Introduction: Lung cancer, the leading cause of cancer-related mortality worldwide, is a heterogeneous malignancy comprising distinct histological and molecular subtypes, with non-small cell lung cancer (NSCLC) accounting for approximately 85% of cases and adenocarcinoma (ADC) representing the most prevalent histotype. An emerging pathological feature of NSCLC, spread through air spaces (STAS)—defined as the extension of tumor cells into the lung parenchyma beyond the main tumor margin—has been associated with worse disease-free and overall survival and has been proposed as a possible predictor of recurrence to guide surgical extent. Concurrently, recent comprehensive genomic profiling of early-stage NSCLC has highlighted the need to interpret multi-omics data and their relationship with pathological variables, including IASLC histological subtypes, to better personalize treatment strategies. In this context, we investigated the overall distribution of STAS and its association with tumor mutational profiles and IASLC histological subtypes in a large real-world cohort of lung adenocarcinoma patients from the LANTERN project. Materials and Methods: In a prospective, multicenter observational study (March 2023–December 2024), 271 NSCLC patients were enrolled, and clinicopathological, immunohistochemical, and genomic data were collected; comprehensive genomic profiling was performed using the TruSight Oncology 500 assay to analyze 523 cancer-related genes, tumor mutational burden (TMB), and microsatellite instability; and STAS was assessed according to IASLC criteria. Adenocarcinoma accounted for roughly 90% of the cases, with a median age of 69 years and a predominance of stage IV disease (49.5%). STAS was evaluable in 162 cases and was detected in 17.9% of tumors. Results: STAS-positive tumors showed a higher trend towards locally advanced and advanced disease; no differences were observed in sex, age, smoking status, tumor mutational burden, or PD-L1 expression. Additionally, STAS-positive tumors showed a higher association with micropapillary, mucinous, and papillary patterns, whereas the acinar pattern was more frequent in STAS-negative tumors. The most frequently mutated genes were TP53, KRAS, EGFR, and STK11, with no significant differences between groups; ROS1 alterations were absent in STAS-negative tumors but detected more frequently in STAS-positive cases. Conclusions: Overall, these findings indicate that STAS positivity is associated with high-risk histological subtypes and advanced disease, suggesting its importance as a marker of tumor aggressiveness and emphasizing the need for its systematic evaluation in lung adenocarcinoma to better guide surgical planning and patient risk assessment.

Non-small cell lung cancer (NSCLC) remains the leading cause of cancer-related mortality globally. While multi-modal artificial intelligence (AI) models offer significant predictive potential, their translation into routine clinical practice is delayed by the “black box” nature of complex algorithms and the fragmentation of heterogeneous data. We present LANTERN-XGB, a hierarchical machine learning workflow designed to bridge this gap by generating interpretable “digital human avatars” for precision oncology. The methodology employs a multi-stage scalable tree boosting system (XGBoost) architecture utilizing shapley additive explanations (SHAP) for rigorous hierarchical feature selection, missing value management, and patient-specific decision support. The workflow was developed and benchmarked using a retrospective cohort of 437 patients with clinical N0 NSCLC, followed by validation on a prospective dataset (n = 100) and an independent external dataset (n = 100). The pipeline integrates diverse data modalities to predict occult lymph node metastasis (OLM). LANTERN-XGB identified a robust consensus signature driven by non-linear interactions among CT textural fragmentation, PET metabolic heterogeneity, tumor density distribution, and systemic clinical modulators. Exploratory transcriptomic pathway analysis (GSVA) revealed that high-risk predictions strongly correlate with systemic molecular dysregulation, such as the enrichment of immune-inflammatory signaling and metabolic stress pathways. The model achieved robust discrimination in external validation (AUC ≈ 0.77), performing comparably to state-of-the-art nomogram benchmarks. Crucially, the LANTERN-XGB framework demonstrated superior utility in handling diagnostic ambiguity; local force plots allowed for the correct reclassification of “borderline” prediction by visualizing feature interactions that standard linear models fail to capture. LANTERN-XGB provides a validated, open-source framework that successfully balances predictive power with clinical transparency. By empowering clinicians to visualize and verify the logic behind AI predictions, this workflow offers a pragmatic path for integrating reliable multi-modal avatars into daily medical decision-making.

Objectives: Precision medicine has improved outcomes in advanced lung cancer, but its implementation remains limited. The prospective LANTERN project (NCT05802771) involves comprehensive clinical and multi-omics data collection. Here, we report preliminary genomic findings, focusing on the prognostic impact of gene mutations, pairwise co-mutational patterns, and their stage-specific association with PD-L1 analysis on patients speciments. Methods: This prospective, multicenter observational study enrolled 271 consecutive patients between March 2023 and December 2024, including 137 with stage I–III and 134 with stage IV disease. Comprehensive genomic profiling was performed using the TruSight Oncology 500 High Throughput assay to assess single nucleotide variants, insertions/deletions, copy number alterations, tumor mutational burden and microsatellite instability. The prognostic role of mutations and co-mutations was evaluated, together with associations between genomic alterations and PD-L1 expression. Results: At least one Tier I–II alteration was identified in 84% of patients. TP53 (39%), KRAS (27%), EGFR (15%), and STK11 (11%) were the most frequently altered genes. Several co-mutations were associated with worse overall survival, including TP53/STK11, STK11/LRP1B, STK11/RB1 and RB1/LRP1B, and with reduced disease-free survival, particularly TP53/KRAS, TP53/STK11, KRAS/STK11 and STK11/LRP1B. In advanced disease, STK11 and KEAP1 mutations were associated with low PD-L1 expression, whereas PIK3CA, PTEN and ARID1A mutations correlated with higher PD-L1 expression in early-stage tumors. Conclusions: Comprehensive genomic profiling reveals substantial molecular heterogeneity across lung cancer stages and identifies co-mutational patterns with prognostic and immune-related relevance, supporting broader molecular testing at diagnosis.