# Lung cancer metastasis model needs wider validation > Spatial profiling of tumors from 52 people found organ-linked molecular patterns, but the high internal-test scores do not yet make a clinical prediction tool. _Source: Peer-reviewed open-access study in Signal Transduction and Targeted Therapy, cross-checked against PubMed, Crossref, and OpenAlex · 2026-09-07 · 7 min read · Verified against primary sources_ Canonical: https://iyu.app/e/lung-adenocarcinoma-metastasis-signatures ## The 60-second version Spatial profiling in 52 lung adenocarcinoma patients found organ-linked molecular signatures, but the prediction models have only internal validation. **Key points** - The retrospective cohort included 23 brain, 26 liver, and three adrenal matched tissue pairs, with no paired bone metastasis tissue. - Internal test AUCs ranged from 0.907 to 0.975 after 60/40 splits of the same small cohort. - Signals such as FKBP1A, MOCOS, ADAMTSL2, and CKAP2 are candidate associations, not proven causes or drug targets. - The authors reported no competing interests and called for larger, diverse cohorts plus functional validation. **Verdict.** This is a useful biological hypothesis map, not a clinical test; surveillance and treatment should not change on its basis. ## Full explainer > **⚑ Caveat:** The reported AUCs came from internal train/test splits of one 52-person retrospective cohort. There was no independent external validation, so these models are not ready to guide surveillance or treatment. ### Study design — What the researchers measured The team retrospectively studied **52 people with advanced lung adenocarcinoma and distant metastases**. Digital spatial profiling measured gene activity separately in tumor, immune, and stromal regions; multiplex immunofluorescence and clinical records added cell and outcome information. - **52** — patients in one retrospective cohort - **60/40** — internal training and test split - **4** — organ-specific risk models The paired tissue collection was uneven: **23 primary-brain pairs, 26 primary-liver pairs, and three primary-adrenal pairs**. No paired bone metastasis tissue was available. That imbalance is central to interpreting the organ-specific claims. ### Spatial map — Different compartments carried different signals Spatial profiling avoids treating a tumor as one blended sample. It let the researchers ask whether an association appeared mainly in cancer cells, immune cells, or supporting stroma. The resulting candidates differed by metastatic site. - **Brain:** Tumor-cell FKBP1A and stromal CKAP2 were among the selected signals; cell-death-related pathways were enriched. - **Liver:** Tumor-cell MOCOS and immune-region ADAMTSL2 were among the candidates; extracellular-matrix and chromatin-related patterns appeared. - **Adrenal gland:** PTGR2 and IGKC in tumor regions and IRF3 in stroma were selected, but only three matched adrenal pairs were available. - **Bone:** ARMCX6 and PLPP1 were among the primary-tumor signals; there was no matched bone metastasis tissue for lesion-to-lesion analysis. > **i** These genes are associations selected within this dataset. The paper did not functionally prove that they cause organ-specific metastasis or that targeting them improves outcomes. ### Model results — High AUC is not external validation - **Brain metastasis:** Internal test AUC 0.974; 95% CI 0.903-1.000. - **Liver metastasis:** Internal test AUC 0.975; 95% CI 0.906-1.000. - **Adrenal metastasis:** Internal test AUC 0.929; 95% CI 0.789-1.000. - **Bone metastasis:** Internal test AUC 0.907; 95% CI 0.755-1.000. AUC describes discrimination in a particular sample. It does not establish calibration, benefit to patients, or performance at another hospital. All four models were developed and tested by splitting the same small cohort, and the study did not include an independent external cohort. > The study offers a map of candidates, not a clinical navigation system. ### Survival analysis — An unexpected immune association needs caution Post-metastasis analyses highlighted stromal genes including **PKM** for overall survival and **VCAM1** for progression-free survival. Greater M2 macrophage infiltration correlated with better outcomes in this cohort, which runs against a common immunosuppression narrative. The authors cautioned that organ site, treatment regimens, tumor burden, performance status, and other confounders could influence that association. It does not show that M2 macrophages are protective or provide a treatment target. ### Evidence boundary — What the study can and cannot support - **Supported:** primary tumors in this cohort showed spatial molecular patterns associated with different metastatic sites. - **Supported:** internally tested random-forest models separated groups with high AUCs in this dataset. - **Not established:** that these signatures predict accurately in an independent population. - **Not established:** that changing any highlighted gene prevents metastasis or improves survival. - **Not appropriate today:** changing an individual patient's imaging, prevention, or treatment plan from these models. The paper was peer reviewed and indexed in PubMed; the authors reported ethics approval and no competing interests. Its next necessary steps are multicenter external validation, functional experiments, and prospective testing of calibration and patient benefit. Until then, clinical decisions should follow established oncology guidance. ## Primary sources - [Telegram post 1471](https://t.me/CNSmydream/1471) - [Signal Transduction and Targeted Therapy paper](https://doi.org/10.1038/s41392-026-02817-y) - [PubMed record (PMID 42493498)](https://pubmed.ncbi.nlm.nih.gov/42493498/) - [Crossref metadata](https://api.crossref.org/works/10.1038/s41392-026-02817-y) - [OpenAlex record](https://openalex.org/W7170153028) --- _Published by iyu (https://iyu.app) — the day's AI news, checked against primary sources and rewritten in plain language. Free to quote with attribution and a link to the canonical URL._