Abstract

Introduction

   The mechanism of thymoma-associated myasthenia gravis (TAMG) is
   currently unknown, although patients with TAMG experience more severe
   myasthenic symptoms and have worse prognoses compared to regular
   thymoma patients. The objective of this research is to create a
   transcriptome map of TAMG using genes linked to disulfidptosis, detect
   possible biomarkers, and examine the disparities in the tumor immune
   microenvironment (TIME) among different thymoma patients. The findings
   will offer valuable knowledge for personalized treatment alternatives.

Methods

   Thymoma samples' RNA-seq data, along with their corresponding clinical
   data, were acquired from the TCGA database using methods. Next, genes
   and disulfidptosis-related lncRNAs(DRLs) were chosen through
   correlation analysis. Then, a prediction model of TAMG was established
   by LASSO regression. Subsequent to that, an analysis of the mutation
   data, the tumor mutational burden (TMB), and the assessment of immune
   and stromal elements within the tumor microenvironment were conducted.

Results

   A total of 87 patients diagnosed with thymoma were included in the
   study, with 29 of them having TAMG. We discovered a group of 325
   lncRNAs in this sample that showed significant associations with genes
   related to disulfidptosis, with 25 of them displaying significantly
   altered expression. Moreover, utilizing LASSO regression, we
   constructed a predictive model incorporating 11 DRLs. The analysis
   revealed an area under the curve (AUC) of 0.934 (CI 0.879–0.989), a
   cut-off value of 0.797, along with a sensitivity of 82.8 % and
   specificity of 93.1 %. Furthermore, we examined the TIME in both the
   high-risk and low-risk groups, and observed noteworthy disparities in B
   cells, T cells, and APC among the two groups (p < 0.05).

Conclusion

   This research offers the initial examination of genes associated with
   disulfidptosis and TAMG through genomic and transcriptomic analysis.
   Furthermore, a strong risk forecasting model was created and the
   significance of TIME in TAMG was also clarified. The discoveries offer
   significant understanding into the molecular processes of TAMG and
   present possible indicators for categorizing risk.

   Keywords: Thymoma-associated myasthenia gravis(TAMG), Disulfidptosis,
   RNA-seq, Tumor immune microenvironment(TIME)

1. Introduction

   Thymic epithelial tumors (TETs) are rare cancers found in the chest
   area, which present considerable difficulties in treatment because of
   their aggressive characteristics [[33]1]. Cancer is caused by many
   changes within organisms and in the environment [[34]2]. The DNA of the
   most vertebrates especially mammals is depleted in CpG dinucleotides
   [[35]3]. The remaining CpGs clustering in DNA regions is generally
   referred to as CpG islands (CGIs). There has been growing interest in
   CGI because they are enriched in the promoters of genes [[36]3] and by
   altering DNA methylation in CGIs, they play important roles in the
   regulation of gene expression and gene silencing in biological
   processes [[37]4] and considerably, would help to discover the
   epigenetic causes of cancer [[38]5].

   Patients diagnosed with thymoma often experience autoimmune disorders,
   with around 33 % of cases showing these additional health conditions.
   Out of all the autoimmune conditions, myasthenia gravis (MG) stands as
   the prevailing one [[39]6]. Individuals diagnosed with
   thymoma-associated myasthenia gravis (TAMG) experience heightened
   myasthenic symptoms and face bleaker prognoses [[40]7]. While there are
   well-established diagnostic criteria for categorizing individuals with
   MG, a significant limitation exists due to the absence of objective
   biomarkers that can anticipate the advancement of the disease either in
   clinical settings or during clinical trials. Consequently, there is a
   pressing requirement for the creation of dependable biomarkers that can
   assist in the handling of these circumstances [[41]8].

   Currently, the precise understanding of TAMG remains uncertain,
   although scientists theorize that either the tumor cells or the
   atypical immune cells in the thymus gland might initiate an immune
   reaction against AChR or other elements of the neuromuscular junction
   [[42][9], [43][10], [44][11]]. The overexpression of specific genes has
   been linked to TAMG, and the reduced expression of CTLA-4 in thymoma
   could potentially heighten the vulnerability to developing MG
   [[45]12,[46]13].

