Prognostic modeling of oral cancer by gene profiles and clinicopathological co-variables

Exploring patient stratification in head and neck squamous cell carcinoma using machine learning techniques: Preliminary results

BACKGROUND

Head and Neck Squamous Cell Carcinoma (HNSCC) presents a significant challenge in oncology due to its inherent heterogeneity. Traditional staging systems, such as TNM (Tumor, Node, Metastasis), provide limited information regarding patient outcomes and treatment responses. There is a need for a more robust system to improve patient stratification.

METHOD

In this study, we utilized advanced statistical techniques to explore patient stratification beyond the limitations of TNM staging. A comprehensive dataset, including clinical, radiomic, genomic, and pathological data, was analyzed. The methodology involved correlation analysis of variable pairs and triples, followed by clustering techniques.

RESULTS

The analysis revealed that HNSCC subpopulations exhibit distinct characteristics, which challenge the conventional one-size-fits-all approach.

CONCLUSION

This study underscores the potential for personalized treatment strategies based on comprehensive patient profiling, offering a pathway towards more individualized therapeutic interventions.

An Approach for Systems-Level Understanding of Prostate Cancer from High-Throughput Data Integration to Pathway Modeling and Simulation

To understand complex diseases, high-throughput data are generated at large and multiple levels. However, extracting meaningful information from large datasets for comprehensive understanding of cell phenotypes and disease pathophysiology remains a major challenge. Despite tremendous advances in understanding molecular mechanisms of cancer and its progression, current knowledge appears discrete and fragmented. In order to render this wealth of data more integrated and thus informative, we have developed a GECIP toolbox to investigate the crosstalk and the responsible genes'/proteins' connectivity of enriched pathways from gene expression data. To implement this toolbox, we used mainly gene expression datasets of prostate cancer, and the three datasets were GSE17951, GSE8218, and GSE1431. The raw samples were processed for normalization, prediction of differentially expressed genes, and the prediction of enriched pathways for the differentially expressed genes. The enriched pathways have been processed for crosstalk degree calculations for which number connections per gene, the frequency of genes in the pathways, sharing frequency, and the connectivity have been used. For network prediction, protein-protein interaction network database FunCoup2.0 was used, and cytoscape software was used for the network visualization. In our results, we found that there were enriched pathways 27, 45, and 22 for GSE17951, GSE8218, and GSE1431, respectively, and 11 pathways in common between all of them. From the crosstalk results, we observe that focal adhesion and PI3K pathways, both experimentally proven central for cellular output upon perturbation of numerous individual/distinct signaling pathways, displayed highest crosstalk degree. Moreover, we also observe that there were more critical pathways which appear to be highly significant, and these pathways are HIF1a, hippo, AMPK, and Ras. In terms of the pathways' components, GSK3B, YWHAE, HIF1A, ATP1A3, and PRKCA are shared between the aforementioned pathways and have higher connectivity with the pathways and the other pathway components. Finally, we conclude that the focal adhesion and PI3K pathways are the most critical pathways, and since for many other pathways, high-rank enrichment did not translate to high crosstalk degree, the global impact of one pathway on others appears distinct from enrichment.

Prognostic signature related to the immune environment of oral squamous cell carcinoma

Oral squamous cell carcinoma (OSCC) prognosis remains poor. Here we aimed to identify an effective prognostic signature for predicting the survival of patients with OSCC. Gene-expression and clinical data were obtained from the Cancer Genome Atlas database. Immune microenvironment-associated genes were identified using bioinformatics. Subtype and risk-score analyses were performed for these genes. Kaplan–Meier analysis and immune cell infiltration level were explored in different subtypes and risk-score groups. The prognostic ability, independent prognosis, and clinical features of the risk score were assessed. Furthermore, immunotherapy response based on the risk score was explored. Finally, a conjoint analysis of the subtype and risk-score groups was performed to determine the best prognostic combination. We found 11 potential prognostic genes and constructed a risk-score model. The subtype cluster 2 and a high-risk group showed the worst overall survival; differences in survival status might be due to the different immune cell infiltration levels. The risk score showed good performance, independent prognostic value, and valuable clinical application. Higher risk scores showed higher Tumor Immune Dysfunction and Exclusion scores, indicating that patients with a high-risk score were less likely to benefit from immunotherapy. Finally, conjoint analysis for the subgroups and risk groups showed the best predictive ability.

