Clinical bottom line
Artificial intelligence (AI) approaches integrating tumor immune cell metabolic states offer a promising but still emerging avenue for predicting pathological features relevant to cancer progression and treatment response. Recent studies highlight that metabolic heterogeneity within tumor-infiltrating immune cells—such as macrophages and T cells—reflects spatially organized tumor states and immune escape mechanisms. AI models leveraging spatially resolved metabolomic and transcriptomic data can identify metabolic phenotypes linked to immune suppression or activation, potentially aiding in patient stratification and therapeutic targeting. However, current evidence remains largely preclinical and hypothesis-generating, with limited validation in diverse clinical cohorts. Pathologists should view AI-based metabolic profiling as a complementary research tool that warrants further validation before routine clinical application.
What the evidence shows
Tumor immune cell metabolism is increasingly recognized as a critical determinant of the tumor microenvironment (TME) and immune response efficacy. Shukla Kirtikar et al. (2026) demonstrated that spatially organized immune-metabolic-redox ecosystems define distinct tumor states in head and neck squamous cell carcinoma, with AI models successfully predicting pathological features by integrating metabolic and immune cell spatial data. These metabolic states correlate with immune cell exhaustion and tumor immune escape, underscoring their biological relevance.
Foglia et al. (2026) reviewed the dynamic metabolic tug-of-war between tumor cells and immune cells, emphasizing how metabolic rewiring supports immune evasion and resistance to therapy. AI-driven analyses of metabolic profiles can potentially identify metabolic checkpoints that modulate immune cell function, such as glycolysis and oxidative phosphorylation pathways, which are critical in macrophage polarization and T cell activation.
Tsukamoto et al. (2021) highlighted macrophages’ dual roles in anti-tumor immunity and tumor progression, noting that their metabolic states influence their functional phenotypes. AI models that incorporate metabolic markers may better characterize macrophage heterogeneity within tumors, aiding pathological assessment.
Spatially resolved metabolomics combined with 3D tumor-immune coculture models have revealed metabolic alterations during antitumor immune responses, providing mechanistic insights that AI can exploit to predict pathological features (source 4). These approaches enable mapping of metabolic gradients and immune cell interactions in situ, which traditional bulk analyses cannot capture.
Collectively, these studies suggest AI can integrate complex metabolic and spatial data to predict tumor immune microenvironment features linked to prognosis and treatment response, potentially guiding immunotherapy strategies.
Caveats and uncertainty
Despite promising findings, the clinical utility of AI models predicting pathological features from tumor immune cell metabolism remains unproven. Most evidence derives from preclinical models or limited patient cohorts, often focusing on specific cancer types such as head and neck squamous cell carcinoma. The generalizability of these metabolic signatures across diverse tumors and patient populations is unclear.
Technical challenges include standardizing spatial metabolomic data acquisition, integrating multi-omics datasets, and ensuring reproducibility of AI algorithms. The metabolic plasticity of immune cells and tumor heterogeneity complicate interpretation, and causal relationships between metabolic states and pathological outcomes require further elucidation.
Moreover, current AI models often lack prospective clinical validation and have not yet demonstrated improved diagnostic accuracy or patient outcomes compared to established pathological assessments. Regulatory and ethical considerations around AI-driven metabolic profiling also remain to be addressed.
How this may change practice
In the near term, AI-based analyses of tumor immune cell metabolism are best viewed as research tools that deepen understanding of tumor-immune interactions and identify candidate metabolic biomarkers. Pathologists may increasingly collaborate with computational biologists to incorporate spatial metabolomic data into research protocols.
With further validation, these AI approaches could augment pathological evaluation by providing metabolic context to immune infiltration patterns, potentially refining prognostic stratification and predicting response to immunotherapies. This could lead to more personalized treatment decisions based on metabolic-immune phenotyping.
Ultimately, integrating AI-driven metabolic profiling into routine pathology workflows will require robust clinical validation, standardization of assays, and demonstration of added clinical value. Until then, clinicians should interpret such data cautiously and in conjunction with established diagnostic criteria.