Multidimensional transcriptomic and immune profiling of papillary thyroid cancer: bridging molecular networks, disease stage, survival outcomes, and drug discovery

Abstract

Thyroid cancer ranks tenth among the deadliest cancers globally. Papillary thyroid cancer (PTC) represents the majority (> 80%) of all thyroid cancer types. Albeit radioactive iodine therapy achieves a great deal of success rate and efficacy, it is not easily available in all affected regions and its inappropriate use can drive the development of new malignancies. This motivated us to rigorously analyze the transcriptomes of PTC so as to infer novel prognostic markers and hotspot targets that are druggable using the approved drugs. We downloaded two datasets from gene expression omnibus and subsequently processed them via GEO2R tool. The mutual differentially expressed genes (DEGs) from the two datasets included 452 genes. This was followed by the prediction of the interacting network through NetworkAnalyst followed by cytoscape to shortlist them into four hub genes (FN1, MAPK13, KRT19 and EGFR) by applying four cytohubba filters (MCC, MNC, closeness and betweenness). The hub genes were then evaluated for their interaction with their neighbor genes, differential expression in tumor versus healthy control samples, stage correlation, immune infiltration, overall survival and availability of specific drugs. The four hub genes displayed higher level of expression in tumor samples, varied correlation with PTC stage, high immune cells infiltrate rate. Moreover, MAPK13 and KRT19 were strongly associated with overall survival of PTC, while FN1 and KRT19 showed significantly low tumor purity. In conclusion, FN1, MAPK13, KRT19 and EGFR can be considered both good prognostic markers and druggable targets for PTC.

Introduction

Thyroid cancer, a prevalent malignancy within the endocrine system, has been on the rise globally, with papillary thyroid cancer (PTC) emerging as the most common subtype, constituting over 80% of thyroid malignancies [1]. PTC is the most common endocrine malignancy, accounting for approximately 80–85% of all thyroid cancers, with a rising incidence globally in recent decades [2]. PTC is characterized by distinctive histological features, including papillae formation, nuclear grooves, and pseudoinclusions, and is typically associated with an excellent prognosis, boasting a 10-year survival rate exceeding 90% [3]. However, a subset of PTC cases exhibits aggressive behavior, including lymph node metastasis, vascular invasion, or resistance to radioactive iodine therapy, necessitating risk-stratified management approaches. Molecular alterations, particularly BRAF V600E mutations and RET/PTC rearrangements, play a critical role in tumorigenesis and have been leveraged for diagnostic and therapeutic advancements [4]. Recent advances in molecular profiling and targeted therapies, such as tyrosine kinase inhibitors, have improved outcomes for advanced or refractory PTC [5]. Ongoing research continues to refine risk assessment, surgical approaches, and systemic treatments to optimize patient care.

Current guidelines suggest a less aggressive surgical approach and limited use of postoperative radioiodine therapy for PTC cases, particularly in low-risk intrathyroidal PTMCs without lymph node or distant metastases, where active surveillance may be a viable alternative with a low progression rate of 1–5% and no increased risk of poorer outcomes related to delayed surgery [6, 7]. Treatment for PTC often combines surgery, radioiodine therapy, and hormone replacement therapy with levothyroxine drugs, aiming to provide optimal care in line with international standards [8, 9]. Nevertheless, such combined therapies failed to radically get rid of the cancer, necessitating the deep investigation into single cell transcriptomics to find unique biomarkers and targets that could potentially provide a cue for addressing PTC.

Recent advancements in transcriptomic analysis of PTC have revealed critical insights into its molecular landscape, novel therapeutic targets, and biomarkers for risk stratification. Single-cell RNA sequencing has characterized the tumor microenvironment, identifying premalignant thyrocytes and three malignant phenotypes that influence tumor behavior and response to radioactive iodine therapy, along with vascular-immune crosstalk supporting combined anti-angiogenic and immunotherapy approaches [10]. Comprehensive genomic studies have highlighted actionable mutations, such as BRAF V600 and RET fusions, enabling targeted therapies like selective BRAF inhibitors; however, resistance mechanisms remain a challenge [11]. Transcriptomic profiling has further differentiated high-risk from low-risk PTC by identifying immune-related gene upregulation in low-risk cases and oncogenic signaling activation in high-risk cases, leading to the development of a 31-gene signature for personalized treatment strategies [12]. Additionally, immune-escape markers (e.g., CTLA4, PDCD1) and novel biomarkers like HOXD9 have been associated with recurrence risk, suggesting potential for precision medicine approaches [13]. Multiomics studies continue to refine diagnostic and prognostic tools, emphasizing the importance of integrating molecular insights into clinical practice [14].