   Notably, the recently identified process of disulfidptosis differs from
   other well-known types of cell demise, as it is linked to the actin
   cytoskeleton, an essential component of cells that is vital for
   preserving cellular morphology and enhancing viability [[47]14].
   Moreover, it is believed that actin plays a role in the development of
   TAMG, which is linked to thymoma. Thymic selection, a crucial step in
   the maturation of immune cells, is thought to involve the actin
   cytoskeleton, a protein filament network that offers cells structure
   and support. Nevertheless, there is presently a lack of studies on
   TAMG-associated genes related to disulfidptosis.

   Moreover, the epigenome comprising different mechanisms e.g. DNA
   methylation, remodeling, histone tail modifications, chromatin
   microRNAs and lncRNAs, interact with environ-mental factors like
   nutrition, pathogens, climate to influence the expression profile of
   genes and the emergence of specific phenotypes [[48]4,[49]15,[50]16].
   Multi-level interactions between the genome, epigenome and
   environmental factors might occur [[51][17], [52][18], [53][19]].
   Furthermore, numerous lines of evidence suggest the influence of
   epigenome variation on health [[54]3,[55]20,[56]21]. Also, the
   information obtained from the analysis of biological data by
   bioinformatics, in aligning sequences in information banks to find
   genetic similarities and differences, predicting the structure and
   function of gene products [[57]22,[58]23]. It helps to find the
   phylogenetic relationship between genes and protein sequences
   [[59]24,[60]25].

   Hence, the objective of this research was firstly to chart the
   transcriptome of TAMG using genes linked to disulfidptosis. Our goal
   was to detect potential biomarkers for the diagnosis and treatment of
   TAMG by screening risk genes and developing a prediction model.
   Furthermore, we examined the disparities in tumor immune
   microenvironment (TIME) among TAMG patients classified as high and
   low-risk groups, aiming to offer recommendations for personalized
   treatment choices. The findings of this research could greatly enhance
   our comprehension of the molecular processes that underlie TAMG and
   contribute to the creation of improved and tailored therapeutic
   approaches for this critical ailment.

2. Materials and methods

2.1. Raw data

   Data from the TCGA database (https//portal.gdc.cancer.gov/) included 94
   thymoma patients with clinical data and corresponding transcriptome
   RNA-seq data of 92 samples (2 normal cases and 90 tumor cases).
   Additionally, mutome data of the patients were also obtained. Out of
   the 90 tumor samples in the transcriptome data, 3 cases did not have
   information on MG. [61]Supplementary Table 1 showed the detailed raw
   data. Gene IDs from the Human Genome Annotated Data (GRCh38) were used
   to annotate the clinical data of the patients. The annotations were
   carried out utilizing the online resource found at
   [62]https://asia.ensembl.org/index.html. The research follows the data
   access protocols and publication standards set by TCGA.

2.2. Development of a genetic marker linked to myasthenia gravis and
disulfidptosis

   Genes associated with disulfidptosis were obtained from the available
   literature [[63]14]. In particular, disulfidptosis-related lncRNAs
   (DRLs) were chosen using correlation analysis (with a corFilter of 0.4
   and pvalueFilter of 0.001). The variable selection and shrinkage
   process was performed using the ‘glmnet’ R package, utilizing the Least
   Absolute Shrinkage and Selection Operator (LASSO) algorithm. In the
   TCGA cohort, the status of patients' TAMG was determined by the DRLs,
   which served as the independent variable. Tenfold cross-validation was
   used to determine the penalty parameter (λ) for the model, based on the
   minimum criteria. This criteria corresponds to the value of λ that
   yields the lowest partial likelihood deviance. The risk scores of
   patients were determined by evaluating the expression level of every
   gene and their respective regression coefficients, following the
   formula score = sum (expression of each gene × corresponding
   coefficient) [[64]26]. Subsequently, the predictive capability of the
   genetic pattern was assessed through analysis of the ROC curve.
   Ultimately, the individuals were categorized into high-risk and
   low-risk categories by utilizing the median value of the risk score
   [[65]26]. Subsequently, principal component analysis (PCA) was
   performed using the ‘prcomp’ function from the 'stats' R package,
   considering the gene expression in the signature [[66]27].

