Application of Machine Learning in Developing a Novelty Five-Pseudogene Signature to Predict Prognosis of Head and Neck Squamous Cell Carcinoma: A New Aspect of "Junk Genes" in Biomedical Practice

Prognostic value of CDKN2A in head and neck squamous cell carcinoma via pathomics and machine learning

This study aims to enhance the prognosis prediction of Head and Neck Squamous Cell Carcinoma (HNSCC) by employing artificial intelligence (AI) to analyse CDKN2A gene expression from pathology images, directly correlating with patient outcomes. Our approach introduces a novel AI-driven pathomics framework, delineating a more precise relationship between CDKN2A expression and survival rates compared to previous studies. Utilizing 475 HNSCC cases from the TCGA database, we stratified patients into high-risk and low-risk groups based on CDKN2A expression thresholds. Through pathomics analysis of 271 cases with available slides, we extracted 465 distinctive features to construct a Gradient Boosting Machine (GBM) model. This model was then employed to compute Pathomics scores (PS), predicting CDKN2A expression levels with validation for accuracy and pathway association analysis. Our study demonstrates a significant correlation between higher CDKN2A expression and improved median overall survival (66.73 months for high expression vs. 42.97 months for low expression, p = 0.013), establishing CDKN2A's prognostic value. The pathomic model exhibited exceptional predictive accuracy (training AUC: 0.806; validation AUC: 0.710) and identified a strong link between higher Pathomics scores and cell cycle activation pathways. Validation through tissue microarray corroborated the predictive capacity of our model. Confirming CDKN2A as a crucial prognostic marker in HNSCC, this study advances the existing literature by implementing an AI-driven pathomics analysis for gene expression evaluation. This innovative methodology offers a cost-efficient and non-invasive alternative to traditional diagnostic procedures, potentially revolutionizing personalized medicine in oncology.

A hypoxia-glycolysis-lactate-related gene signature for prognosis prediction in hepatocellular carcinoma

BACKGROUND

Liver cancer ranks sixth in incidence and third in mortality globally and hepatocellular carcinoma (HCC) accounts for 90% of it. Hypoxia, glycolysis, and lactate metabolism have been found to regulate the progression of HCC separately. However, there is a lack of studies linking the above three to predict the prognosis of HCC. The present study aimed to identify a hypoxia-glycolysis-lactate-related gene signature for assessing the prognosis of HCC.

METHODS

This study collected 510 hypoxia-glycolysis-lactate genes from Molecular Signatures Database (MSigDB) and then classified HCC patients from TCGA-LIHC by analyzing their hypoxia-glycolysis-lactate genes expression. Differentially expressed genes (DEGs) were screened out to construct a gene signature by LASSO-Cox analysis. Univariate and multivariate regression analyses were used to evaluate the independent prognostic value of the gene signature. Analyses of immune infiltration, somatic cell mutations, and correlation heatmap were conducted by "GSVA" R package. Single-cell analysis conducted by "SingleR", "celldex", "Seurat", and "CellCha" R packages revealed how signature genes participated in hypoxia/glycolysis/lactate metabolism and PPI network identified hub genes.

RESULTS

We classified HCC patients from TCGA-LIHC into two clusters and screened out DEGs. An 18-genes prognostic signature including CDCA8, CBX2, PDE6A, MED8, DYNC1LI1, PSMD1, EIF5B, GNL2, SEPHS1, CCNJL, SOCS2, LDHA, G6PD, YBX1, RTN3, ADAMTS5, CLEC3B, and UCK2 was built to stratify the risk of HCC. The risk score of the hypoxia-glycolysis-lactate gene signature was further identified as a valuable independent factor for estimating the prognosis of HCC. Then we found that the features of clinical characteristics, immune infiltration, somatic cell mutations, and correlation analysis differed between the high-risk and low-risk groups. Furthermore, single-cell analysis indicated that the signature genes could interact with the ligand-receptors of hepatocytes/fibroblasts/plasma cells to participate in hypoxia/glycolysis/lactate metabolism and PPI network identified potential hub genes in this process: CDCA8, LDHA, YBX1.

CONCLUSION

The hypoxia-glycolysis-lactate-related gene signature we built could provide prognostic value for HCC and suggest several hub genes for future HCC studies.