Extensive research efforts have been dedicated to elucidating the molecular mechanisms underlying disease progression, yet the translation of these findings into effective therapeutic strategies has proven challenging. Transcriptomics analysis offers a powerful approach to unravel the complex gene regulatory networks and key molecular alterations driving PTC pathogenesis. By leveraging publicly available transcriptomic data, this study aims to identify hub genes and regulatory factors critically involved in PTC progression. These findings can reveal novel pharmacologic targets, paving the way for the development of innovative chemotherapeutic interventions to address the unmet clinical needs of PTC patients.

Methods

Study design

The overall workflow followed in the present study is illustrated in Fig. 1. Nonetheless, the details about each stage is explained in the text below.

Fig. 1.

The designated flowchart summarizing the main steps to accomplish the goals of this in silico study

Data extraction and processing

Gene expression omnibus (GEO) repository [15] (https://www.ncbi.nlm.nih.gov/geo/) was used a source for retrieving the raw data. GSE205733 dataset (no published report) involved 8 specimens, 4 cancerous thyroid tissues and other 4 paracancer tissues. Also, GSE237841 dataset [16] contained 6 specimens in total sepgreated into 2 control group while 4 as PTC group. The two datasets employed expression profiling by high throughput sequencing.

The transcriptome of two types of data was processed with GEO2R embedded tool [17] (https://www.ncbi.nlm.nih.gov/geo/geo2r/) which can be used for the comparison of two or more transcriptomic samples to identify differentially expressed genes (DEGs) with the assistance of GEO query and limma R packages. The shortlisting is based upon log2FC<-1 for sorting down-regulated genes and log2FC > 1 for up-regulated genes with a p value < 0.05. The differentially-expressed genes (DEGs) of the two data sets are then submitted to Venn online tool (https://bioinformatics.psb.ugent.be/webtools/Venn/) for drawing and concluding the mutual genes.

Analysis of gene enrichment

To give an overall overview of the mutual genes roles, FunRich standalone software [18] was employed. This involved exploration of the implicated biological processes, cellular compartments, molecular function as well as biological pathways. An adjusted p value less than 0.05 was a cut of criteria for statistical significance.

Hub gene network analysis

The shared genes were then submitted to NetworkAnalyst [19] (https://www.networkanalyst.ca/) to infer the protein-protein interactions (PPIs). The tissue specific protein-protein interaction data were collected from the DifferentialNet database where the tissue selected was “Thyroid” at a filter degree of 15.0. Afterward, minimum network option was applied to narrow the generated network since the mutual DEGs were too large. The generated network was then sent to Cytoscape v. 3.9.1 [20] and used CytoHubba plugin to find the 20 most important genes (hub genes) mutual among (i) MCC (maximal clique centrality), (ii) MNC (maximum neighborhood component), (iii) closeness, and (iv) betweenness.

Genemania

In addition, the most interacting network of the predicted hub genes was assessed via Genemania (https://genemania.org/) [21]. The server provides the 20 most interacting network based upon physical connections, co-localization, co-expression, besides the predicted interactions.

Identification of transcriptional and post-transcriptional regulators

The determined hub genes were then investigated for the identification of their transcriptional regulation (via transcriptional factors, TF) and post-transcriptionally (via miRNA). This was accomplished by utilizing NetworkAnalyst [19] (https://www.networkanalyst.ca/). TF-miRNA co-network was generated based on RegNetwork repository. The generated co-network was imported into Cytoscape to get the most significant TF and miRNA according to degree filtration of Cytohubba plugin, setting degree at 15.

Differential expression, stage and immune infiltration

The differential expression of the shortlisted hub genes between cancer and control groups was predicted through utilizing GEPIA 2 web portal (http://gepia2.cancer-pku.cn/#index) [22]. The Log2FC cutoff and p-value were set at 1 and < 0.01, respectively. Besides, the stage of cancer progression was also checked for the hub genes using the same portal. Concerning, the immune infiltration, we inputted the hub genes into TIMER server version 1 (https://cistrome.shinyapps.io/timer/) [23]. The server analyzes tumor purity, CD4+T cell, CD8+T cell, B cell, macrophage, neutrophil, and dendritic cell of the inputted gene.

