Non-small cell lung cancer and the precision oncology revolution. Non-small cell lung cancer (NSCLC) is the leading cause of cancer death worldwide, encompassing adenocarcinoma (LUAD), squamous cell carcinoma (LUSC) and other types. Over the past two decades, identifying specific DNA mutations that drive tumor growth has enabled targeted therapies for a substantial minority of patients - particularly those with mutations in EGFR, ALK, ROS1, RET, KRAS, and other receptor tyrosine kinases.
The treatment gap for 'oncogene-negative' tumors. Tumors with actionable driver mutations represent only about 25-30% of all NSCLC cases. The remaining 70-75% lack these druggable alterations. For these patients, treatment relies primarily on immune checkpoint inhibitors (ICIs) - drugs that release the immune system's brakes - but only about 20% achieve long-term responses with immunotherapy alone, and 20-30% benefit from combined chemo-immunotherapy.
The molecular signature of hard-to-treat NSCLC. Tumors without actionable driver mutations often instead harbor loss-of-function alterations in tumor suppressor genes: TP53 (mutated in over 50% of lung adenocarcinomas), KEAP1 (about 17%), STK11 (about 16%), and NF1. These genes normally restrain cancer growth, and when they are lost or damaged, they promote tumor development through mechanisms that are much harder to target than oncogenic driver mutations.
Why multi-omics approaches are needed. Traditional genomic profiling only reveals DNA mutations and copy number changes. However, a mutation does not always translate into a functional change in the protein it encodes, and some tumors activate cancer pathways through non-genetic mechanisms like altered protein activity or epigenetic changes. By integrating genomics, transcriptomics, proteomics, phosphoproteomics and other data layers, multi-omics studies can expose the actual active signaling networks and identify new therapeutic vulnerabilities invisible to DNA sequencing alone.
Whole genome sequencing reveals what whole exome sequencing misses. Among LUAD cases that appear to lack actionable RTK/RAS/RAF pathway alterations by standard whole exome sequencing, whole genome sequencing identifies such alterations in about one-third more cases. These include KRAS mutations, amplifications of EGFR and MAPK1, deletions of RASA1 and NF1, and NRG1 gene fusions - showing that what appears 'oncogene-negative' by conventional testing is often not truly so.
The genuinely oncogene-negative LUAD molecular profile. Among cases truly negative for RTK/RAS/RAF pathway alterations at the whole genome level, the dominant genetic features are mutations in tumor suppressor genes - particularly TP53 (mutated in approximately 67% of these cases), KEAP1 (24%), STK11 (16%), and SMARCA4 (14%). These tumors also tend to have high tumor mutational burden and complex structural variants including large copy number changes and gene amplifications.
How lung squamous cell carcinoma differs. LUSC has a distinct genomic profile from LUAD. TP53 mutations are nearly universal in LUSC (about 90% of cases), and CDKN2A loss affects approximately 70% of tumors. Actionable gene fusions are rare in LUSC. The most common amplifications involve SOX2, FGFR1, CCND1, PDGFRA, EGFR and MYC, while deletions frequently affect CDKN2A, FOXP1 and PTEN. This profile creates different therapeutic vulnerabilities than LUAD.
Mutation co-occurrences that define treatment subtypes. Certain gene mutation combinations have important clinical implications. KRAS mutations co-occurring with STK11 or KEAP1 mutations create a particularly immunosuppressive tumor microenvironment that often resists immune checkpoint inhibitors. KRAS co-mutations with TP53 show different characteristics and may respond better to immunotherapy. MDM2 and CDK4 amplification co-occurrence suggests potential benefit from combined MDM2 and CDK4/6 inhibitors.
Three foundational transcriptional subtypes of LUAD. The Cancer Genome Atlas identified three transcriptional subtypes of lung adenocarcinoma: terminal respiratory unit (TRU), proximal-inflammatory (PI) and proximal-proliferative (PP). The TRU subtype, enriched for EGFR mutations and RTK fusions, is more common in non-smokers and women. PI tumors, characterized by NF1 and TP53 co-mutations, and PP tumors, linked to KRAS mutations and STK11 loss, are more common in smokers and show higher tumor mutational burden.