Disulfidptosis features and prognosis in head and neck squamous cell carcinoma patients: unveiling and validating the prognostic signature across cohorts

BACKGROUND

Head and neck squamous cell carcinoma (HNSCC) is a significant health concern with a variable global incidence and is linked to regional lifestyle factors and HPV infections. Despite treatment advances, patient prognosis remains variable, necessitating an understanding of its molecular mechanisms and the identification of reliable prognostic biomarkers.

METHODS

We analyzed 959 HNSCC samples and employed batch correction to obtain consistent transcriptomic data across cohorts. We examined 79 disulfidptosis-related genes to determine consensus clusters and utilized high-throughput sequencing to identify genetic heterogeneity within tumors. We established a disulfidptosis prognostic signature (DSPS) using least absolute shrinkage and selection operator (LASSO) regression and developed a prognostic nomogram integrating the DSPS with clinical factors. Personalized chemotherapy prediction was performed using the "pRRophetic" R package.

RESULTS

Batch corrections were used to harmonize gene expression data, revealing two distinct disulfidptosis subtypes, C1 and C2, with differential gene expression and survival outcomes. Subtype C1, characterized by increased expression of the MYH family genes ACTB, ACTN2, and FLNC, had a mortality rate of 48.4%, while subtype C2 had a mortality rate of 38.7% (HR = 0.77, 95% CI: 0.633-0.934, P = 0.008). LASSO regression identified 15 genes that composed the DSPS prognostic model, which independently predicted survival (HR = 2.055, 95% CI: 1.420-2.975, P < 0.001). The prognostic nomogram, which included the DSPS, age, and tumor stage, predicted survival with AUC values of 0.686, 0.704, and 0.789 at 3, 5, and 8 years, respectively, indicating strong predictive capability. In the external validation cohort (cohort B), the DSPS successfully identified patients at greater risk, with worse overall survival outcomes in the high-DSPS subgroup (HR = 1.54, 95% CI: 1.17-2.023, P = 0.002) and AUC values of 0.601, 0.644, 0.636, and 0.748 at 3, 5, 8, and 10 years, respectively, confirming the model's robustness.

CONCLUSION

The DSPS provides a robust prognostic tool for HNSCC, underscoring the complexity of this disease and the potential for tailored treatment strategies. This study highlights the importance of molecular signatures in oncology, offering a step toward personalized medicine and improved patient outcomes in HNSCC management.

Identification of Pseudo-R genes in Vitis vinifera and characterization of their role as immunomodulators in host-pathogen interactions

INTRODUCTION

Duplication events are fundamental to co-evolution in host-pathogen interactions. Pseudogenes (Ψs) are dysfunctional paralogs of functional genes and resistance genes (Rs) in plants are the key to disarming pathogenic invasions. Thus, deciphering the roles of pseudo-R genes in plant defense is momentous.

OBJECTIVES

This study aimed to functionally characterize diverse roles of the resistance Ψs as novel gene footprints and as significant gene regulators in the grapevine genome.

METHODS

PlantPseudo pipeline and HMM-profiling identified whole-genome duplication-derived (WGD) Ψs associated with resistance genes (Ψ-Rs). Further, novel antifungal and antimicrobial peptides were characterized for fungal associations using protein-protein docking with Erysiphe necator proteins. miRNA and tasiRNA target sites and transcription factor (TF) binding sites were predicted in Ψ-Rs. Finally, differential co-expression patterns in Ψ-Rs-lncRNAs-coding genes were identified using the UPGMA method.

RESULTS

2,746 Ψ-Rs were identified from 31,032 WGD Ψs in the genome of grapevine. 69-antimicrobial and 81-antifungal novel peptides were generated from Ψ-Rs. The putative genic potential was predicted for five novel antifungal peptides which were further characterized by docking against E. necator proteins. 395 out of 527 resistance loci-specific Ψ-Rs were acting as parental gene mimics. Further, to explore the diverse roles of Ψ-Rs in plant-defense, we identified 37,026 TF-binding sites, 208 miRNA, and 99 tasiRNA targeting sites on these Ψ-Rs. 194 Ψ-Rs were exhibiting tissue-specific expression patterns. The co-expression network analysis between Ψs-lncRNA-genes revealed six out of 79 pathogen-responsive Ψ-Rs as significant during pathogen invasion.