Survival prediction

KM plotter [24] (https://kmplot.com/analysis/) was deployed for the purpose of scrutinization of the association between hub genes and survival outcomes, primarily overall survival of PTC. The impacts of high and low expression levels of hub genes on survival rates were compared. Hazard ratios with 95% confidence intervals were calculated, and a log-rank P value of less than 0.05 was considered statistically significant without any other restrictions (n = 502).

Hub genes–drug interaction

To decipher the potential drugs able to inhibit, agonize, antagonize or block the hub genes as a treatment availability, we used the most recently updated database DGIdb 5.0 (https://dgidb.org/) [25]. The server extracts drug-gene interactions from databases such as DrugBank, PharmGKB, ChEMBL, and Drug Target Commons. Additionally, genes have been classified as potentially druggable based on their presence in specific pathways, molecular functions, and gene families, according to resources like the Gene Ontology, the Human Protein Atlas, IDG, and “druggable genome”. DGIdb encompasses over 10,000 genes and 20,000 drugs, accounting for nearly 70,000 drug-gene interactions, or fitting into one of 43 potentially druggable gene categories.

Results

Data extraction and processing

The two gene expression dataset were processed to infer the differentially expressed genes (DEGs). Mean-difference plots resulted from GEO2R tool are provided in Fig. 2. Up-regulated and down-regulated genes that displayed significant difference were colored in red and blue, respectively.

Fig. 2.

Mean-difference plots of the two datasets analyzed in this study. Up-regulated and down-regulated genes are colored in red and blue, respectively. Data are provided for GSE237841 (A) and GSE205733 (B) datasets

From GSE205733, a total of 2191 DEGs were obtained, from which 829 were up-regulated and 1371 down-regulated. Likewise, 703 DEGs were extracted from GSE237841 (165 up-regulated and 538 down-regulated). In both datasets the majority of DEGs showed low expression. The shared DEGs from the two datasets were 452 (Fig. 3) which were selected further for comprehensive analysis.

Fig. 3.

Venn diagram for identification of mutual DEGs between the two datasets processed

Enrichment of DEGs

Gene enrichment analysis of the mutual DEGs ranked plasma membrane as the main cellular compartment (38.9%), followed by extracellular milieu (24.4%) and extracellular space (7.7%) as shown in Fig. 4. This gives us a clue that the driving signal of PTC might come from outside the cell and interact with plasma membrane receptors or involved activation of a signaling cascade and the subsequent mediators.

Fig. 4.

Output of the DEGs enrichments by using FunRich program. A Cellular compartment, B biological process, C biological pathways, and D molecular function

This interpretation is enhanced by the biological process enrichments where signal transduction (28.9%) and cell communication (27.8%) ranked as top 2. In addition, cell growth ranked third with a percentage of 14.1%. This enforces the reasoning that a signal from outside the cell interacted with membrane receptors of thyroid gland and triggered cellular growth therein.

Although not statistically significant, integrin cell surface interactions exhibited maximal enrichment in biological pathway of FunRich (34.6%). The transition between mesenchymal-to-epithelial ranked second with enrichment of 15%. This emphasizes our hypothesis further where the signal transduction thyroid cells underwent, led to such transformation.

The molecular functions of the DEGs were found to converge at extracellular matrix structural constituent and cell adhesion activity (each constituting 6.2%). Next, 4.6% of the DEGs exhibited receptor activity. These glorify the aberrant functions that thyroid cells subject upon transformation into PTC.

Hub gene network

As the shared DEGs were very high, minimum network option was applied. This outputted 627 nodes, 1282 edges, and 248 seeds as illustrated in Fig. 5. The figure visualized the DEGs according to their number of connections where the large circles indicate higher degree of connections. This network was then sent to cytoscape program to compute the hub genes.

Fig. 5.

Minimum network generated for the DEGs using NetworkAnalyst server. Gene circles are designated according to their connections

To increase the reliability of the prediction, four filtering approaches were selected and converged, namely, MCC, MNC, closeness and betweenness. Among the shortlisted 20 genes from each category, only 4 (FN1, MAPK13, KRT19 and EGFR) were consensus (Fig. 6). The four hub genes were the most implicated in the pathogenesis of PTC. So, these hub genes were then completed the comprehensive investigation of the present study.

