Multi-omics profiling exposes deep biological diversity in cancer of unknown primary
Cancer of unknown primary (CUP) remains one of oncology’s most frustrating diagnostic categories because metastatic disease arrives before the tissue of origin can be assigned. In this preprint, investigators used whole-genome sequencing and RNA sequencing to characterize the disease at scale, aiming to move beyond the familiar but incomplete question of where the tumor began. The work framed CUP as a biologically heterogeneous group of cancers rather than a single clinicopathologic entity.
The cohort was substantial for a disease this uncommon: whole-genome sequencing data from 283 patients were evaluated for tissue-of-origin prediction, mutational signatures, driver alterations, genetic immune escape, and clinically actionable findings. RNA sequencing was available in 79 cases, allowing a second layer of analysis focused on gene expression programs and transcriptomic subtypes. Clinical and molecular data were then integrated with weighted gene co-expression network analysis, giving the study enough resolution to connect genomic lesions with broader tumor behavior.
High-confidence tissue-of-origin predictions were generated in 58% of tumors when likelihood exceeded 80%. Lung, gastrointestinal, and biliary tract origins were the most frequent assignments. In practice, that matters because CUP management often begins with empirical chemotherapy in the absence of a clear primary site; a molecular assignment can create a rational path toward tumor-type-specific treatment selection, even when pathology alone leaves the case unresolved.
A single diagnostic label was never going to capture the biology here.
Mutational processes point to several distinct evolutionary routes
The genomic data separated CUP into recognizable mutational program groups. Four subgroups were dominated by tobacco smoking, defective DNA mismatch repair, UV exposure, or APOBEC activity. A fifth group, which included most tumors in the cohort at 169 cases or 60%, showed mixed mutational processes rather than a single dominant signature. This pattern suggests that CUP develops through multiple evolutionary routes, with some tumors carrying a clear carcinogenic imprint and others accumulating alterations through overlapping or poorly resolved mechanisms.
Driver alterations were common and familiar to oncologists: TP53, CDKN2A, and KRAS appeared frequently, alongside epigenetic modifiers including ARID1A, SMARCA4, and ARID2. The presence of these genes is consistent with a tumor population that has undergone broad disruption of cell-cycle control, chromatin regulation, and oncogenic signaling. From a therapeutic standpoint, the findings are less about one magic mutation and more about the cumulative architecture of malignancy, where several pathways may be vulnerable at once.
The study also reported genetic immune escape alterations in 35% of tumors, mostly involving HLA-I genes. Loss or dysfunction in antigen presentation gives the tumor a way to disappear from immune surveillance, a mechanism that fits well with the broader clinical aggressiveness of CUP. If a metastatic cancer can both proliferate quickly and blunt immune recognition, the result is a disease that is harder to detect, harder to classify, and harder to treat with standard site-specific assumptions.
Transcriptomic subtypes add a microenvironment layer that genomics alone misses
RNA profiling identified five transcriptomic subtypes with differing levels of immune infiltration, fibrosis, and tumor proliferation. These subtypes were not significantly linked to the clinical or genomic variables examined, which means the tumor microenvironment appears to behave as a separate axis of heterogeneity rather than a simple reflection of tissue-of-origin or mutation burden. In a disease often treated as a diagnostic dead end, this distinction is clinically relevant because it suggests that two CUP tumors with similar genomic findings may still live in very different immunologic and stromal contexts.
Fibrosis and immune infiltration can influence drug penetration, T-cell access, and sensitivity to immunotherapy. A highly proliferative, immune-cold tumor may behave very differently from a fibrotic but inflamed one, even if both are labeled CUP in the clinic. The transcriptomic data therefore add practical texture to the usual pathology report: they help explain why some tumors may be more resistant to treatment despite harboring actionable-looking mutations.
WGCNA linked proliferative programs to cell-cycle activity, DNA replication, and neuroendocrine features. Those connections are biologically coherent. Tumors driven by rapid replication often accumulate genomic instability, which can sustain aggressive growth while also creating additional therapeutic vulnerabilities, particularly in pathways related to cell-cycle control and DNA damage response.
Actionable alterations were common, but translation still depends on context
Potentially targetable alterations were identified in 64% of patients based on whole-genome sequencing. The phrase “potentially targetable” should be read carefully: it does not mean every alteration will produce a usable treatment option, and it does not guarantee response. It does mean the genomic profile of many CUP tumors already contains clues that could guide therapy selection, trial enrollment, or off-label consideration in the right clinical setting.
The most important message is not that CUP suddenly becomes easy to treat. It is that the disease is molecularly legible in many patients if the right assays are applied. Tissue-of-origin assignment can still matter, but this work argues that origin alone is not the full story. Immune escape, proliferative state, chromatin disruption, and mutational history all help define the tumor’s behavior, and each may point toward a different therapeutic opening.
Because this is a medRxiv preprint, the findings have not yet been validated by peer review. Even so, the dataset is large enough to support a strong biological claim: CUP is not one disease with one solution. It is a collection of metastatic cancers whose shared feature is diagnostic uncertainty, not shared molecular identity.
What this means at the bedside
For clinicians facing CUP, the practical implication is a broader molecular workup when tissue is available. Sequencing can supply a probabilistic origin, expose DNA repair defects or immune-evasion features, and identify alterations that may be worth discussing in a molecular tumor board. The transcriptomic layer may be even more informative in the future if it can help separate tumors that are immunologically active from those that are fibrotic and proliferative.
The study leaves several questions open. RNA-seq was available in only a subset of cases, and the results were derived from retrospective molecular profiling rather than prospective treatment assignment. The next step is not simply more sequencing, but prospective testing of whether these genomic and transcriptomic programs can improve treatment selection, response, and survival in patients with CUP. The biological framework is now clearer; the therapeutic proof still has to be earned.
- • medRxiv preprint DOI 10.64898/2026.11.14.26368097: https://doi.org/10.64898/2026.11.14.26368097
- • ClinicalTrials.gov Trial Protocol (NCT01855477): https://clinicaltrials.gov/study/NCT01855477
- • ClinicalTrials.gov Trial Protocol (NCT02925234): https://clinicaltrials.gov/study/NCT02925234