2.3. Gene set enrichment

   The ‘clusterProfiler’ R package was utilized to conduct analysis on the
   differentially expressed genes (DEGs) with a fold change of
   |log2FC| ≥ 1 and a false discovery rate (FDR) of <0.05 between the
   high-risk and low-risk groups, using Gene Ontology (GO) and Kyoto
   Encyclopedia of Genes and Genomes (KEGG) methods. The target sets for
   Gene Set Enrichment Analysis (GSEA) were downloaded from the Molecular
   Signatures Database as Analysis Hallmark and
   c2.cp.kegg.v7.2.symbols.gmt. GSEA-3.0 software from Broad Institute was
   used to perform the analysis.

2.4. Mutation landscape

   We utilized the R package maftools [[67]28] to analyze the mutation
   data, enabling us to assess the enrichment of oncogenic pathways in
   primary tumors from TCGA. TMB, which stands for tumor mutational
   burden, was determined by analyzing SNV data from primary tumors. TMB
   represents the overall count of somatic mutations found in a tumor
   sample per megabase (Mb).

2.5. Generation of ImmuneScore, StromalScore, and ESTIMATEScore

   For each sample, the immune-stromal component ratio in the tumor
   microenvironment (TME) was calculated using the ESTIMATE algorithm,
   which was implemented in R version 4.0.2 and utilized the estimate
   package. The algorithm produced three scores: ImmuneScore,
   StromalScore, and ESTIMATEScore, which represented the proportion of
   immune, stromal, and the combined sum of both elements, respectively
   [[68]29]. A greater percentage of the corresponding component in the
   TME was indicated by higher scores. Following the analysis, the
   CIBERSORT computational technique was utilized to estimate the
   abundance profile of tumor-infiltrating immune cells (TIC) in every
   tumor sample [[69]30].

2.6. Statistical analysis

   Before conducting the analysis, numerical and percentage
   representations were used to describe categorical variables, while
   means and standard deviations were employed to express measurement data
   that followed a normal distribution. Data that did not follow a normal
   distribution were described by calculating the medians and
   interquartile range (IQR). To compare the variations in proportions
   between groups, we conducted Pearson's χ2 test, Fisher's exact test,
   and nonparametric tests. Additionally, we calculated correlations using
   Pearson correlation analysis and Spearman rank correlation analysis.
   The statistical analysis was conducted utilizing R software (version
   4.2.2). In conclusion, any test with a p value below 0.05 was deemed to
   have statistically significant results, and all p values mentioned in
   the document are two-tailed unless specified otherwise.

3. Results

3.1. 1Clinical characteristics

   [70]Fig. 1 displays the depicted workflow diagram for this study. A
   total of 87 patients diagnosed with thymoma were included in our study,
   with 29 of them having MG. The individuals diagnosed with MG had a
   median age of 55 years (ranging from 17 to 79 years), comprising of 14
   male and 15 female patients. On the other hand, those without MG had a
   median age of 61 years (ranging from 39 to 81 years), consisting of 30
   male and 28 female patients. Six distinct categories were assigned to
   the thymic tumors, including A type (n = 11), AB type (n = 26), B1 type
   (n = 10), B2 type (n = 22), B3 type (n = 9), and thymic carcinoma (TC)
   (n = 9) ([71]Supplementary Table 2). No significant disparities in age,
   gender, or Masaoka stage were found when comparing patient
   characteristics in the two groups. Nevertheless, we noticed a notable
   disparity in the allocation of WHO categorization [[72]31] (p < 0.01)
   ([73]Supplementary Table 2).