Fig. 6.

The shortlisted hub genes resulted by applying four cytohubba filters (MCC, MNC, Betweenness and Closeness)

Genemania

Afterwards, the hub genes were submitted to genemania to infer the most significant interactions and the corresponding neighbors in the normal function and signaling pathway or compartmentalization. This resulted in the identification of 20 most interacting genes that are within the same family or complement each other in terms of signaling cascade in the function in question. The network involved significant genes such as MAPKAPK3, PIK3CA, RALB, ADAM8, EGF, HGF, among others (Fig. 7).

Fig. 7.

The most interacting genes related to the four hub genes (shown in the middle) as predicted by GeneMania

It is worth noting that > 77% of connections are physical (pink), 8% co-expression (violet), 3% co-localization (blue) whilst only 5% prediction-based (orange). Moreover, these neighbor-genes are involved in protein kinase B signaling, ERBB signaling pathway, phosphatidylinositol 3-kinase signaling, and peptidyl-tyrosine phosphorylation. It is noteworthy also that these signaling pathways form a significant portion of the overall signal transduction pathways as summarized in the Reactome depiction of Fig. 8.

Fig. 8.

Signaling pathways linked to the identified hub genes and their interacting neighbors as outlined by red dashed rectangles

TF-miRNA regulatory elements

Every gene in the human genome is subjected to control by transcriptional factors as well as miRNA. This ensures the spatiotemporal regulation of this gene to perform its role as needed. The hub genes predicted in this study were found to be under control of at least 8 transcriptional factors, i.e. E2F1, ESR1, EGR1, SP1, CEBPB, JUN, ATF2, and SMAD4. On the flip side, has-miR-1, has-miR-27a and has-miR-27b are the main epigenetic regulators as depicted in Fig. 9. This would help pharmacologists to seek genetic-based therapies such as miRNA sponges and circRNA governing miRNA to control these hub genes.

Fig. 9.

TF-miRNA co-network of the hub genes as detected by NetworkAnalyst. miRNAs are shown in violet whereas TF in green

Differential expression

Differential expression of hub genes was also studied in tumor versus healthy sample groups via GEPIA platform. It turned out that FN1, MAPK13, KRT19 and EGFR all displayed higher level of expression compared to control. FN1 was the highest in terms of upregulation while EGFR was slightly elevated. Among all, only FN1 and KRT19 groups showed statistically significant (p < 0.01) difference (Fig. 10). This signifies the pivotal role of these genes in the etiology and pathophysiology of PTC.

Fig. 10.

Differential expression of the hub genes in tumor versus control group as retrieved from GEPIA 2 server. Red color represents PTC group while the grey color represents the control group

The correlation between hub gene expression and PTC stage was also checked. The results exhibited high correlation in the PTC stage with FN1 and, to a lesser degree, MAPK13 and KRT19. This correlation was statistically significant (P < 0.01). Nonetheless, EGFR failed to show significant association (p = 0.29) as provided in Fig. 11. This implies that all hub genes, except for EGFR, continue to differ in their expression level throughout the PTC progression.

Fig. 11.

Correlation of the hub genes expression in the different stages of PTC

We also evaluated the type and degree of infiltration of immune cells in PTC in response to hub genes differential expression. Our findings suggest the marked significant correlation between FN1 and KRT19 and tumor purity (p < 0.05). This means that the upregulation of these genes is highly expressed in tumor microenvironment. It should be also noted that as the correlation is negative (< 0) this indicates the reduced fraction of tumor cells in response to immune cells infiltration. With respect to immune cell type, only CD8 + T cells were significantly varied with both FN1 and KRT19 genes expression. Otherwise, all other immune cells (B cell, CD4 + T cell, macrophage, neutrophil and dendritic cells) displayed substantial degree of infiltration and highly correlated with all hub genes (p < 0.001) as depicted in Fig. 12.

Fig. 12.

Association among the expressed hub genes and the tumor purity as well as the infiltrated immune cells as predicted by TIMER tool

Survival prediction

The association of the four hub genes with overall survival in patients with PTC is also explored. It turned out that two (MAPK13 and KRT19) out of the four hub genes exhibited statistically significant difference (P < 0.05) between the low and high expression groups (Fig. 13). Hazard ratio (HR) predicts the death risk in high-expression group. As all the four genes had an HR value < 6, this mirrors the diminished probability of death in high-expression group. Collectively, it can be said that MAPK13 and KRT19 are significantly correlated with overall survival of PTC and, thereby, can be considered good prognostic markers.

