Unusual Morphologic Presentation of Perineural Spread From Cutaneous Squamous Cell Carcinoma: Diagnosis Aided by Comprehensive Molecular Analysis and Machine Learning.

Neoplasms of unknown primary frequently pose a diagnostic challenge due to their nonspecific morphological and immunohistochemical features. Definitive classification of these neoplasms has a profound impact on treatment decisions. Mutational and gene expression profiling can provide diagnostic and prognostic information in these challenging cases. We present a case of pontine and cranial nerve lesions in an elderly male with no clinically identifiable index lesion at the time of presentation. The lesion's morphology and immunoprofile did not provide a definitive diagnosis. The whole-exome and transcriptome sequencing identified a UV signature confirming the tumor's cutaneous origin. In addition, pathogenic mutations in multiple genes, including those frequently associated with squamous cell carcinoma (e.g., NOTCH1), were identified. The molecular data was also analyzed by "Caris MI GPSai," a machine-learning algorithm that compares the neoplasm's gene expression and mutational profile against an extensive reference database of genomic and transcriptomic alterations observed in various neoplasms. This predicted the cancer to be cutaneous squamous cell carcinoma with a 66% probability, enabling appropriate treatment for the patient. This case highlights the deceptive morphology of cutaneous squamous cell carcinoma with perineural spread and demonstrates how molecular profiling with machine learning can aid in achieving a definitive diagnosis.