Semi-supervised empirical Bayes group-regularized factor regression

The features in a high-dimensional biomedical prediction problem are often well described by low-dimensional latent variables (or factors). We use this to include unlabeled features and additional information on the features when building a prediction model. Such additional feature information is often available in biomedical applications. Examples are annotation of genes, metabolites, or p-values from a previous study. We employ a Bayesian factor regression model that jointly models the features and the outcome using Gaussian latent variables. We fit the model using a computationally efficient variational Bayes method, which scales to high dimensions. We use the extra information to set up a prior model for the features in terms of hyperparameters, which are then estimated through empirical Bayes. The method is demonstrated in simulations and two applications. One application considers influenza vaccine efficacy prediction based on microarray data. The second application predicts oral cancer metastasis from RNAseq data.

A robust and interpretable gene signature for predicting the lymph node status of primary T1/T2 oral cavity squamous cell carcinoma

Oral cavity squamous cell carcinoma (OSCC) affects more than 30 000 individuals in the United States annually, with smoking and alcohol consumption being the main risk factors. Management of early-stage tumors usually includes surgical resection followed by postoperative radiotherapy in certain cases. The cervical lymph nodes (LNs) are the most common site for local metastasis, and elective neck dissection is usually performed if the primary tumor thickness is greater than 3.5 mm. However, postoperative histological examination often reveals that many patients with early-stage disease are negative for neck nodal metastasis, posing a pressing need for improved risk stratification to either avoid overtreatment or prevent the disease progression. To this end, we aimed to identify a primary tumor gene signature that can accurately predict cervical LN metastasis in patients with early-stage OSCC. Using gene expression profiles from 189 samples, we trained K-top scoring pairs models and identified six gene pairs that can distinguish primary tumors with nodal metastasis from those without metastasis. The signature was further validated on an independent cohort of 35 patients using real-time polymerase chain reaction (PCR) in which it achieved an area under the receiver operating characteristic (ROC) curve and accuracy of 90% and 91%, respectively. These results indicate that such signature holds promise as a quick and cost effective method for detecting patients at high risk of developing cervical LN metastasis, and may be potentially used to guide the neck treatment regimen in early-stage OSCC.

Shooting at Moving and Hidden Targets-Tumour Cell Plasticity and the Notch Signalling Pathway in Head and Neck Squamous Cell Carcinomas

Head and Neck Squamous Cell Carcinoma (HNSCC) is often aggressive, with poor response to current therapies in approximately 40-50% of the patients. Current therapies are restricted to operation and irradiation, often combined with a small number of standard-of-care chemotherapeutic drugs, preferentially for advanced tumour patients. Only very recently, newer targeted therapies have entered the clinics, including Cetuximab, which targets the EGF receptor (EGFR), and several immune checkpoint inhibitors targeting the immune receptor PD-1 and its ligand PD-L1. HNSCC tumour tissues are characterized by a high degree of intra-tumour heterogeneity (ITH), and non-genetic alterations that may affect both non-transformed cells, such as cancer-associated fibroblasts (CAFs), and transformed carcinoma cells. This very high degree of heterogeneity likely contributes to acquired drug resistance, tumour dormancy, relapse, and distant or lymph node metastasis. ITH, in turn, is likely promoted by pronounced tumour cell plasticity, which manifests in highly dynamic and reversible phenomena such as of partial or hybrid forms of epithelial-to-mesenchymal transition (EMT), and enhanced tumour stemness. Stemness and tumour cell plasticity are strongly promoted by Notch signalling, which remains poorly understood especially in HNSCC. Here, we aim to elucidate how Notch signal may act both as a tumour suppressor and proto-oncogenic, probably during different stages of tumour cell initiation and progression. Notch signalling also interacts with numerous other signalling pathways, that may also have a decisive impact on tumour cell plasticity, acquired radio/chemoresistance, and metastatic progression of HNSCC. We outline the current stage of research related to Notch signalling, and how this pathway may be intricately interconnected with other, druggable targets and signalling mechanisms in HNSCC.

