Matija Snuderl

Deep learning-based classifier for carcinoma of unknown primary using methylation quantitative trait loci

Abstract Cancer of unknown primary (CUP) constitutes between 2% and 5% of human malignancies and is among the most common causes of cancer death in the United States. Brain metastases are often the first clinical presentation of CUP; despite extensive pathological and imaging studies, 20%-45% of CUP are never assigned a primary site. DNA methylation array profiling is a reliable method for tumor classification but tumor-type-specific classifier development requires many reference samples. This is difficult to accomplish for CUP as many cases are never assigned a specific diagnosis. Recent studies identified subsets of methylation quantitative trait loci (mQTLs) unique to specific organs, which could help increase classifier accuracy while requiring fewer samples. We performed a retrospective genome-wide methylation analysis of 759 carcinoma samples from formalin-fixed paraffin-embedded tissue samples using Illumina EPIC array. Utilizing mQTL specific for breast, lung, ovarian/gynecologic, colon, kidney, or testis (BLOCKT) (185k total probes), we developed a deep learning-based methylation classifier that achieved 93.12% average accuracy and 93.04% average F1-score across a 10-fold validation for BLOCKT organs. Our findings indicate that our organ-based DNA methylation classifier can assist pathologists in identifying the site of origin, providing oncologists insight on a diagnosis to administer appropriate therapy, improving patient outcomes.

DNA Methylation Profiling Classifies and Reveals Origin of Gynecologic Central Nervous System-Like Tumors

Gynecologic neuroectodermal tumors either exhibit central nervous system (CNS) differentiation (CNS-like) or represent Ewing sarcoma (EWS), which lacks CNS features and harbors FET-ETS gene fusions. DNA methylation profiling reclassified CNS primitive neuroectodermal tumors into common CNS neoplasms or embryonal tumors with specific epigenetic/genetic characteristics. Its utility in classifying gynecologic neuroectodermal tumors is unknown. Whole-genome DNA methylation profiling was performed on 26 gynecologic neuroectodermal tumors (22 CNS-like tumors, 4 EWS) arising in the ovary, paratubal soft tissue, uterus, and vulva, which were classified by using sarcoma and CNS tumor DNA methylation classifiers. Sarcoma-related gene fusions were confirmed by fluorescence in situ hybridization or targeted RNA next-generation sequencing. Tumor-only whole-exome sequencing (WES) was performed in 13 cases. Copy number alterations and zygosity were inferred from DNA methylation array and WES data. Methylation abnormalities associated with imprinting were examined. The sarcoma methylation classifier identified EWS (n = 3) and high-grade endometrial stromal sarcoma (n = 1), confirmed by fluorescence in situ hybridization or next-generation sequencing detection of EWSR1 and YWHAE rearrangements, respectively. The remaining CNS-like tumors were classified by DNA methylation with positive/valid (n = 4), indeterminate (n = 9), and negative (n = 9) scores at the family level. Methylation subclasses included teratoma; embryonal tumor with multilayered rosettes, atypical; medulloblastoma, SHH-activated, subtype 3; medulloblastoma, group 3; intraocular medulloepithelioma; supratentorial ependymoma, ZFTA::RELA fused, subclass A; and diffuse pediatric-type high-grade glioma, MYCN subtype. Male biological sex was predicted in 54% of methylation-confirmed CNS-like tumors and none of the sarcomas. Among CNS-like tumors, copy number analyses identified genome-wide chromosomal gains and losses, and WES revealed genome-wide allelic imbalance suggestive of genome-wide duplications. Epigenetic imprinting analyses showed increased paternal or maternal imprinting signal across multiple chromosomes, suggesting uniparental duplication. DNA methylation profiling successfully classified gynecologic neuroectodermal tumors as known CNS tumors or sarcoma entities. Epigenetic and exomic studies indicate a male genome and increased maternal allelic contribution in CNS-like tumors, suggesting development via conception or chimerism.