An early invasive signature from the adenocarcinoma in situ-to-invasive adenocarcinoma single-cell trajectory stratifies stage I lung adenocarcinoma overall survival.

BACKGROUND: Stage I lung adenocarcinoma (LUAD) shows significant variability in prognosis that the traditional tumor-node-metastasis (TNM) system fails to fully detect. Most current prognostic signatures are based on bulk-level statistical analyses and do not have a clear biological basis. This research focused on identifying an early-invasive gene set (EIGS) through single-cell pseudotime trajectory analysis and creating a simple risk score for better prognostic classification in stage I LUAD. METHODS: Single-cell RNA sequencing data from adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA), and invasive adenocarcinoma (IAC) were analyzed to reconstruct malignant epithelial pseudotime trajectories using Monocle 2. A 200-gene EIGS was derived from the invasion-associated branch and condensed into a multi-gene EIGS Score via stepwise Cox forward selection with Ridge penalization in The Cancer Genome Atlas (TCGA) stage I training cohort (n=269). The model was externally validated in three independent cohorts (GSE37745, n=70; GSE50081, n=92; GSE72094, n=254) and assessed by pooled analysis. Immune microenvironment characterization, drug-sensitivity profiling, and virtual gene-knockout analysis were conducted to assess the biological and therapeutic relevance of the EIGS Score. RESULTS: The EIGS Score stratified overall survival (OS) in the training cohort [hazard ratio (HR) = 4.53, P<0.001; concordance index (C-index) =0.698] and was confirmed in pooled external validation across three cohorts [pooled OS: HR =2.50, P<0.001, C-index =0.640; pooled disease-free survival (DFS): HR =3.26, P=0.001, C-index =0.653]. Multivariable Cox regression confirmed the EIGS Score as an independent prognostic factor after adjusting for age, sex, and substage (HR =1.93; P<0.001). High-EIGS tumors exhibited lower immune infiltration scores, enrichment of the high-plasticity cell state, and a selective drug-sensitivity profile. The virtual knockout of ARL4C disrupted immune-related transcriptional programs, supporting a regulatory role for ARL4C in connecting the EIGS network to immune microenvironment remodeling. CONCLUSIONS: The EIGS Score, derived from a single-cell invasive trajectory, consistently demonstrates prognostic value in stage I LUAD and may complement existing risk-stratification tools, though further prospective validation is needed.