Background: Loss of STK11 occurs in 20-25% of NSCLC and confers poor prognosis and resistance to ICB. STK11 loss affects lineage plasticity in murine models and differentiation phenotypes in patients. However, interactions between differentiation state, TME, and clinical outcomes are not well characterized. Methods: This is a retrospective and descriptive study across eight transcriptomic datasets of 2118 non-squamous NSCLC with 314 STK11 mutations, including 628 patients randomized to docetaxel or atezolizumab in the OAK and POPLAR clinical trials. STK11-DPC applies a validated gene signature to assess functional STK11 loss, then uses a rule-based classifier to define three differentiation lineages: “Neuroendocrine,”“TTF1-Low,” and TTF1-Positive.” TME is assessed by xCell immune deconvolution, and patient outcomes assessed by Cox proportional hazards regression models. Results: STK11-DPC significantly outperforms STK11 mutations in predicting decreased MHC-1, MHC-2, interferon-stimulated genes, and immune infiltration (P < 1e-8 for each). Each of the three differentiation groups has low immune infiltration compared to STK11 WT tumors, but we observe distinctive TME phenotypes corresponding to the Neuroendocrine and TTF1-Low groups. Neuroendocrine tumors have low expression of pro-inflammatory genes in the TNF/NFKB and JAK/STAT signaling pathways (P < 1e-10), while TTF1-Low tumors show hallmarks of increased inflammation, with higher PMN-MDSC, CD8 and NK signatures, but relative exclusion of dendritic cells (P < 1e-6 for each). These differences persist after controlling for KRAS, KEAP1, and TP53 mutations. Conversely, STK11-deficient tumors in WT-like or TTF1-Positive groups have TME more similar to STK11-WT tumors and show increased immune infiltration. STK11-DPC was prognostic when applied to OAK and POPLAR, with worse OS in Neuroendocrine (HR 1.6, P = 0.006) and TTF1-Low (HR 2.4, P = 5.8e-10) groups. Survival in both arms was equally poor in these groups. In contrast, STK11-deficient tumors with WT-Like or TTF1-Positive phenotypes had better OS with atezolizumab compared to docetaxel (HR 0.5, P = 0.03). Conclusions: STK11-DPC is highly predictive of immune exclusion and NSCLC patient outcomes. It identifies STK11-deficient subsets with increased immune infiltration and improved ICB response, and two putatively immune-resistant subgroups with markedly different TME phenotypes: Neuroendocrine tumors have an anti-inflammatory phenotype, while TTF1-Low tumors exhibit chronic inflammation. Performance of STK11-DPC as a predictive biomarker warrants further validation in additional patient cohorts, while the apparent TME differences suggest distinct and non-overlapping mechanisms of ICB resistance and may guide development of precision treatment strategies.