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

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Cancers present with diverse histologies and origins, complicating diagnosis.
• Current diagnostic methods rely on morphology and immunohistochemistry, which can be ambiguous.
• Cancers of Unknown Primary (CUP) have poor outcomes and limited therapeutic options.

Purpose of the Study:

• To describe and validate the Tempus Tumor Origin (Tempus TO) assay.
• To assess the accuracy of an RNA-sequencing-based machine learning classifier for cancer subtyping.
• To evaluate the assay's utility in diagnosing cancers of unknown primary.

Main Methods:

• Development of an RNA-sequencing-based machine learning classifier (Tempus TO) to distinguish between 68 cancer subtypes.
• Validation of the model using a large cohort of samples with known diagnoses (9210 total samples).
• Assessment of model accuracy on both retrospective and prospective sample sets.

Main Results:

• The Tempus TO model achieved 91% accuracy in classifying cancer subtypes across a large, diverse sample set.
• Evaluation on a cohort of cancers of unknown primary (CUP) demonstrated that the model identified established genomic-subtype associations.
• The assay shows promise in differentiating between various cancer origins.

Conclusions:

• The Tempus TO assay is a highly accurate tool for identifying cancer subtypes.
• This assay can aid in the diagnosis of cancers of unknown primary (CUP).
• Integrating diagnostic prediction tests with genomic variant analysis may improve therapeutic strategies for CUP patients.