Lung cancer remains the leading cause of cancer-related mortality worldwide, and although immunotherapy has transformed the treatment landscape of advanced non-small cell lung cancer (NSCLC), durable benefit is limited to a subset of patients. PD-L1 immunohistochemistry and tumor mutational burden, while clinically utilized, demonstrate imperfect predictive capacity, underscoring the need for more robust biomarkers. This review highlights emerging multimodal biomarkers—including circulating tumor DNA (ctDNA), the gut microbiome, and artificial intelligence (AI)-driven radiomics—as promising tools to enhance the prediction of immunotherapy response. Longitudinal ctDNA monitoring offers a minimally invasive method to assess tumor burden dynamics, detect early molecular response, distinguish pseudo-progression from true progression, and stratify risk, with ctDNA clearance correlating with improved survival outcomes. The gut microbiome has also been associated with ICI efficacy, as specific bacterial taxa and composite scoring systems correlate with treatment response, though methodological heterogeneity limits clinical translation. Radiomic analyses leveraging CT and PET imaging extract quantitative tumor features that, when integrated with clinical and molecular data, demonstrate improved predictive performance compared to single-modality approaches. Despite promising advances, challenges including assay standardization, external validation, data harmonization, interpretability of AI models, and infrastructure requirements remain barriers to widespread adoption. Multimodal integration of genomic, microbiome, and imaging biomarkers represents a critical step toward precision immuno-oncology, with prospective validation needed to translate these approaches into improved outcomes for patients with advanced NSCLC.

Immune checkpoint inhibitors (ICIs) have transformed non-small cell lung cancer (NSCLC) treatment; however, durable responses occur in only a subset of patients, underscoring the need for robust predictive biomarkers. Serine/threonine kinase 11 (STK11) is an emerging biomarker that portends poor prognosis and predicts therapeutic resistance. Loss of STK11 disrupts AMPK signaling, leading to unchecked mTOR activation, metabolic reprogramming, angiogenesis, and epithelial–mesenchymal transition, fostering tumor progression and immune evasion. STK11 mutations frequently co-occur with KRAS and KEAP1 alterations, exhibit low PD-L1 expression, an immunosuppressive tumor microenvironment that leads to the development of PD-1/PD-L1 resistance. Clinical studies consistently demonstrate inferior outcomes with ICIs in STK11-mutant NSCLC, particularly in the presence of KRAS and KEAP1 co-mutations. Dual checkpoint inhibition combining PD-1/PD-L1 and CTLA-4 blockade shows promise in overcoming resistance, results remain inconsistent, and prospective trials are ongoing. Beyond immunotherapy, STK11 mutations confer poor outcomes across targeted therapies, including KRAS G12C inhibitors, with KEAP1 co-mutation serving as a strong negative predictor of efficacy. In this review we present an overview of STK11 function and its role in tumor biology, highlight the prognostic and predictive potential of STK11 mutations in the context of NSCLC treatment and summarize the emerging treatment strategies.

Lung cancer remains the leading cause of cancer-related death globally, despite significant advances in diagnosis and treatment. Single biomarker approaches used clinically, such as programmed death ligand-1 (PD-L1) expression levels, have limited capacity for predicting treatment response. Multimodal data analysis using artificial intelligence (AI) offers an innovative scope to integrate diverse data sources—including radiologic imaging, digital pathology, genomics, immunohistochemistry, and Cell Painting morphology—to improve clinical predictions. This review aims to examine multimodal AI applications across the lung cancer treatment landscape related to such data sources. We analyze technical architectures spanning convolutional neural networks for imaging, vision transformers for pathology, and graph neural networks for genomics. We discuss how integrating and learning from heterogeneous data sources requires cross-attention fusion mechanisms. We further analyze critical studies demonstrating that multimodal AI clinical applications achieve superior predictive performance compared to unimodal biomarker methods. Multimodal AI models can augment clinicians in treatment selection, longitudinal monitoring using circulating tumor DNA (ctDNA), and variant interpretation through morphological profiling. We propose developing a multimodal AI model to optimize precision oncology for lung cancer.