CONCLUSIONS

Our study provides pathogen responsive Ψ-Rs integral for pathogen invasion, which will offer a useful resource for future experimental validations. In addition, our findings on novel peptide generations from Ψ-Rs offer valuable insights which can serve as a useful resource for predicting novel genes with the futuristic potential of being investigated for their bioactivities in the plant system.

Emerging roles of PIWI-interacting RNAs (piRNAs) and PIWI proteins in head and neck cancer and their potential clinical implications

Head and neck squamous cell carcinoma (HNSCC) are among the well-known neoplasms originating in the oral cavity, pharynx, and larynx. Despite advancements in chemotherapy, radiotherapy, and surgery, the survival rates of the patients are low, which has posed a major therapeutic challenge. A growing number of non-coding RNAs (ncRNAs), for instance, microRNAs, have been identified whose abnormal expression patterns have been implicated in HNSCC. However, more recently, several seminal research has shown that piwi-interacting RNAs (piRNAs), a promising and young class of small ncRNA, are linked to the emergence and progression of cancer. They can regulate transposable elements (TE) and gene expression through multiple mechanisms, making them potentially more powerful regulators than miRNAs. Hence, they can be more promising ncRNAs candidates for cancer therapeutic intervention. Here, we surveyed the roles and clinical implications of piRNAs and their PIWI proteins partners in tumorigenesis and associated molecular processes of cancer, with a particular focus on HNSCC, to offer a new avenue for diagnosis, prognosis, and therapeutic interventions for the malignancy, improving patient's outcomes.

The World of Pseudogenes: New Diagnostic and Therapeutic Targets in Cancers or Still Mystery Molecules?

Pseudogenes were once considered as “junk DNA”, due to loss of their functions as a result of the accumulation of mutations, such as frameshift and presence of premature stop-codons and relocation of genes to inactive heterochromatin regions of the genome. Pseudogenes are divided into two large groups, processed and unprocessed, according to their primary structure and origin. Only 10% of all pseudogenes are transcribed into RNAs and participate in the regulation of parental gene expression at both transcriptional and translational levels through senseRNA (sRNA) and antisense RNA (asRNA). In this review, about 150 pseudogenes in the different types of cancers were analyzed. Part of these pseudogenes seem to be useful in molecular diagnostics and can be detected in various types of biological material including tissue as well as biological fluids (liquid biopsy) using different detection methods. The number of pseudogenes, as well as their function in the human genome, is still unknown. However, thanks to the development of various technologies and bioinformatic tools, it was revealed so far that pseudogenes are involved in the development and progression of certain diseases, especially in cancer.

Comparison of time-to-event machine learning models in predicting oral cavity cancer prognosis

BACKGROUND

Applying machine learning to predicting oral cavity cancer prognosis is important in selecting candidates for aggressive treatment following diagnosis. However, models proposed so far have only considered cancer survival as discrete rather than dynamic outcomes.

OBJECTIVES

To compare the model performance of different machine learning-based algorithms that incorporate time-to-event data. These algorithms included DeepSurv, DeepHit, neural net-extended time-dependent cox model (Cox-Time), and random survival forest (RSF).

MATERIALS AND METHODS

Retrospective cohort of 313 oral cavity cancer patients were obtained from electronic health records. Models were trained on patient data following preprocessing. Predictors were based on demographic, clinicopathologic, and treatment information of the cases. Outcomes were the disease-specific and overall survival. Multivariable analyses were conducted to select significant prognostic features associated with tumor prognosis. Two models were generated per algorithm based on all-prognostic features and significant-prognostic features following statistical analysis. Concordance index (c-index) and integrated Brier scores were used as performance evaluators and model stability was assessed using intraclass correlation coefficients (ICC) calculated from these measures obtained from the cross-validation folds.

