Abstract 4589: Molecular subtypes identified in lung adenocarcinoma using a large-scale, real-world dataset

Abstract Introduction: Despite advances in targeted and IO therapies, lung adenocarcinoma (LUAD) remains a high unmet clinical need due to its complex molecular landscape. Previous studies have focused on early-stage disease with smaller patient populations to identify molecular subtypes in LUAD. This study characterized the molecular profiles and clinical features of LUAD patient tumors from a large real-world dataset (RWD) containing early and late-stage patients as well as primary and metastatic sites. Methods: The analysis cohort consisted of de-identified clinical and genomic records of patients diagnosed with LUAD and profiled with Tempus’ targeted DNA and whole-transcriptome RNA assays. Molecular subtyping was performed using non-negative matrix factorization on RNA-seq data from primary lung site tumors (n=3, 975; Stage III and IV: 68%). A subtype classifier was trained on tumor-intrinsic features and applied to tumors from primary non-lung and metastatic sites (n=3, 981; Stage III and IV: 99%). quanTIseq was used to estimate immune cell proportions. Progression free survival (PFS), defined as the time from the start of first metastatic therapy to the date of first progression or death from any cause, was assessed in LUAD patients with available data (17.6%). Results: Six distinct molecular subtypes were identified: C1-C6. The C1 subtype (31.8%; median PFS 10.4 months, 95% CI 7.4-13.2) had the highest tumor mutation burden (22.1% with TMB≥10) and PD-L1 status (54.2% with TPS≥50), and was characterized by TP53 mutations (78.62%) with co-mutations in NF1 (log odds ratio 1.8, p<0.05). The C2 subtype (18.82%; PFS 8.8 months, 95% CI 6.6-10.7) showed high EGFR mutation frequency (33.31%), and lower immune cell proportion (41%). In contrast, the C3 subtype (12.63%) showed high prevalence of EGFR mutations (32.37%) with high immune activation (57%), and had the best prognosis (PFS 17.4 months, 95% CI 11.7-36). C3 comprised a large portion of primary lung samples (23%) and was overrepresented in early-stage primary tumors (Stage I and II: 61%) . The C4 subtype (7.03%; PFS 9.8 months, 95% CI 5.5-16.7) was enriched with a fibrotic tumor microenvironment (TME) and an increased proportion of immune cells (54%). The C5 subtype (23.15%) had the worst prognosis (PFS 5.6 months, 95% CI 4.3-7.4), lowest immune cell proportion (36%), and a mutational profile characterized by decreased EGFR mutations (5.6%) and increased STK11 (36.85%), KEAP1 (19.67%), and SMARCA4 (15.95%) mutations. Lastly, the smallest subtype, C6 (6.6%; PFS 9.7 months, 95% CI 6.4-15.4) was characterized with a fibrotic TME. Conclusions: This study highlights the heterogeneity of LUAD, revealing patient states with distinct mutational landscapes, TME profiles, and clinical outcomes in a real-world setting. Future work is needed to explore how these subtypes may inform strategies for disease management and new treatment modalities. Citation Format: Akul Singhania, Prerna Jain, Ezgi Karaesmen Rizvi, Radia M. Johnson, Justin Guinney. Molecular subtypes identified in lung adenocarcinoma using a large-scale, real-world dataset [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 4589.