Adaptive group-regularized logistic elastic net regression

In high-dimensional data settings, additional information on the features is often available. Examples of such external information in omics research are: (i) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$p$\end{document}-values from a previous study and (ii) omics annotation. The inclusion of this information in the analysis may enhance classification performance and feature selection but is not straightforward. We propose a group-regularized (logistic) elastic net regression method, where each penalty parameter corresponds to a group of features based on the external information. The method, termed gren, makes use of the Bayesian formulation of logistic elastic net regression to estimate both the model and penalty parameters in an approximate empirical–variational Bayes framework. Simulations and applications to three cancer genomics studies and one Alzheimer metabolomics study show that, if the partitioning of the features is informative, classification performance, and feature selection are indeed enhanced.

Understanding the role of potential pathways and its components including hypoxia and immune system in case of oral cancer

There are a few biological functions or phenomenon which are universally associated with majority of the cancers and hypoxia and immune systems are among them. Hypoxia often occurs in most of the cancers which helps the cells in adapting different responses with respect to the normal cells which may be the activation of signaling pathways which regulate proliferation, angiogenesis, and cell death. Similar to it, immune signaling pathways are known to play critical roles in cancers. Moreover, there are a number of genes which are known to be associated with these hypoxia and immune system and appear to direct affect the tumor growth and propagations. Cancer is among the leading cause of death and oral cancer is the tenth-leading cause due to cancer death. In this study, we were mainly interested to understand the impact of alteration in the expression of hypoxia and immune system-related genes and their contribution to head and neck squamous cell carcinoma. Moreover, we have collected the genes associated with hypoxia and immune system from the literatures. In this work, we have performed meta-analysis of the gene and microRNA expression and mutational datasets obtained from public database for different grades of tumor in case of oral cancer. Based on our results, we conclude that the critical pathways which dominantly enriched are associated with metabolism, cell cycle, immune system and based on the survival analysis of the hypoxic genes, we observe that the potential genes associated with head and neck squamous cell carcinoma and its progression are STC2, PGK1, P4HA1, HK1, SPIB, ANXA5, SERPINE1, HGF, PFKM, TGFB1, L1CAM, ELK4, EHF, and CDK2.

Contribution of Genomics to the Surgical Management and Study of Oral Cancer

Background

Oral squamous cell carcinoma (OSCC) is the most frequent type of tumor arising from the oral cavity. Surgery is the cornerstone of the treatment of these cancers. Tumor biology has long been overlooked as an important contributor to the outcome of surgical procedures, but recent studies are challenging this concept. Molecular analyses of tumor DNA or RNA provide a rich source of information about the biology of OSCC.

Methods

We searched for relevant articles using PubMed. We examined in particular the prospect of applying molecular methods for minimally invasive exploration of OSCC biology.

Results

We examined five potential applications of genomics to the surgical management and study of OSCC: i) assessing oral potentially malignant lesions; ii) tumor staging prior to surgery; iii) predicting postoperative risk in locally advanced tumors; iv) measuring minimal residual disease and optimizing the longitudinal monitoring of OSCC; and v) predicting the efficacy of medical treatment.

Conclusions

Genomic information can be harnessed in order to identify new biomarkers that could improve the staging, choice of therapy and management of OSCC. The identification of new biomarkers is awaited for better personalization of the surgical treatment of OSCC.

Bioinformatics analysis reveals that ANXA1 and SPINK5 are novel tumor suppressor genes in patients with oral squamous cell carcinoma

BACKGROUND

Oral squamous cell carcinoma (OSCC) is a solid tumor of squamous epithelial origin. Currently, surgery is still the main treatment for OSCC, with radiotherapy and chemotherapy as important adjuvant treatments. However, the problem of poor prognosis of OSCC patients still exists in clinical practice. To explore further potential biomarkers or treatment targets in OSCC patients, this study used a high-throughput gene expression database to study the potential molecular mechanisms of OSCC carcinogenesis.