Expanded subtyping identifies immune-sensitive and resistant groups. More detailed transcriptomic analysis of a larger LUAD cohort identified five subtypes (S1-S5). Critically, the S3 subtype - marked by high immune and inflammatory signatures along with TP53 and NF1 mutations - showed CD274 (PD-L1) amplification and overexpression, suggesting responsiveness to immune checkpoint inhibitors. Furthermore, the S3 subtype was a better predictor of progression-free survival on ICIs than PD-L1 expression alone in an independent patient cohort.
KEAP1/STK11 mutations define an immunotherapy-resistant subtype. The S4 subtype, enriched for KEAP1 and STK11 mutations, is associated with resistance to anti-PD-1 and anti-PD-L1 therapies across multiple independent studies. These mutations promote an immunosuppressive tumor microenvironment with abundant suppressive myeloid cells and depleted CD8+ cytotoxic T cells. Adding anti-CTLA4 therapy (which activates CD4+ effector T cells) can reprogram suppressive myeloid cells and may overcome this resistance.
Gene signatures that go beyond mutation testing. A 46-gene signature predicting activation of the KEAP1/NRF2 antioxidant pathway outperforms mutation testing in predicting patient survival. A T-effector/interferon-gamma gene expression signature - capturing immune activation through transcripts of PD-L1, CXCL9, and IFN-gamma - identifies NSCLC patients most likely to benefit from atezolizumab, particularly those lacking RTK/RAS/RAF alterations, demonstrating the clinical utility of transcriptomic profiling beyond genomics.
Why proteomics adds unique information beyond genomics. The correlation between gene messenger RNA levels and protein levels is often weak, because protein abundance is also controlled by post-transcriptional regulation (splicing, microRNA activity) and post-translational modification (ubiquitination, protein degradation). Phosphoproteomic analysis - measuring the active, phosphorylated forms of proteins - reveals which signaling pathways are actually active in a tumor, even when the upstream driver gene is not itself mutated. This reveals therapeutic targets invisible to standard DNA sequencing.
Pathway activation without corresponding driver mutations. Phosphoproteomic analysis of LUAD shows that activation of key growth pathways like RAS/MAPK and PI3K/AKT/mTOR is only partially driven by mutations in those same pathways. Drugs targeting proteins within a pathway may be effective even when the direct target protein is not mutated, because pathway activation can arise from alterations elsewhere in the signaling network. This explains cases where genomically guided treatment predictions fail clinically.
Multi-omics clusters uncover ancestry-linked molecular patterns. A comprehensive proteogenomic study of 110 treatment-naive LUAD samples from patients of different ancestries identified four distinct molecular clusters. Cluster C1, enriched for TP53 mutations and high tumor mutational burden, showed immune signaling activation. Cluster C3, enriched for Vietnamese patients and STK11 mutations, showed histone deacetylase and cell cycle pathway activation. Cluster C4, enriched for EGFR mutations in Chinese and female patients, showed MAPK and chromatin organization pathway activation.
Four molecular subgroups in EGFR/ALK-negative LUAD from never-smokers. A proteogenomic study of 99 Korean never-smokers with EGFR- and ALK-negative LUAD identified four actionable subgroups: Proliferation-high (enriched for CDK2, CDK5 and polo-like kinase vulnerabilities), Immune-high (high B-cell infiltration and immune checkpoint expression), Angiogenesis-high (TP53/KRAS co-mutations and proangiogenic factor upregulation), and Metabolism (oxidative phosphorylation and lipid metabolism upregulation with partial CD8+ T-cell suppression). Each subgroup has distinct candidate therapeutic agents identified through drug sensitivity databases.
