Establishment of a Machine Learning Model for the Risk Assessment of Perineural Invasion in Head and Neck Squamous Cell Carcinoma.

Perineural invasion is a prevalent pathological finding in head and neck squamous cell carcinoma and a risk factor for unfavorable survival. An adequate diagnosis of perineural invasion by pathologic examination is limited due to the availability of tumor samples from surgical resection, which can arise in cases of definitive nonsurgical treatment. To address this medical need, we established a random forest prediction model for the risk assessment of perineural invasion, including occult perineural invasion, and characterized distinct cellular and molecular features based on our new and extended classification. RNA sequencing data of head and neck squamous cell carcinoma from The Cancer Genome Atlas were used as a training cohort to identify differentially expressed genes that are associated with perineural invasion. A random forest classification model was established based on these differentially expressed genes and was validated by inspection of H&E-stained whole image slides. Differences in epigenetic regulation and the mutational landscape were detected by an integrative analysis of multiomics data and single-cell RNA-sequencing data were analyzed. We identified a 44-gene expression signature related to perineural invasion and enriched for genes mainly expressed in cancer cells according to single-cell RNA-sequencing data. A machine learning model was trained based on the expression pattern of the 44-gene set with the unique feature to predict occult perineural invasion. This extended classification model enabled a more accurate analysis of alterations in the mutational landscape and epigenetic regulation by DNA methylation as well as quantitative and qualitative differences in the cellular composition in the tumor microenvironment between head and neck squamous cell carcinoma with or without perineural invasion. In conclusion, the newly established model could not only complement histopathologic examination as an additional diagnostic tool but also guide the identification of new drug targets for therapeutic intervention in future clinical trials with head and neck squamous cell carcinoma patients at a higher risk for treatment failure due to perineural invasion.

A Novel Subgroup of UCHL1-Related Cancers Is Associated with Genomic Instability and Sensitivity to DNA-Damaging Treatment.

PURPOSE: Identification of molecularly-defined cancer subgroups and targeting tumor-specific vulnerabilities have a strong potential to improve treatment response and patient outcomes but remain an unmet challenge of high clinical relevance, especially in head and neck squamous cell carcinoma (HNSC). EXPERIMENTAL DESIGN: We established a UCHL1-related gene set to identify and molecularly characterize a UCHL1-related subgroup within TCGA-HNSC by integrative analysis of multi-omics data. An extreme gradient boosting model was trained on TCGA-HNSC based on GSVA scores for gene sets of the MSigDB to robustly predict UCHL1-related cancers in other solid tumors and cancer cell lines derived thereof. Potential vulnerabilities of UCHL1-related cancer cells were elucidated by an in-silico drug screening approach. RESULTS: We established a 497-gene set, which stratified the TCGA-HNSC cohort into distinct subgroups with a UCHL1-related or other phenotype. UCHL1-related HNSC were characterized by higher frequencies of genomic alterations, which was also evident for UCHL1-related cancers of other solid tumors predicted by the classification model. These data indicated an impaired maintenance of genomic integrity and vulnerability for DNA-damaging treatment, which was supported by a favorable prognosis of UCHL1-related tumors after radiotherapy, and a higher sensitivity of UCHL1-related cancer cells to irradiation or DNA-damaging compounds (e.g., Oxaliplatin). CONCLUSION: Our study established UCHL1-related cancers as a novel subgroup across most solid tumor entities with a unique molecular phenotype and DNA-damaging treatment as a specific vulnerability, which requires further proof-of-concept in pre-clinical models and future clinical trials.

Establishment of a Plasticity-Associated Risk Model Based on a SOX2- and SOX9-Related Gene Set in Head and Neck Squamous Cell Carcinoma.

Recent studies highlighted SOX2 and SOX9 as key determinants for cancer-cell plasticity and demonstrated that cisplatin-induced adaptation in oral squamous cell carcinoma (SCC) is acquired by an inverse regulation of both transcription factors. However, the association between SOX2/SOX9-related genetic programs with risk factors and genetic or epigenetic alterations in primary head and neck SCC (HNSCC), and their prognostic value is largely unknown.Here, we identified differentially-expressed genes (DEG) related to SOX2 and SOX9 transcription in The Cancer Genome Atlas (TCGA)-HNSC, which enable clustering of patients into groups with distinct clinical features and survival. A prognostic risk model was established by LASSO Cox regression based on expression patterns of DEGs in TCGA-HNSC (training cohort), and was confirmed in independent HNSCC validation cohorts as well as other cancer cohorts from TCGA. Differences in the mutational landscape among risk groups of TCGA-HNSC demonstrated an enrichment of truncating NSD1 mutations for the low-risk group and elucidated DNA methylation as modulator of SOX2 expression. Gene set variation analysis (GSVA) revealed differences in several oncogenic pathways among risk groups, including upregulation of gene sets related to oncogenic KRAS signaling for the high-risk group. Finally, in silico drug screen analysis revealed numerous compounds targeting EGFR signaling with significantly lower efficacy for cancer cell lines with a higher risk phenotype, but also indicated potential vulnerabilities. IMPLICATIONS: The established risk model identifies patients with primary HNSCC, but also other cancers at a higher risk for treatment failure, who might benefit from a therapy targeting SOX2/SOX9-related gene regulatory and signaling networks.