RESULTS

While all models were satisfactory, better discriminatory performance and calibration was observed for disease-specific than overall survival (mean c-index: 0.85 vs 0.74; mean integrated Brier score: 0.12 vs 0.17). DeepSurv performed best in terms of discrimination for both outcomes (c-indices: 0.76 -0.89) while RSF produced better calibrated survival estimates (integrated Brier score: 0.06 -0.09). Model stability of the algorithms varied with the outcomes as Cox-Time had the best intraclass correlation coefficient (mean ICC: 1.00) for disease-specific survival while DeepSurv was most stable for overall survival prediction (mean ICC: 0.99).

CONCLUSIONS

Machine learning algorithms based on time-to-event outcomes are successful in predicting oral cavity cancer prognosis with DeepSurv and RSF producing the best discriminative performance and calibration.

Identification of Immune Subtypes for Predicting the Prognosis of Patients in Head and Neck Squamous Cell Carcinoma

Head and neck squamous cell carcinoma (HNSCC) is a common malignancy with poor prognosis and immune response, which plays an important role in tumor progression. Recently, immunotherapies have revolutionized the therapeutic means of malignancies including HNSCC. However, the relationship between immunophenotypes of HNSCC and its clinical response to immune-checkpoint inhibitors remains unclear. We aim to identify molecular subtyping related to distinct immunophenotypes in HNSCC. Consensus clustering algorithm was conducted for subtyping. Immunophenotypes between subtypes were compared according to infiltrating immunocytes, immune reactions, major histocompatibility complex (MHC) family, immunoinhibitory, immunostimulatory and immune scores. The relationship between immunophenotype and genotype was investigated from gene mutation and tumor mutation burden. The potential response of Immune-checkpoint blockade (ICB) therapy was estimated with TIDE and ImmuCellAI algorithms, and immune-checkpoint genes. The immune characteristics were also investigated. Biological functions were annotated by the gene-set enrichment analysis (GSEA) algorithm. Two distinct immune subtypes of HNSCC with different survival outcomes, biological characteristics, immunophenotype, and ICB response were identified. The subtype-1 was featured with better prognosis, more infiltrated immunocytes, stronger immune reaction, higher immune-related gene expression, higher immune-checkpoint gene expression (PD-1, PD-L1, and CTLA-4), and better ICB response. A higher immune response in subtype-1 was also revealed by GSEA. Subtype-1 possessed a higher immune response and more sensitivity to ICB therapy leading to a better prognosis. These findings may shed promising light on the immunotherapy strategy in HNSCC

Pseudogene Transcripts in Head and Neck Cancer: Literature Review and In Silico Analysis

Once considered nonfunctional, pseudogene transcripts are now known to provide valuable information for cancer susceptibility, including head and neck cancer (HNC), a serious health problem worldwide, with about 50% unimproved overall survival over the last decades. The present review focuses on the role of pseudogene transcripts involved in HNC risk and prognosis. We combined current literature and in silico analyses from The Cancer Genome Atlas (TCGA) database to identify the most deregulated pseudogene transcripts in HNC and their genetic variations. We then built a co-expression network and performed gene ontology enrichment analysis to better understand the pseudogenes’ interactions and pathways in HNC. In the literature, few pseudogenes have been studied in HNC. Our in silico analysis identified 370 pseudogene transcripts associated with HNC, where SPATA31D5P, HERC2P3, SPATA31C2, MAGEB6P1, SLC25A51P1, BAGE2, DNM1P47, SPATA31C1, ZNF733P and OR2W5 were found to be the most deregulated and presented several genetic alterations. NBPF25P, HSP90AB2P, ZNF658B and DPY19L2P3 pseudogenes were predicted to interact with 12 genes known to participate in HNC, DNM1P47 was predicted to interact with the TP53 gene, and HLA-H pseudogene was predicted to interact with HLA-A and HLA-B genes. The identified pseudogenes were associated with cancer biology pathways involving cell communication, response to stress, cell death, regulation of the immune system, regulation of gene expression, and Wnt signaling. Finally, we assessed the prognostic values of the pseudogenes with the Kaplan–Meier Plotter database, and found that expression of SPATA31D5P, SPATA31C2, BAGE2, SPATA31C1, ZNF733P and OR2W5 pseudogenes were associated with patients’ survival. Due to pseudogene transcripts’ potential for cancer diagnosis, progression, and as therapeutic targets, our study can guide new research to HNC understanding and development of new target therapies.