METHODS

The GEO database related to OSCC was searched and analyzed using GEO2R. Oncomine and the Human Protein Atlas were used to evaluate the expression level of differentially-expressed genes (DEGs). The cBioPortal dataset was used to analyze the mutations of the potential DEGs and patient survival.

RESULTS

Three GEO datasets, GSE146483, GSE138206, and GSE148944, were downloaded and 7 DEGs were found in common in OSCC tissues. Using Oncomine and the Human Protein Atlas, ANXA1, IL1RN, and SPINK5 were decreased in cancer tissues, while protein levels of APOE and IFI35 were increased accordingly. Interestingly, low levels of ANXA1 and SPINKS were associated with the TNM stage of OSCC patients. No mutations in DEGs were found in OSCC patients, based on the cBioPortal dataset. Survival analysis indicated OSCC patients with high MSR1 had poor overall survival (OS), while low expression of CXCR4, ANXA1, IL1RN, and SPINK5 also predicted poor OS in OSCC patients.

CONCLUSIONS

Our findings uncovered 7 potential biomarkers of OSCC patients, with ANXA1 and SPINK5 serving as potential tumor suppressor genes in OSCC.

PSMC2, ORC5 and KRTDAP are specific biomarkers for HPV-negative head and neck squamous cell carcinoma

The prognosis of patients with human papillomavirus (HPV)-negative head and neck squamous cell carcinoma (HNSCC) is poorer than those with HPV-positive HNSCC. The present study aimed to identify novel and specific biomarkers of HPV-negative HNSCC using bioinformatics analysis and associated experiments. The gene expression profiles of HPV-negative HNSCC tissues and corresponding clinical data were downloaded from The Cancer Genome Atlas database and used in a weighted gene co-expression network analysis. Genes in clinically significant co-expression modules were used to construct a protein-protein interaction (PPI) network. The genes demonstrating a high degree score in the PPI network and a high correlation with tumor grade were considered hub genes. The diagnostic value of the hub genes associated with HPV-negative and HPV-positive HNSCC was analyzed using differential expression gene (DEG) analysis, immunohistochemical (IHC) staining and a receiver operating characteristic (ROC) curve analysis. Seven genes [Serrate RNA effector molecule (SRRT), checkpoint kinase 2 (CHEK2), small nuclear ribonucleoprotein polypeptide E (SNRPE), proteasome 26S subunit ATPase 2 (PSMC2), origin recognition complex subunit 5 (ORC5), S100 calcium binding protein A7 and keratinocyte differentiation associated protein (KRTDAP)] were demonstrated to be hub genes in clinically significant co-expression modules. DEG, IHC and ROC curve analyses revealed that SRRT, CHEK2 and SNRPE were significantly upregulated in HPV-negative and HPV-positive HNSCC tissues compared with in adjacent tissues, and these genes demonstrated a high diagnostic value for distinguishing HNSCC tissues. However, PSMC2, ORC5 and KRTDAP were the only differentially expressed genes identified in HPV-negative HNSCC tissues, and these genes demonstrated a high diagnostic value for HPV-negative HNSCC. PSMC2, ORC5 and KRTDAP may therefore serve as novel and specific biomarkers for HPV-negative HNSCC, potentially improving the diagnosis and treatment of patients with HPV-negative HNSCC.

Genome-wide study of salivary miRNAs identifies miR-423-5p as promising diagnostic and prognostic biomarker in oral squamous cell carcinoma

Survival rates of oral squamous cell carcinoma (OSCC) remained substantially unchanged over the last decades; thus, additional prognostic tools are strongly needed. Salivary miRNAs have emerged as excellent non-invasive cancer biomarker candidates, but their association with OSCC prognosis has not been investigated yet. In this study, we analyzed global salivary miRNA expression in OSCC patients and healthy controls, with the aim to define its diagnostic and prognostic potential.

Methods: Saliva was collected from patients with newly diagnosed untreated primary OSCC and healthy controls. Global profiling of salivary miRNAs was carried out through a microarray approach, while signature validation was performed by quantitative real-time PCR (RT-qPCR). A stringent statistical approach for microarray and RT-qPCR data normalization was applied. The diagnostic performance of miRNAs and their correlation with OSCC prognosis were comprehensively analyzed.