Three proteomic subtypes of lung squamous cell carcinoma. A multi-omics study of LUSC identified three primary proteomic subtypes: inflamed, redox, and mixed. The inflamed subtype features intense immune infiltration and often contains tertiary lymphoid structures (organized immune cell aggregates) associated with better prognosis - combination immunotherapy strategies may be particularly effective here, potentially enhanced by anti-CD33 drugs to eliminate immunosuppressive myeloid cells.
The redox subtype: metabolic vulnerabilities in LUSC. The redox subtype of LUSC is distinguished by high rates of copy number alterations and mutations in NFE2L2 and KEAP1 (affecting 84% of redox tumors). It shows metabolic vulnerabilities in serine biosynthesis, glycolysis, and reactive oxygen species production. Three genes - TP63, PSAT1, and TFRC - were identified as promising therapeutic targets through computational screening of public gene knockout datasets. TFRC, encoding the transferrin receptor involved in ferroptosis (iron-dependent programmed cell death), is of particular interest.
Universal cell cycle dysregulation in LUSC. A comprehensive proteogenomic study identified loss of cell cycle inhibitors CDKN2A/p16INK4a and RB1 as a near-universal feature of LUSC, occurring through genetic, epigenetic, or unknown mechanisms. Amplification of cyclin D (CCND) or CDK4/6 genes is also frequent. CDK4/6 inhibitors have shown limited success in unselected LUSC, but patients with high RB protein phosphorylation levels may specifically benefit - suggesting that protein-level measurement of RB activation, rather than gene-level testing, is the right biomarker.
Cross-histological multi-omics reveals convergent therapeutic opportunities. An analysis of 1,023 NSCLC cases spanning both LUAD and LUSC identified nine molecular subtypes crossing traditional histological boundaries. Another study of 229 Korean NSCLC patients identified five subtypes including a 'hypoxic' subtype with the worst prognosis - characterized by PI3K-AKT and hypoxia pathway activation across all histological types - and a 'proliferative' subtype enriched for WGD-positive LUSC that showed sensitivity to the nuclear export inhibitor selinexor in organoid models.
TP53: the most mutated gene in NSCLC and a longstanding drug target. TP53 is mutated in over 50% of LUAD and over 85% of LUSC cases. Missense mutations in p53 can cause varying degrees of functional loss, and some gain oncogenic properties. Current approaches under clinical development include small molecules that restore normal p53 protein conformation (such as eprenetapopt and rezatapopt, which selectively reactivates the Y220C p53 mutant), MDM2-inhibiting drugs that prevent normal p53 from being degraded, PROTACs (targeted protein degradation molecules), gene therapy, therapeutic vaccines, and bispecific antibodies.
The KEAP1/NRF2 axis: a master regulator of drug resistance. KEAP1 normally targets the transcription factor NRF2 for destruction by tagging it for proteasomal degradation. When KEAP1 is mutated or lost, NRF2 accumulates and activates a broad antioxidant and metabolic program that promotes cancer cell survival and resistance to multiple therapies including chemotherapy, radiotherapy, RTK inhibitors and immune checkpoint inhibitors. KEAP1 mutations are present in approximately 17% of LUAD and are even more prevalent in the oncogene-negative LUAD subset.
Emerging strategies to target NRF2 pathway activation. Multiple approaches are under investigation to overcome KEAP1/NRF2-driven drug resistance. PROTACs targeting NRF2 for degradation and 'molecular glues' such as R16 (which binds mutant KEAP1 and restores its ability to suppress NRF2) are in preclinical development. NRF2's role in promoting glutamine metabolism creates a vulnerability to glutaminase inhibitors in some tumor contexts. The orphan nuclear receptor NR0B1, expressed selectively in KEAP1-mutant NSCLC, represents another druggable target. Inhibitors of mitochondrial respiratory complex I also show selective activity in NRF2-overactive tumors.
