Background
The unknown primary in cervical cancer of unknown primary (cCUP) results in invasive diagnostic and unspecific treatment options for patients. The most common histological subtype, squamous cell carcinoma, can stem from various tumor primary sites including the oral cavity, oropharynx, larynx, facial skin, lungs, and esophagus. DNA methylation profiles are highly tissue specific and have been successfully used to classify tissue origin.
We aimed to implement a SVM based classifier with publicly available DNA methylation profiles of commonly cervically metastasizing squamous cell carcinomas (n=1103) in order to compare the tissue of origin of own squamous cell cCUP samples (n=28). Methylation analysis was performed with Infinium MethylationEPIC v1.0 BeadChip by Illumina.
Results
The highest overall accuracy of tested classifiers was achieved by support vector machine algorithm with 87%. Sqamous cell cCUP samples on DNA methylation level resembled squamous cell carcinomas commonly metastasizing into cervical lymph nodes.
The most frequently predicted entity was oral cavity (11 39%) followed by oropharynx and larynx (both 7, 25%), skin (2, 7%) and esophagus (1, 4%). These rates did not differ from expected distribution of lymph node metastases in literature.
In tSNE clustering, cCUP samples graphically distributed into three different groups corresponding to clusters of cervically metastasizing cancer entities.
Conclusions
On DNA Methylation level cCUP is comparable to cancer entities that commonly metastasize to cervical lymph nodes. Our SVM based classifier can accurately predict these cancers and could greatly reduce invasiveness of cCUP diagnostics and therapy after clinical validation.