Results: In total, 25 miRNAs emerged as differentially expressed between OSCC patients and healthy controls and, among them, seven were significantly associated with disease-free survival (DFS). miR-106b-5p, miR-423-5p and miR-193b-3p were expressed at high levels in saliva of OSCC patients and their combination displays the best diagnostic performance (ROC - AUC = 0.98). Moreover, high expression of miR-423-5p was an independent predictor of poor DFS, when included in multivariate survival analysis with the number of positive lymph nodes - the only significant clinical prognosticator. Finally, we observed a significant decrease in miR-423-5p expression in matched post-operative saliva samples, suggesting its potential cancer-specific origin.

Conclusion: Salivary miRNAs identified in our cohort of patients show to be accurate in OSCC detection and to effectively stratify patients according to their likelihood of relapse. These results, if validated in an independent set of patients, could be particularly promising for screening/follow-up of high-risk populations and useful for preoperative prognostic assessment.

Development of a multiomics database for personalized prognostic forecasting in head and neck cancer: The Big Data to Decide EU Project

Background

Despite advances in treatments, 30% to 50% of stage III‐IV head and neck squamous cell carcinoma (HNSCC) patients relapse within 2 years after treatment. The Big Data to Decide (BD2Decide) project aimed to build a database for prognostic prediction modeling.

Methods

Stage III‐IV HNSCC patients with locoregionally advanced HNSCC treated with curative intent (1537) were included. Whole transcriptomics and radiomics analyses were performed using pretreatment tumor samples and computed tomography/magnetic resonance imaging scans, respectively.

Results

The entire cohort was composed of 71% male (1097)and 29% female (440): oral cavity (429, 28%), oropharynx (624, 41%), larynx (314, 20%), and hypopharynx (170, 11%); median follow‐up 50.5 months. Transcriptomics and imaging data were available for 1284 (83%) and 1239 (80%) cases, respectively; 1047 (68%) patients shared both.

Conclusions

This annotated database represents the HNSCC largest available repository and will enable to develop/validate a decision support system integrating multiscale data to explore through classical and machine learning models their prognostic role.

Outcome prediction of head and neck squamous cell carcinoma by MRI radiomic signatures

OBJECTIVES

Head and neck squamous cell carcinoma (HNSCC) shows a remarkable heterogeneity between tumors, which may be captured by a variety of quantitative features extracted from diagnostic images, termed radiomics. The aim of this study was to develop and validate MRI-based radiomic prognostic models in oral and oropharyngeal cancer.

MATERIALS AND METHODS

Native T1-weighted images of four independent, retrospective (2005-2013), patient cohorts (n = 102, n = 76, n = 89, and n = 56) were used to delineate primary tumors, and to extract 545 quantitative features from. Subsequently, redundancy filtering and factor analysis were performed to handle collinearity in the data. Next, radiomic prognostic models were trained and validated to predict overall survival (OS) and relapse-free survival (RFS). Radiomic features were compared to and combined with prognostic models based on standard clinical parameters. Performance was assessed by integrated area under the curve (iAUC).

RESULTS

In oral cancer, the radiomic model showed an iAUC of 0.69 (OS) and 0.70 (RFS) in the validation cohort, whereas the iAUC in the oropharyngeal cancer validation cohort was 0.71 (OS) and 0.74 (RFS). By integration of radiomic and clinical variables, the most accurate models were defined (iAUC oral cavity, 0.72 (OS) and 0.74 (RFS); iAUC oropharynx, 0.81 (OS) and 0.78 (RFS)), and these combined models outperformed prognostic models based on standard clinical variables only (p < 0.001).

CONCLUSIONS

MRI radiomics is feasible in HNSCC despite the known variability in MRI vendors and acquisition protocols, and radiomic features added information to prognostic models based on clinical parameters.

KEY POINTS

• MRI radiomics can predict overall survival and relapse-free survival in oral and HPV-negative oropharyngeal cancer. • MRI radiomics provides additional prognostic information to known clinical variables, with the best performance of the combined models. • Variation in MRI vendors and acquisition protocols did not influence performance of radiomic prognostic models.