STK11 mutations: metabolic reprogramming and immunotherapy resistance. STK11 encodes liver kinase B1 (LKB1), which activates the energy sensor AMPK and regulates cell metabolism, polarity and migration. STK11 mutations promote metabolic reprogramming and an immunosuppressive microenvironment. Combination approaches under investigation include AXL inhibitors plus pembrolizumab, and strategies targeting the enhanced metabolic dependencies of STK11-mutant tumors. In KRAS/STK11 co-mutated cell lines, HSP90 inhibitors show particular sensitivity.
Single-cell analyses reveal the complexity of tumor evolution. The landmark TRACERx study showed that each NSCLC on average harbors 4.2 truncal driver mutations (present in all tumor cells from early on) and 2.8 subclonal mutations (arising later in subsets of cells). Seventy-seven percent of tumors show at least one whole genome doubling event. Subclonal driver mutations create intratumor heterogeneity - variation within the same tumor - that complicates treatment because therapies targeting one clone may select for resistant clones harboring different alterations.
The tumor microenvironment as a therapeutic target. Multi-omics studies consistently reveal that the immune composition of the tumor microenvironment - the mixture of immune and stromal cells surrounding cancer cells - is as important as the cancer cells themselves for predicting treatment response. KEAP1/STK11-mutant tumors show immunosuppressive microenvironments depleted of CD8+ T cells. Tumors with tertiary lymphoid structures (organized lymphoid aggregates within the tumor) generally show better prognosis. These microenvironmental features can be measured and potentially targeted.
The challenge of translating multi-omics findings to clinical use. Despite the wealth of molecular targets identified by multi-omics studies, translating these findings into approved therapies faces multiple challenges: most identified targets still lack validated clinical-grade inhibitors; multi-omics profiling of patient tumors is expensive and time-consuming; and the molecular heterogeneity within individual patients means that no single therapeutic target will be present in all cells. Future clinical validation will require prospective studies designed with specific molecular selection criteria.
A proposed roadmap for multi-omics clinical translation. The review authors suggest a roadmap that begins with large-scale proteogenomic profiling of well-annotated clinical cohorts, followed by identification of druggable molecular subgroups, development of clinically practical biomarkers for subgroup identification, early-phase basket trials in molecularly selected populations, and ultimately randomized trials comparing molecularly guided to standard therapy. Epigenetic and metabolic vulnerabilities, combination strategies targeting multiple dependencies simultaneously, and tumor microenvironment manipulation are all highlighted as particularly promising directions.
Shifting precision oncology beyond driver mutations. Current precision oncology for NSCLC focuses almost exclusively on the 25-30% of patients with actionable driver mutations. Multi-omics approaches represent the path to extending precision medicine benefits to the remaining 70-75% of patients whose tumors lack these obvious targets. The molecular diversity revealed by proteogenomic studies suggests that 'oncogene-negative' NSCLC is not a single disease but many distinct molecular diseases requiring different treatment strategies.
Combination strategies as the future of treatment. Single-agent targeted therapies have shown limited success against tumor suppressor gene losses, because blocking one pathway in a cell with lost growth restraints is rarely sufficient to halt tumor growth. Multi-omics studies consistently point toward combination strategies - combining ICIs with metabolic inhibitors, cell cycle inhibitors with DNA repair agents, or NRF2 inhibitors with chemotherapy - matched to specific molecular subgroups rather than applied broadly.
New biomarkers for patient selection. Several multi-omics-derived biomarkers are moving toward clinical use: the KEAP1/NRF2 pathway gene signature outperforms mutation testing for predicting survival; the S3 transcriptomic subtype predicts ICI benefit better than PD-L1 expression; and proteogenomic subgrouping can identify patients with specific cell cycle or metabolic vulnerabilities. These richer molecular portraits could replace or complement current single-marker tests.
The promise and the path ahead. The multi-omics revolution has dramatically expanded the catalog of potential therapeutic targets in oncogene-negative NSCLC. Translating this knowledge into improved patient outcomes will require close collaboration between molecular biologists, clinical trialists, and regulators to design studies that test molecularly matched treatments in precisely defined patient subgroups - a more complex but ultimately more scientifically rigorous approach than the broad indication trials that have characterized immunotherapy development.