Fig. 1.

   [74]Fig. 1
   [75]Open in a new tab

   The workflow chart of this study. DRLs, disulfidptosis-related lncRNAs.

3.2. Disulfidptosis-related gene expression in TAMG

   Through the examination of gene co-expression, we discovered a
   collection of 325 lncRNAs that displayed a strong correlation with
   genes related to disulfidptosis ([76]Fig. 2A). In thymoma patients, we
   examined the expression of these DRLs in both MG+ and MG-groups. We
   discovered 25 DRLs that exhibited significant differences in
   expression, with 10 showing up-regulated expression and 15 showing
   down-regulated expression ([77]Fig. 2B-C, [78]Supplementary Table 3).

Fig. 2.

   [79]Fig. 2
   [80]Open in a new tab

   Identification of the DRLs signature using LASSO regression. (A) The
   co-expression analysis of disulfidptosis-related genes and lncRNAs. (B)
   Heatmap analysis of DRLs in patients with thymoma with MG and patients
   without MG. (C) Differential expression of DRLs between thymoma with MG
   and thymoma without MG. (D) The log (lambda) sequence plot of DRLs
   using LASSO regression. (E) The LASSO coefficient profiles of DRLs. (F)
   DRLs and coefficients selected by LASSO regression. DRLs,
   disulfidptosis-related lncRNAs; LASSO, Least Absolute Shrinkage and
   Selection Operator; MG, myasthenia gravis.

   In order to further investigate the possibility of inferring TAMG
   status using gene expression profiles, we additionally utilized a LASSO
   regression model ([81]Fig. 2D–E). The findings showed that a model
   consisting of 11 genes ([82]AC010307.4, [83]AC137735.2, [84]AC137735.2,
   [85]AL596442.3, [86]AL596442.3, [87]AL596442.3, ZBTB20-AS4,
   [88]AL353804.2, [89]AL353804.2, [90]AL353804.2, and [91]AL353804.2)
   demonstrated the highest accuracy in predicting outcomes ([92]Fig. 2F).
   Further investigation was conducted on the SLC7A11, SLC3A2, RPN1,
   NCKAP1, NUBPL, NDUFA11, LRPPRC, OXSM, NDUFS1, and GYS1 mRNAs, as well
   as the co-expressed DRLs that were identified. The correlation analysis
   indicated that [93]AL359232.1 exhibited a positive correlation with
   NUBPL (r = 0.506, p < 0.001), GYS1 (r = 0.408, p < 0.001), and OXSM
   (r = 0.364, p < 0.01), while demonstrating a negative correlation with
   RPN1 (r = −0.345, p < 0.001). Furthermore, NDUFS1 showed a positive
   correlation with ZBTB20−AS4 (r = 0.557, p < 0.001), [94]AC018521.6
   (r = 0.769, p < 0.001), NCKAP1 (r = 0.705, p < 0.001), LRPPRC
   (r = 0.690, p < 0.001), and NUBPL (r = 0.570, p < 0.001), while
   exhibiting a negative correlation with NDUFA11 (r = −0.506, p < 0.001)
   ([95]Fig. 3A).

Fig. 3.

   [96]Fig. 3
   [97]Open in a new tab

   A predictive model incorporating 11 DRLs for predicting TAMG. (A)ROC
   curves verified the TAMG of the risk score. (B) Association of the 11
   DRLs included in the prediction TAMG model with disulfidptosis-related
   genes. DRLs, disulfidptosis-related lncRNAs; TAMG, thymoma-associated
   myasthenia gravismyasthenia gravis; ROC, receiver operator
   characteristic curve.*p < 0.05, **p < 0.01, ***p < 0.001.