Fig. 13.

Correlation of the expression of hub genes and the overall survival of the PTC as predicted by KM plotter

Hub genes–drug interaction

Finally, we scrutinized the GDIDB to identify the most potential drugs capable of blocking the four hub genes. Indeed, our scrutinization resulted in 231 total interactions segregated into EGFR (189), FN1 (13), KRT19 (2), and MAPK13 (27). The interaction score mirrors how strongly the hub genes are associated with PTC based on evidence from literature, experimental data, and computational predictions. Among these, only 60 drugs gained FDA approval while 171 unapproved yet. These drugs were found to fall within two categories, inhibitors (92) and antibodies (6, all of them are blockers of EGFR). The shortlisted approved drugs of each hub gene are summarized in Table 1. The list included efficacious inhibitors like Erlotinib, Imatinib, Dacarbazine, Pirfenidone and Dexamethasone. This list provides ready-to-use drugs against the most potential hub genes that pose the possible benefit for patients with PTC.

Table 1.

The identified drugs that already gained FDA-approval against the four hub genes

Gene	Drug	Regulatory approval	Indication	Interaction score
FN1	Dacarbazine	Approved	Antineoplastic agents	0.26927
EGFR	Gefitinib	Approved	Antineoplastic agent	0.085797
EGFR	Erlotinib	Approved	Antineoplastic agent	0.550928
EGFR	Cetuximab	Approved	Antineoplastic agent	0.648243
EGFR	Vandetanib	Approved	Antineoplastic agent	0.110169
EGFR	Lapatinib	Approved	Antineoplastic agent	0.277818
EGFR	Pemetrexed Disodium	Approved	Antineoplastic agent	0.022225
EGFR	Neratinib	Approved	Antineoplastic agent	0.396883
EGFR	Ibrutinib	Approved	Antineoplastic agent	0.016837
EGFR	Afatinib	Approved	Antineoplastic agent	1.184384
EGFR	Panitumumab	Approved	Antineoplastic agent	0.180582
EGFR	Leucovorin Calcium	Approved	Adjuvant to chemotherapy	0.015017
EGFR	Sorafenib	Approved	Antineoplastic agent	0.007311
EGFR	Imatinib	Approved	Antineoplastic agent	0.007408
EGFR	PEGinterferon α-2 A	Approved	Hepatitis B and C, antineoplastic agents, immunomodulatory agents	0.01916
EGFR	Rituximab	Approved	Antineoplastic agent	0.012347
EGFR	Bevacizumab-Awwb	Approved	Antineoplastic agent	0.006945
EGFR	Sirolimus	Approved	For treatment of wet age-related macular degeneration immunosuppressant	0.029244
EGFR	Tucatinib	Approved	Antineoplastic agent	0.034727
EGFR	Nivolumab	Approved	Antineoplastic agent	0.018521
EGFR	Necitumumab	Approved	Antineoplastic agent	0.694546
EGFR	Nitroglycerin	Approved	Raynaud’s disease	0.021371
EGFR	Ramucirumab	Approved	Antineoplastic agent	0.092606
EGFR	Paclitaxel	Approved	Peripheral arterial disease (PAD) Anti-inflammatory agent Antineoplastic agent	0.007717
EGFR	Tremelimumab	Approved	Antineoplastic agent	0.030869
EGFR	Ipilimumab	Approved	Antineoplastic agent	0.013891
EGFR	Pacritinib	Approved	Antineoplastic agent	0.030869
EGFR	Dacomitinib Anhydrous	Approved	Antineoplastic agent	0.60194
EGFR	Regorafenib	Approved	Antineoplastic agent	0.008419
EGFR	Uracil	Approved	Antineoplastic agent	0.046303
EGFR	Crizotinib	Approved	Antineoplastic agent	0.010685
MAPK13	Sorafenib	Approved	Antineoplastic agent	0.034118
MAPK13	Pirfenidone	Approved	For treatment of fibrotic conditions	0.277818
KRT19	Dexamethasone	Approved	For treatment of Meniere’s disease, glucocorticoid Anti-inflammatory agent	0.610554