Assessing the risk of dengue severity using demographic information and laboratory test results with machine learning

BACKGROUND

Dengue virus causes a wide spectrum of disease, which ranges from subclinical disease to severe dengue shock syndrome. However, estimating the risk of severe outcomes using clinical presentation or laboratory test results for rapid patient triage remains a challenge. Here, we aimed to develop prognostic models for severe dengue using machine learning, according to demographic information and clinical laboratory data of patients with dengue.

METHODOLOGY/PRINCIPAL FINDINGS

Out of 1,581 patients in the National Cheng Kung University Hospital with suspected dengue infections and subjected to NS1 antigen, IgM and IgG, and qRT-PCR tests, 798 patients including 138 severe cases were enrolled in the study. The primary target outcome was severe dengue. Machine learning models were trained and tested using the patient dataset that included demographic information and qualitative laboratory test results collected on day 1 when they sought medical advice. To develop prognostic models, we applied various machine learning methods, including logistic regression, random forest, gradient boosting machine, support vector classifier, and artificial neural network, and compared the performance of the methods. The artificial neural network showed the highest average discrimination area under the receiver operating characteristic curve (0.8324 ± 0.0268) and balance accuracy (0.7523 ± 0.0273). According to the model explainer that analyzed the contributions/co-contributions of the different factors, patient age and dengue NS1 antigenemia were the two most important risk factors associated with severe dengue. Additionally, co-existence of anti-dengue IgM and IgG in patients with dengue increased the probability of severe dengue.

CONCLUSIONS/SIGNIFICANCE

We developed prognostic models for the prediction of dengue severity in patients, using machine learning. The discriminative ability of the artificial neural network exhibited good performance for severe dengue prognosis. This model could help clinicians obtain a rapid prognosis during dengue outbreaks. However, the model requires further validation using external cohorts in future studies.

TP53-Associated Ion Channel Genes Serve as Prognostic Predictor and Therapeutic Targets in Head and Neck Squamous Cell Carcinoma

TP53 mutations are the most occurred mutation in HNSCC which might affect the ion channel genes. We aim to investigate the ion channel gene alteration under TP53 mutation and their prognostic implication. The overall mutation status of HNSCC were explored. By screening the TP53-associated ion channel genes (TICGs), an ion channel prognostic signature (ICPS) was established through a series of machine learning algorithms. The ICPS was then evaluated and its clinical significance was explored. 82 TICGs differentially expressed between TP53WT and TP53MUT were screened. Using univariate regression analysis and LASSO regression analysis and multivariate regression analysis, an ICPS containing 7 ion channel genes was established. A series of evaluation was carried out which proved the predictive ability of ICPS. Functional analysis of ICPS revealed that cancer-related pathways were enriched in high-risk group. Next, for clinical application, a nomogram was constructed based on ICPS and other independent clinicopathological factors. TP53 mutation status strongly affects the expression of ion channel genes. The ICPM we have identified is a strong indicator for HNSCC prognosis and could help with patient stratification as well as identification of novel drug targets.

Machine learning models to predict length of stay and discharge destination in complex head and neck surgery

BACKGROUND

This study develops machine learning (ML) algorithms that use preoperative-only features to predict discharge-to-nonhome-facility (DNHF) and length-of-stay (LOS) following complex head and neck surgeries.

METHODS

Patients undergoing laryngectomy or composite tissue excision followed by free tissue transfer were extracted from the 2005 to 2017 NSQIP database.

RESULTS

Among the 2786 included patients, DNHF and mean LOS were 421 (15.1%) and 11.7 ± 8.8 days. Four classification models for predicting DNHF with high specificities (range, 0.80-0.84) were developed. The generalized linear and gradient boosting machine models performed best with receiver operating characteristic (ROC), accuracy, and negative predictive value (NPV) of 0.72-0.73, 0.75-0.76, and 0.88-0.89. Four regression models for predicting LOS in days were developed, where all performed similarly with mean absolute error and root mean-squared errors of 3.95-3.98 and 5.14-5.16. Both models were developed into an encrypted web-based interface: https://uci-ent.shinyapps.io/head-neck/.

CONCLUSION

Novel and proof-of-concept ML models to predict DNHF and LOS were developed and published as web-based interfaces.