Validation of a Transcriptome-Based Assay for Classifying Cancers of Unknown Primary Origin

• DOI: 10.1007/s40291-023-00650-5
• Publication Date: 2023-04-26T08:02:32Z
• Authors: Jackson Michuda; Alessandra Breschi; Joshuah Kapilivsky; Kabir Manghnani; Calvin McCarter; Adam J. Hockenberry; Brittany Mineo; Catherine Igartua; Joel T. Dudley; Martin C. Stumpe; Nike Beaubier; Maryam Shirazi; Ryan Jones; Elizabeth Morency; Kim Blackwell; Justin Guinney; Kyle A. Beauchamp; Timothy Taxter
• Keywords: 0301 basic medicine; 03 medical and health sciences; 0302 clinical medicine; Gene Expression Profiling; Humans; Neoplasms, Unknown Primary; Original Research Article; Genomics; Transcriptome; 3. Good health; Retrospective Studies

Cancers assume a variety of distinct histologies, and may originate from myriad sites including solid organs, hematopoietic cells, connective tissue. Clinical decision-making based on consensus guidelines such as the National Comprehensive Cancer Network (NCCN) is often predicated specific histologic anatomic diagnosis, supported by clinical features pathologist interpretation morphology immunohistochemical (IHC) staining patterns. However, in patients with nonspecific morphologic IHC findings—in addition to ambiguous presentations recurrence versus new primary—a definitive diagnosis not be possible, resulting patient being categorized having cancer unknown primary (CUP). Therapeutic options outcomes are poor for CUP, median survival 8–11 months. Here, we describe validate Tempus Tumor Origin (Tempus TO) assay, an RNA-sequencing-based machine learning classifier capable discriminating between 68 clinically relevant subtypes. Model accuracy was assessed using and/or metastatic samples known subtype. We show that TO model 91% accurate when both retrospectively held out cohort set sequenced after freeze collectively contained 9210 total diagnoses. When evaluated CUPs, recapitulated established associations genomic alterations Combining diagnostic prediction tests (e.g., sequencing-based variant reporting xT) expand therapeutic cancers or uncertain histology.