At Veracyte, each thyroid nodule fine needle aspiration (FNA) sample we receive undergoes whole-transcriptome analysis, meaning that over 21,000 genes and 425,000 probes are examined across the transcriptome1. The result is a tremendous amount of data generated on every nodule tested. This data forms the foundation used to perform the Afirma Genomic Sequencing Classifier (GSC), which is comprised of multiple classifiers that work in sequence to provide the binary benign or suspicious molecular result of an indeterminate thyroid nodule, and any available variant or fusion information if detected.2,3
With Afirma GSC, samples first pass through a series of classifiers designed to detect rare neoplasms and lesions, such as parathyroid tumors or medullary thyroid carcinoma. If none are detected, the sample will then proceed to the test’s “ensemble” model: The foundational classifier that drives Afirma GSC and leverages multiple advanced machine learning algorithms to provide the binary benign or suspicious result.
The Hürthle Cell and Hürthle Neoplasm Indexes are designed to address the challenge of classifying oncocytic (Hürthle) cells as either high or low risk and have enabled Afirma GSC to reclassify significantly more oncocytic cell lesions as benign as compared to the original Afirma GEC test.4-7
At this stage, the nodule has been classified as benign or suspicious. Forty-four percent of indeterminate nodules reported as suspicious carry a variant or fusion that may further refine the risk of malignancy.2 This is where Xpression Atlas, the largest thyroid gene and fusion panel available, may provide more granular information on which variants and fusions are present for more personalized treatment.3,8
All of these components make up Afirma GSC, which has helped to prevent unnecessary surgeries or personalized treatment for over 250,000 patients.8 But since Afirma GSC represents approximately 10,000 out of the 21,000 genes sequenced, this only scratches the surface of the data that Afirma GSC can potentially deliver.2
Outside of this data used for Afirma GSC, there is the potential that any additional genomic information can be used for further research into molecular diagnostics. A great example of this was demonstrated at ENDO 2023, where Dr. Allan Golding presented the oral abstract “mRNA Expression Based Tumor Behavior Prediction Models In Thyroid Nodules,” which leveraged RNA-seq data with machine learning to develop thyroid tumor behavior signatures regarding invasion and lymph node metastases.9
At Veracyte, we’re excited about the potential discoveries waiting to be unlocked through the whole-transcriptome Afirma testing platform. Afirma GSC has primarily allowed patients to avoid unnecessary surgery and has helped guide surgical decision-making.2,3 Now, we are on the cusp of providing novel research tools moving forward.