Discussion

Papillary thyroid carcinoma (PTC) is the most common type of thyroid cancer, accounting for 80–93% of cases. It is defined by distinctive nuclear features including enlargement, elongation, overlapping, chromatin clearing, irregular contours, grooves, and pseudoinclusions [26]. PTC has a variable gross appearance and may appear anywhere in the thyroid gland, often as a firm, white lesion with calcifications. Microscopically, PTC shows papillary structures with fibrovascular cores lined by cells with the characteristic nuclear features. PTC frequently invades lymphatics, leading to multifocal lesions and lymph node metastases, which are common at presentation. While PTC usually has an indolent course, certain variants like tall cell, columnar cell, and solid variant have a poorer prognosis. The most frequent genetic alteration in PTC is the BRAF V600E mutation, particularly in tall cell and classic variants [27]. However, the full spectrum of genetic and molecular aberrations driving PTC pathogenesis remains an active area of research. Further elucidation of the key genes and pathways involved could enable the development of targeted therapies for high-risk and advanced PTC. So, the current study was sought, through utilizing the power of network pharmacology and systems biology, to fill this research gap.

Gene enrichment analysis of the mutual DEGs revealed that the plasma membrane (38.9%) and extracellular regions (24.4%) are key cellular compartments, suggesting that external signals interacting with plasma membrane receptors may drive PTC. Biological processes such as signal transduction (28.9%) and cell communication (27.8%) were highly enriched, indicating external signals triggering thyroid cell growth. Although not statistically significant, integrin cell surface interactions showed the highest enrichment (34.6%), supporting the hypothesis of signal-driven transformation. DEGs were primarily associated with extracellular matrix structural components and cell adhesion (6.2% each), with 4.6% showing receptor activity, highlighting thyroid cell dysfunction in PTC.

The PPI network generated using NetworkAnalyst was shortlisted to increase the predictions reliability via cytoscape by applying four-filtered criteria, namely MCC, MNC, closeness and betweenness. This is resulted in the identification of four hub genes (FN1, MAPK13, KRT19 and EGFR).

FN1 expression is highly correlated with poor prognosis and progression-free survival in PTC patients [28]. FN1 is involved in cellular adhesion and migration processes, and its upregulation is associated with increased tumor aggressiveness, invasiveness, and metastasis [29]. Mechanistically, FN1 appears to promote an invasive phenotype in PTC cells, potentially through epithelial-mesenchymal transition pathways. Furthermore, FN1 expression is correlated with immune infiltration levels and immune marker expression in PTC, suggesting it may serve as an immunity-related biomarker and therapeutic target. Importantly, the increased FN1 levels are detectable at the initial diagnosis, highlighting its potential as a prognostic biomarker for PTC [30].

MAPK13 is a distinct gene encoding one of the four p38 MAPK proteins, which are involved in various cellular processes including cell growth, differentiation, and apoptosis. Upregulation of MAPK13 expression has been observed in PTC samples compared to normal thyroid tissue [31]. Mechanistically, MAPK13 appears to promote an aggressive phenotype in PTC cells, potentially by maintaining cancer stem cell properties and driving tumorigenesis [32]. Furthermore, enhanced MAPK13 expression is associated with poor prognosis and clinical outcomes in PTC patients. These findings suggest that MAPK13 may serve as a prognostic biomarker and a therapeutic target for PTC [33].

EGFR is a receptor tyrosine kinase that is overexpressed in a significant proportion of PTC cases. Upregulation of EGFR expression and activation of EGFR-mediated signaling pathways, such as the MAPK and PI3K/Akt pathways, have been associated with increased tumor aggressiveness, invasion, and metastasis in PTC [34]. EGFR overexpression is also linked to poor prognosis and clinical outcomes in PTC patients. Mechanistically, EGFR appears to promote an invasive phenotype in PTC cells, potentially through epithelial-mesenchymal transition processes [35]. Furthermore, EGFR expression is correlated with immune infiltration and immune marker levels in PTC, suggesting it may serve as a prognostic biomarker and therapeutic target. Importantly, increased EGFR levels are detectable at the initial diagnosis, highlighting its potential clinical utility in PTC management [36].

