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

Assessing the Reporting Quality of Machine Learning Algorithms in Head and Neck Oncology

OBJECTIVE

This study aimed to assess reporting quality of machine learning (ML) algorithms in the head and neck oncology literature using the TRIPOD-AI criteria.

DATA SOURCES

A comprehensive search was conducted using PubMed, Scopus, Embase, and Cochrane Database of Systematic Reviews, incorporating search terms related to "artificial intelligence," "machine learning," "deep learning," "neural network," and various head and neck neoplasms.

REVIEW METHODS

Two independent reviewers analyzed each published study for adherence to the 65-point TRIPOD-AI criteria. Items were classified as "Yes," "No," or "NA" for each publication. The proportion of studies satisfying each TRIPOD-AI criterion was calculated. Additionally, the evidence level for each study was evaluated independently by two reviewers using the Oxford Centre for Evidence-Based Medicine (OCEBM) Levels of Evidence. Discrepancies were reconciled through discussion until consensus was reached.

RESULTS

The study highlights the need for improvements in ML algorithm reporting in head and neck oncology. This includes more comprehensive descriptions of datasets, standardization of model performance reporting, and increased sharing of ML models, data, and code with the research community. Adoption of TRIPOD-AI is necessary for achieving standardized ML research reporting in head and neck oncology.

CONCLUSION

Current reporting of ML algorithms hinders clinical application, reproducibility, and understanding of the data used for model training. To overcome these limitations and improve patient and clinician trust, ML developers should provide open access to models, code, and source data, fostering iterative progress through community critique, thus enhancing model accuracy and mitigating biases.

LEVEL OF EVIDENCE

NA Laryngoscope, 2024.

Dendritic cell-related hub genes in head-and-neck squamous cell carcinoma: implications for prognosis and immunotherapy

Background

In the context of head-and-neck squamous cell carcinoma (HNSCC), dendritic cells (DCs) assume pivotal responsibilities, acting as architects of antigen presentation and conductors of immune checkpoint modulation. In this study, we aimed to identify hub genes associated with DCs in HNSCC and explore their prognostic significance and implications for immunotherapy.

Methods

Integrated clinical datasets from The Cancer Genome Atlas (TCGA)-HNSCC and GSE65858 cohorts underwent meticulous analysis. Employing weighted gene co-expression network analysis (WGCNA), we delineated candidate genes pertinent to DCs. Through the application of random survival forest and least absolute shrinkage and selection operator (LASSO) Cox's regression, we derived key genes of significance. Lisa (epigenetic Landscape In Silico deletion Analysis and the second descendent of MARGE) highlighted transcription factors, with Dual-luciferase assays confirming their regulatory role. Furthermore, immunotherapeutic sensitivity was assessed utilizing the Tumor Immune Dysfunction and Exclusion online tool.

Results

This study illuminated the functional intricacies of HNSCC DC subsets to tailor innovative therapeutic strategies. We leveraged clinical data from the TCGA-HNSCC and GSE65858 cohorts. We subjected the data to advanced analysis, including WGCNA, which revealed 222 DC-related candidate genes. Following this, a discerning approach utilizing random survival forest analysis and LASSO Cox's regression unveiled seven genes associated with the prognostic impact of DCs, notably ACP2 and CPVL, associated with poor overall survival. Differential gene expression analysis between ACP2 + and ACP2 - DC cells revealed 208 differential expressed genes. Lisa analysis identified the top five significant transcription factors as STAT1, SPI1, SMAD1, CEBPB, and IRF1. The correlation between STAT1 and ACP2 was confirmed through quantitative reverse transcription polymerase chain reaction (qRT-PCR) and Dual-luciferase assays in HEK293T cells. Additionally, TP53 and FAT1 mutations were more common in high-risk DC subgroups. Importantly, the sensitivity to immunotherapy differed among the risk clusters. The low-risk cohorts were anticipated to exhibit favorable responses to immunotherapy, marked by heightened expressions of immune system-related markers. In contrast, the high-risk group displayed augmented proportions of immunosuppressive cells, suggesting a less conducive environment for immunotherapeutic interventions.

Conclusions

Our research may yield a robust DC-based prognostic system for HNSCC; this will aid personalized treatment and improve clinical outcomes as the battle against this challenging cancer continues.

Context-Dependent Regulation of Peripheral Nerve Abundance by the PI3K Pathway in the Tumor Microenvironment of Head and Neck Squamous Cell Carcinoma

Recent studies have highlighted neurons and their associated Schwann cells (SCs) as key regulators of cancer development. However, the mode of their interaction with tumor cells or other components of the tumor microenvironment (TME) remains elusive. We established an SC-related 43-gene set as a surrogate for peripheral nerves in the TME. Head and neck squamous cell carcinoma (HNSCC) from The Cancer Genome Atlas (TCGA) were classified into low, intermediate and high SC score groups based on the expression of this gene set. Perineural invasion (PNI) and TGF-β signaling were hallmarks of SChigh tumors, whereas SClow tumors were enriched for HPV16-positive OPSCC and higher PI3K-MTOR activity. The latter activity was partially explained by a higher frequency of PTEN mutation and PIK3CA copy number gain. The inverse association between PI3K-MTOR activity and peripheral nerve abundance was context-dependent and influenced by the TP53 mutation status. An in silico drug screening approach highlighted the potential vulnerabilities of HNSCC with variable SC scores and predicted a higher sensitivity of SClow tumors to DNA topoisomerase inhibitors. In conclusion, we have established a tool for assessing peripheral nerve abundance in the TME and provided new clinical and biological insights into their regulation. This knowledge may pave the way for new therapeutic strategies and impart proof of concept in appropriate preclinical models.

Tumor Neurobiology in the Pathogenesis and Therapy of Head and Neck Cancer

The neurobiology of tumors has attracted considerable interest from clinicians and scientists and has become a multidisciplinary area of research. Neural components not only interact with tumor cells but also influence other elements within the TME, such as immune cells and vascular components, forming a polygonal relationship to synergistically facilitate tumor growth and progression. This review comprehensively summarizes the current state of the knowledge on nerve-tumor crosstalk in head and neck cancer and discusses the potential underlying mechanisms. Several mechanisms facilitating nerve-tumor crosstalk are covered, such as perineural invasion, axonogenesis, neurogenesis, neural reprogramming, and transdifferentiation, and the reciprocal interactions between the nervous and immune systems in the TME are also discussed in this review. Further understanding of the nerve-tumor crosstalk in the TME of head and neck cancer may provide new nerve-targeted treatment options and help improve clinical outcomes for patients.