+90 850 460 63 34+90 532 445 81 40info@mutludemiray.com
Prof. Dr. Mutlu DemirayMedical Oncology
TürkçeEnglishРусскийفارسی

A New Era in Cancer of Unknown Primary: Finding the Lost Origin and Guiding Treatment with Artificial Intelligence and Genomic Mapping

Published: 15.09.2026Dr. Ebru Gül Karakoç

Cancer of unknown primary (CUP) is one of the most challenging areas of oncology: the cancer has metastasized (spread through the body), but its original starting point, that is, the tissue in which it originated, cannot be identified. In the past, standard empirical chemotherapy was regarded as the main option for tumors of uncertain origin; today, a major clinical transformation is under way thanks to Comprehensive Genomic Profiling (CGP), epigenomic tests and the use of artificial intelligence (AI). In oncology, AI algorithms and liquid biopsy technologies are now used not only to determine a cancer's tissue of origin, but also to predict responses to targeted therapies. Our Molecular Tumor Boards (MTB), in turn, process all of this data in detail and design standard chemotherapy regimens for our patients only when they are needed and precisely targeted smart-drug combinations when those are needed, allowing us to deliver "the right treatment to the right patient."

1. Finding the Tumor's Origin with WES, WTS, Epigenomic Methods and Artificial Intelligence (AI)

Today, thanks to advances in Whole Exome Sequencing (WES), Whole Transcriptome Sequencing (WTS) and epigenomic methods, in cancers of unknown primary we can not only find targetable mutations but also identify, with high accuracy, the actual organ in which the tumor originated, using deep learning and machine learning models.

  • With innovative methods developed by global testing companies that perform liquid biopsy-based epigenomic assessment, a cancer's tissue of origin can be determined from a single blood sample.
  • For example, Guardant's methylation-based next-generation sequencing test, which works on blood (using plasma cell-free DNA), can determine the origin of 12 different cancer types by examining more than 2,000 cancer-specific differentially methylated genomic regions.
  • Caris Life Sciences, another leading CGP testing company, has developed a clinically validated artificial intelligence (AI) tool called GPSai that uses WES and WTS data to predict the tissue in which the tumor originated.
  • Trained on the genomic and transcriptomic (RNA) profiles of 201,612 cases in the company's own database, this AI-supported oncology model can successfully identify the tissue of origin in 84% of cancer of unknown primary cases and match them to the appropriate indication.

2. Molecularly Targeted Therapy Versus Standard Chemotherapy

The Phase II CUPISCO study, published in the Journal of Clinical Oncology, has provided strong data supporting the importance of mapping the tumor's genetics in metastatic cancer cases and of guiding treatment through Molecular Tumor Boards.

  • In patients with unfavorable cancer of unknown primary whose disease had been brought under control with initial chemotherapy, continuing standard platinum-based chemotherapy was compared with receiving molecularly guided therapy (MGT) based on comprehensive genetic test results.
  • In patients who received molecularly guided therapy, median progression-free survival (the time during which the disease does not progress) was 6.1 months, while in the group that simply continued chemotherapy it remained at 4.4 months.
  • In terms of overall survival, median survival rose from 12.8 months in the chemotherapy arm to 15.2 months with molecular therapies.
  • The most striking finding of the study was seen in patients in whom genetic mapping revealed an "actionable" mutation.
  • In this group, progression-free survival reached 8.2 months in patients receiving molecular therapy, whereas in those who received standard chemotherapy despite having an actionable mutation it remained at only 5.5 months.

3. Our Clinic's Perspective: From Single-Target Therapy to a Combination Approach

Although the CUPISCO study clearly demonstrates the benefit that Molecular Tumor Board decisions and targeted therapies provide compared with standard chemotherapy, the molecular strategy applied in the study was largely based on a "one target, one drug" logic.

Cancer cells have highly complex escape routes and, when a single genetic pathway is blocked, they can quickly produce different resistance clones. As we emphasized in one of our earlier blog posts, where we discussed targeted combination therapies in detail, rational combination approaches that strike several genetic weaknesses of the tumor at the same time deliver far superior and more durable results in breaking cellular resistance. Based on our clinical experience and assessment, if combination therapies aimed at molecular targets had been designed in the CUPISCO study instead of single drugs, the survival advantage achieved could have been considerably more pronounced.

4. The Critical Role of the Molecular Tumor Board (MTB) in Clinical Decisions

Simply reading the vast amount of data obtained from AI algorithms that support cancer diagnosis, from epigenomic data and from broad CGP panels is not enough; real success lies in correctly interpreting this genetic code and choosing the right treatment strategy to overcome all of the cancer's resistance mechanisms.

At our clinic, in cases of cancer of unknown primary, our Molecular Tumor Board (MTB), made up of 6 specialist cancer geneticists and experienced oncologists, does not confine our patients to rigid protocols. Using the most advanced validated tests available worldwide in this field, it creates the plan best suited to the patient's biology as decoded through artificial intelligence and comprehensive analyses. Thanks to this approach, which does not exclude chemotherapy but uses it strategically only where it is rational, and with the deep experience we have gained from more than 2,500 cases, we continue to develop the most accurate, effective and personalized molecular solutions for each of our patients, even in the most challenging cancer types.

To learn more about identifying the tissue of origin and personalized treatment planning in cancer of unknown primary, you can contact our clinic.

Sources and Further Reading

  • Detecting the Signal of Origin via Blood Methylation (AACR 2025): Forouzmand, E., He, Y., Gittelman, R., Selewa, A., Singer, M., Tsai, J., ... & Chudova, D. (2025). A novel methylation-based classifier to identify cancer signal of origin using blood based testing [Abstract 6365]. Proceedings of the American Association for Cancer Research Annual Meeting 2025; Cancer Res 2025;85(8_Suppl_1).
  • CUPISCO Study: Krämer, A., Bochtler, T., Pauli, C., Shiu, K.-K., Cook, N., Janoski de Menezes, J., ... & Mileshkin, L. (2026). Molecularly Guided Therapy Versus Continued Chemotherapy in Unfavorable Cancer of Unknown Primary: Updated Efficacy and Safety From the Randomized, Phase II CUPISCO Study. Journal of Clinical Oncology.
  • GPSai (Caris Life Sciences): Ghani, H., Helmstetter, A., Ribeiro, J. R., Maney, T., Rock, S., Feldman, R. A., ... & Oberley, M. J. (2025). GPSai: A Clinically Validated AI Tool for Tissue of Origin Prediction during Routine Tumor Profiling. Cancer Research Communications, 5(9), 1477-1489.