Transcriptomics and Epigenomics in head and neck cancer: available repositories and molecular signatures

For many years, head and neck squamous cell carcinoma (HNSCC) has been considered as a single entity. However, in the last decades HNSCC complexity and heterogeneity have been recognized. In parallel, high-throughput omics techniques had allowed picturing a larger spectrum of the behavior and characteristics of molecules in cancer and a large set of omics web-based tools and informative repository databases have been developed. The objective of the present review is to provide an overview on biological, prognostic and predictive molecular signatures in HNSCC. To contextualize the selected data, our literature survey includes a short summary of the main characteristics of omics data repositories and web-tools for data analyses. The timeframe of our analysis was fixed, encompassing papers published between January 2015 and January 2019. From more than 1000 papers evaluated, 61 omics studies were selected: 33 investigating mRNA signatures, 11 and 13 related to miRNA and other non-coding-RNA signatures and 4 analyzing DNA methylation signatures. More than half of identified signatures (36) had a prognostic value but only in 10 studies selection of a specific anatomical sub-site (8 oral cavity, 1 oropharynx and 1 both oral cavity and oropharynx) was performed. Noteworthy, although the sample size included in many studies was limited, about one-half of the retrieved studies reported an external validation on independent dataset(s), strengthening the relevance of the obtained data. Finally, we highlighted the development and exploitation of three gene-expression signatures, whose clinical impact on prognosis/prediction of treatment response could be high. Based on this overview on omics-related literature in HNSCC, we identified some limits and strengths. The major limits are represented by the low number of signatures associated to DNA methylation and to non-coding RNA (miRNA, lncRNA and piRNAs) and the availability of a single dataset with multiple omics on more than 500 HNSCC (i.e. TCGA). The major strengths rely on the integration of multiple datasets through meta-analysis approaches and on the growing integration among omics data obtained on the same cohort of patients. Moreover, new approaches based on artificial intelligence and informatic analyses are expected to be available in the next future.

Prediction of survival of HPV16-negative, p16-negative oral cavity cancer patients using a 13-gene signature: A multicenter study using FFPE samples

OBJECTIVES

To test the performance of an oral cancer prognostic 13-gene signature for the prediction of survival of patients diagnosed with HPV-negative and p16-negative oral cavity cancer.

MATERIALS AND METHODS

Diagnostic formalin-fixed paraffin-embedded oral cavity cancer tumor samples were obtained from the Fred Hutchinson Cancer Research Center/University of Washington, University of Calgary, University of Michigan, University of Utah, and seven ARCAGE study centers coordinated by the International Agency of Research on Cancer. RNA from 638 Human Papillomavirus (HPV)-negative and p16-negative samples was analyzed for the 13 genes using a NanoString assay. Ridge-penalized Cox regressions were applied to samples randomly split into discovery and validation sets to build models and evaluate the performance of the 13-gene signature in predicting 2-year oral cavity cancer-specific survival overall and separately for patients with early and late stage disease.

RESULTS

Among AJCC stage I/II patients, including the 13-gene signature in the model resulted in substantial improvement in the prediction of 2-year oral cavity cancer-specific survival. For models containing age and sex with and without the 13-gene signature score, the areas under the Receiver Operating Characteristic Curve (AUC) and partial AUC were 0.700 vs. 0.537 (p < 0.001), and 0.046 vs. 0.018 (p < 0.001), respectively. Improvement in predicting prognosis for AJCC stage III/IV disease also was observed, but to a lesser extent.

CONCLUSIONS

If confirmed using tumor samples from a larger number of early stage oral cavity cancer patients, the 13-gene signature may inform personalized treatment of early stage HPV-negative and p16-negative oral cavity cancer patients.