All of them exhibited high level of expression in tumor tissue compared to control (p < 0.01). Moreover, the most interacting genes to those 4 hub genes involved converged role in protein kinase B signaling, ERBB signaling pathway, phosphatidylinositol 3-kinase signaling, in addition to peptidyl-tyrosine phosphorylation. This confirms the essential role of these genes in the development and progression of PTC across different stages since previous reports confirms the deep implication of these kinases in PTC [37, 38].

The findings presented in the TIMER server suggest that the upregulation of the hub genes FN1 and KRT19 is highly expressed in the tumor microenvironment of PTC, and the negative correlation between these genes and tumor purity indicates a reduced fraction of tumor cells, likely due to increased immune cell infiltration. While CD8 + T cells were the only immune cell type that showed a significant variation in relation to the expression of FN1 and KRT19 genes, all other immune cell types, including B cells, CD4 + T cells, macrophages, neutrophils, and dendritic cells, displayed a substantial degree of infiltration and were highly correlated with the expression of all hub genes.

Wang et al. [39] in this year found that CD4 + T cells represented 40% of the tumor infiltrate of PTC. Also, patients with thyroglobulin mutations typically exhibit higher infiltration of M2 macrophages and activated NK cells within the tumor microenvironment. In contrast, those with HRAS mutations tend to display decreased infiltration of M0 macrophages but increased presence of activated dendritic cells in the TME [40]. Our immune infiltration studies comply with these validated results.

Recent reports concluded that the upregulation of FN1 and KRT19 is highly expressed in the PTC tumor microenvironment [41, 42]. This is in-line with our findings obtained in this study.

Concerning survival analysis of our predicted hub genes, KMplotter found that MAPK13 and KRT19 are significantly correlated with overall survival of PTC and, thereby, can be considered good prognostic markers.

KRT19 (keratin 19) is a gene that is consistently upregulated in PTC tumors compared to normal thyroid tissue. High expression of KRT19 was found to be part of a 6-gene prognostic signature that can stratify PTC patients into high and low risk groups for survival [43]. Another study identified key oncogenes in PTC including LRRK2, SLC34A2, MUC1, FOXQ1 and KRT19. Their results showed overexpression and enrichment of these genes in oncogenic signaling axis of MAPK and PI3K/AKT in BRAF-enriched subtype [44]. These findings are in close agreement with our results in the present study.

The available drugs against the identified hub genes showed that EGFR had the highest number of interacting drugs, whether small molecules or monoclonal antibodies, since it is not specific to thyroid tissue but is widely distributed within the body and was a hotspot druggable target for other cancer types [45, 46].

In summary, the identified hub genes (FN1, MAPK13, KRT19, and EGFR) were predicted to be biomarkers of PTC which can be utilized for both diagnostic, prognostic, as well as druggable targets of treatment as found in the previous literature.

Limitations

First of all, this in silico study is computation-based and its findings should be validated in an experimental work using at least PCR, Second, confounding factors such as age, gender and disease stage have not taken into consideration in the initial data processing since all the provided specimens involved only diseased versus control information. These factors greatly affect the output results.

Conclusion

The investigation into PTC underscores the pressing need for enhanced prognostic markers and therapeutic targets to combat this prevalent malignancy effectively. The identification of four hub genes—FN1, MAPK13, KRT19, and EGFR—highlights their potential dual role as both prognostic indicators and druggable targets, particularly given their correlation with tumor expression levels, PTC staging, immune infiltration, and overall survival rates. Despite the promising findings, there remains a significant research gap regarding the comprehensive understanding of the molecular mechanisms underpinning PTC pathogenesis and the optimization of treatment strategies across diverse populations. To address these gaps, future research should focus on multi-omics approaches to elucidate the interactions among these hub genes and their pathways. Additionally, clinical trials assessing the efficacy of existing drugs targeting these genes in PTC patients are essential. Expanding access to radioactive iodine therapy and alternative treatment modalities in underserved regions is also crucial to mitigate disparities in outcomes. Finally, establishing international collaborations can facilitate data sharing and resource allocation, ultimately improving management strategies for PTC globally.

Author contributions

HA: Conceptualization, methodology, data curation, validation, drafting and reviewing the manuscript.

Funding

The present study received no funding.

Data availability

Two transcriptomes were downloaded from Gene expression Omnibus (GEO) database and analyzed in this study: GSE205733 and GSE237841.

Declarations

Competing interests

The authors declare no competing interests.