activity were associated with worse survival and a more immunosuppressive tumor microenvironment. Based on these findings, we developed a nucleotide metabolism-related signature that accurately predicts patient prognosis and potential response to immunotherapy across multiple independent cohorts. Our results provide new insights into how tumor metabolism shapes cancer progression and immune behavior and may help improve personalized treatment strategies for lung adenocarcinoma patients.

Background: Lung adenocarcinoma (LUAD) exhibits pronounced cellular and molecular heterogeneity that shapes tumor progression and therapeutic response. Although nucleotide metabolism is essential for sustaining tumor proliferation and coordinating immune interactions, its single-cell heterogeneity and clinical implications remain incompletely defined. Methods: We integrated a publicly available scRNA-seq dataset derived from independent LUAD patients to construct a comprehensive LUAD cellular atlas, identified malignant epithelial cells using inferCNV, and reconstructed differentiation trajectories via Monocle2. Cell–cell communication patterns under distinct nucleotide metabolic states were assessed using CellChat. A nucleotide metabolism-related signature (NMRS) was subsequently developed across TCGA-LUAD and multiple GEO cohorts using 101 combinations of machine learning algorithms. Its prognostic and immunological predictive value was systematically evaluated. The functional relevance of the key gene ENO1 was further verified through pan-cancer analyses and in vitro experiments. Results: We identified substantial nucleotide metabolic heterogeneity within malignant epithelial cells, closely linked to elevated proliferative activity, glycolytic activation, and increased CNV burden. Pseudotime analysis showed that epithelial cells gradually acquire enhanced immune-modulatory and complement-related functions along their differentiation continuum. High-metabolism epithelial cells exhibited stronger outgoing communication—particularly via MIF, CDH5, and MHC-II pathways—highlighting their potential role in shaping an immunosuppressive microenvironment. The NMRS built from metabolism-related genes provided robust prognostic stratification across multiple cohorts and surpassed conventional clinical parameters. Immune profiling revealed that high-NMRS tumors displayed increased T-cell dysfunction, stronger exclusion, higher TIDE scores, and lower IPS, suggesting poorer responses to immune checkpoint blockade. ENO1, markedly upregulated in high-NMRS tumors and functioning as a risk factor in several cancer types, was experimentally shown to promote invasion in LUAD cell lines. Conclusions: This study delineates the profound impact of nucleotide metabolic reprogramming on epithelial cell states, immune ecology, and malignant evolution in LUAD. The NMRS provides a robust predictor of prognosis and immunotherapy response across cohorts, while ENO1 emerges as a pivotal metabolic–immune mediator and promising therapeutic target.