Development and Validation of a Novel and Rapid Molecular Detection Method for High-Risk Human Papillomavirus in Formalin-Fixed, Paraffin-Embedded Tumor Tissue

The most widely applied algorithm for human papillomavirus (HPV) detection in formalin-fixed, paraffin-embedded (FFPE) specimens of oropharyngeal head and neck squamous cell carcinoma (HNSCC) consists of p16INK4A immunostaining followed by PCR-based detection of high-risk HPV DNA on p16INK4A-immunopositive samples. However, in nonoropharyngeal HNSCC this algorithm fails, hampering correct interpretation of the prevalence and prognosis of HPV in these cases. In this study, we developed and validated a molecular HPV detection method for FFPE specimens of oropharyngeal and nonoropharyngeal HNSCC. Sectioning of FFPE blocks was circumvented by using punch biopsies from tumor-enriched regions of FFPE tissue blocks, and combined extraction was applied to obtain high-quality DNA and RNA from the punch biopsy. Next, PCR-based detection of HPV DNA was performed for 15 high-risk HPV types with subsequent detection of E6 mRNA for validation. The combined DNA/RNA FFPE test of tissue cores was assessed in well-characterized cohorts with known HPV status based on earlier work, that is, a cohort of oropharyngeal HNSCC (n = 80) and oral cavity HNSCC (n = 25), and reached an accuracy of 97% and 100%, respectively. In conclusion, our method is rapid, simple, and shows an excellent diagnostic performance for detection of HPV type 16. Ultimately, it can be applied for large cohort studies to determine the etiologic fraction and prognostic implication of HPV in nonoropharyngeal HNSCC.

Molecular Classification of Lymph Node Metastases Subtypes Predict for Survival in Head and Neck Cancer

PURPOSE

In advanced stage head and neck squamous cell cancers (HNSCC), approximately half of the patients with lymph node metastases (LNM) are not cured. Given the heterogeneous outcomes in these patients, we profiled the expression patterns of LNMs to identify the biological factors associated with patient outcomes.Experimental Design: We performed mRNAseq and miRNAseq on 72 LNMs and 29 matched primary tumors from 34 patients with HNSCC. Clustering identified molecular subtypes in LNMs and in primary tumors. Prediction Analysis of Microarrays algorithm identified a 73-gene classifier that distinguished LNM subtypes. Gene-set enrichment analysis identified pathways upregulated in LNM subtypes.

RESULTS

Integrative clustering identified three distinct LNM subtypes: (i) an immune subtype (Group 1), (ii) an invasive subtype (Group 2), and (iii) a metabolic/proliferative subtype (Group 3). Group 2 subtype was associated with significantly worse locoregional control and survival. LNM-specific subtypes were not observed in matched primary tumor specimens. In HNSCCs, breast cancers, and melanomas, a 73-gene classifier identified similar Group 2 LNM subtypes that were associated with worse disease control and survival only when applied to lymph node sites, but not when applied to other primary tumors or metastatic sites. Similarly, previously proposed prognostic classifiers better distinguished patients with worse outcomes when applied to the transcriptional profiles of LNMs, but not the profiles of primary tumors.

CONCLUSIONS

The transcriptional profiles of LNMs better predict outcomes than transcriptional profiles of primary tumors. The LNMs display site-specific subtypes associated with worse disease control and survival across multiple cancer types.

Biomarkers: paving stones on the road towards the personalized precision medicine for oral squamous cell carcinoma

Traditional therapeutics have encountered a bottleneck caused by diagnosis delay and subjective and unreliable assessment. Biomarkers can overcome this bottleneck and guide us toward personalized precision medicine for oral squamous cell carcinoma. To achieve this, it is important to efficiently and accurately screen out specific biomarkers from among the huge number of molecules. Progress in omics-based high-throughput technology has laid a solid foundation for biomarker discovery. With credible and systemic biomarker models, more precise and personalized diagnosis and assessment would be achieved and patients would be more likely to be cured and have a higher quality of life. However, this is not straightforward owing to the complexity of molecules involved in tumorigenesis. In this context, there is a need to focus on tumor heterogeneity and homogeneity, which are discussed in detail. In this review, we aim to provide an understanding of biomarker discovery and application for precision medicine of oral squamous cell carcinoma, and have a strong belief that biomarker will pave the road toward future precision medicine.

A model to predict nodal metastasis in patients with oral squamous cell carcinoma

Difficulty in precise decision making on necessity of surgery is a major problem when managing oral squamous cell carcinomas (OSCCs) with clinically negative neck. Therefore, use of clinical and histopathological parameters in combination would be important to improve patient management. The main objective is to develop a model that predicts the presence of nodal metastasis in patients with OSCC.623 patients faced neck dissections with buccal mucosal or tongue squamous cell carcinoma (SCC) were selected from patients’ records. Demographic data, clinical information, nodal status, Depth of invasion (DOI) and pattern of invasion (POI) were recorded. The parameters which showed a significant association with nodal metastasis were used to develop a multivariable predictive model (PM). Univariate logistic regression was used to estimate the strengths of those associations in terms of odds ratios (OR). This showed statistically significant associations between status of the nodal metastasis and each of the following 4 histopathological parameters individually: size of the tumour (T), site, POI, and DOI. Specifically, OR of nodal metastasis for tongue cancers relative to buccal mucosal cancers was 1.89, P-value < 0.001. Similarly, ORs for POI type 3 and 4 relative to type 2 were 1.99 and 5.83 respectively. A similar relationship was found with tumour size; ORs for T2, T3, and T4 compared to T1 were 2.79, 8.27 and 8.75 respectively. These four histopathological parameters were then used to develop a predictive model for nodal metastasis. This model showed that probability of nodal metastasis is higher among tongue cancers with increasing POI, with increasing T, and with larger depths while other characteristics remained unchanged. The proposed model provides a way of using combinations of histopathological parameters to identify patients with higher risks of nodal metastasis for surgical management.

Prediction of Occult Lymph Node Metastasis in Head and Neck Cancer with CD31 Vessel Quantification

Objective

The management of occult lymph node metastasis (LNM) in head and neck squamous cell carcinoma has been a matter of controversy for decades. The vascular density within the tumor microenvironment, as an indicator of ongoing angiogenesis, could constitute an attractive predictor of LNM. The use of the panvascular endothelial antibody CD31 as a marker of occult LNM has never been reported. The aim of this study was to assess the predictive value of CD31 microvascular density for the detection of occult LNM in squamous cell carcinoma of the oral cavity and oropharynx.

Study Design

Case series with chart review.

Setting

Tertiary university hospital.

Subjects and Methods

Intra- and peritumoral microvascular density values were determined in 56 cases of squamous cell carcinoma of the oral cavity (n = 50) and oropharynx (n = 6) with clinically negative necks using the CD31 marker. Statistical associations of CD31 microvascular densities with clinicopathologic data were then established.

Results

Peritumoral CD31 microvascular density was significantly associated with occult LNM in multivariate analysis ( P < .01). Recursive partitioning analysis for this parameter found a cutoff of 19.33, which identified occult LNM with a sensitivity of 91%, a specificity of 65%, a positive predictive value of 40%, a negative predictive value of 97%, and an overall diagnostic accuracy of 71%.

Conclusion

Peritumoral CD31 microvascular density in primary squamous cell carcinoma of the oral cavity and oropharynx allows accurate prediction of occult LNM.

Improved high-dimensional prediction with Random Forests by the use of co-data

Background

Prediction in high dimensional settings is difficult due to the large number of variables relative to the sample size. We demonstrate how auxiliary ‘co-data’ can be used to improve the performance of a Random Forest in such a setting.

Results

Co-data are incorporated in the Random Forest by replacing the uniform sampling probabilities that are used to draw candidate variables by co-data moderated sampling probabilities. Co-data here are defined as any type information that is available on the variables of the primary data, but does not use its response labels. These moderated sampling probabilities are, inspired by empirical Bayes, learned from the data at hand. We demonstrate the co-data moderated Random Forest (CoRF) with two examples. In the first example we aim to predict the presence of a lymph node metastasis with gene expression data. We demonstrate how a set of external p-values, a gene signature, and the correlation between gene expression and DNA copy number can improve the predictive performance. In the second example we demonstrate how the prediction of cervical (pre-)cancer with methylation data can be improved by including the location of the probe relative to the known CpG islands, the number of CpG sites targeted by a probe, and a set of p-values from a related study.

Conclusion

The proposed method is able to utilize auxiliary co-data to improve the performance of a Random Forest.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-017-1993-1) contains supplementary material, which is available to authorized users.