   In [98]Fig. 3B, the AUC for predicting TAMG was 0.934 (CI 0.879–0.989),
   with a cut-off of 0.797, demonstrating a sensitivity of 82.8 % and
   specificity of 93.1 %. After calculating the median risk score as
   0.347, we categorized thymoma patients into low- and high-risk groups
   depending on whether their risk score was below or above the median.
   PCA was subsequently conducted to assess the capacity of the chosen
   genes in discriminating high- and low-risk TAMG groups from the entire
   gene set. The outcomes demonstrated that the chosen genes were capable
   of effectively distinguishing between the two groups, thereby
   confirming the reliability of our risk score model ([99]Fig. 3C–D).

3.3. Functional and pathway enrichment analysis of genes with differential
expression

   Following the completion of GO functional enrichment analysis, it was
   discovered that the genes associated with differential risk were
   notably enriched in functions pertaining to the extracellular matrix
   containing collagen, organization of external encapsulating structures,
   and the structural constituent of the extracellular matrix ([100]Fig.
   4A). Furthermore, the KEGG pathway enrichment analysis revealed a
   significant enrichment of these genes in the ECM receptor interaction
   pathway, focal adhesion pathway, and the PI3K−Akt signaling pathway
   ([101]Fig. 4B). In addition, GSEA analysis showed that the high-risk
   group had significant enrichment of pathways related to oxidative
   phosphorylation, ribosomes, and systemic lupus erythematosus. On the
   other hand, the low-risk group had significant enrichment of pathways
   such as ECM receptor interaction, focal adhesion, pathways in cancer,
   and the TGF beta signaling pathway ([102]Fig. 4C–D).

Fig. 4.

   [103]Fig. 4
   [104]Open in a new tab

   Functional and pathway enrichment analysis of differentially expressed
   DRLs. (A)Bubble graph of GO enrichment analysis of differentially
   expressed DRLs. (B) KEGG analysis of differentially expressed DRLs.
   (C)GSEA analyses of differentially expressed DRLs in high-risk group.
   (D)GSEA analyses of differentially expressed DRLs in low-risk group.
   DRLs, disulfidptosis-related lncRNAs; GO, Gene Ontology; KEGG, Kyoto
   Encyclopedia of Genes and Genomes; GSEA, Gene Set Enrichment Analysis.

3.4. The mutational landscape of TETs

   Missense mutation was the most frequent variant classification in both
   the high-risk and low-risk groups, while SNP was the most prevalent
   variant type. [105]Fig. 5A shows the high-risk group's top 10 mutation
   genes as GTF2I, TTN, RAB3GAP2, NEXMIF, MYO15A, MAGEB18, CSMD2, COL6A6,
   ABCC1, and ABCA7. The low-risk group had the following top 10 mutation
   genes: GTF2I, HRAS, TTN, MUC16, ADGRV1, TP53, PCLO, L3MBTL3, DNAH17,
   and CYLD ([106]Fig. 5B). Moreover, the TMB of the low-risk cohort
   exceeded that of the high-risk cohort (0.368 compared to 0.249,
   p = 0.00062) ([107]Supplementary Fig. 1).

Fig. 5.

   [108]Fig. 5
   [109]Open in a new tab

   The Mutational Landscape of TETs. (A)The mutation information of genes
   with high mutation frequency in high-risk group. (B)The mutation
   information of genes with high mutation frequency in low-risk group.
   TETs, thymic epithelial tumor.

3.5. The immune microenvironment of the high- and low-risk categories

   The high-risk group, analyzed with the ESTIMATE algorithm, exhibited
   elevated ImmuneScore (2212 vs. 1,826, p < 0.01) and decreased
   StromalScore (−246.2 vs. −567.2, p＜0.01) in comparison to the low-risk
   group ([110]Fig. 6A). In the high- and low-risk groups, [111]Fig. 6B
   displays the proportions of different types of immune cells including B
   cells, T cells, natural killer cells, macrophages, dendritic cells
   (DC), and their subcategories.The group at high risk exhibited a higher
   number of plasma cells, an increased presence of T cell follicular
   helper (Tfh), and a greater activation of DC compared to the low-risk
   group (all p＜0.05). Nevertheless, there was no notable disparity in
   CD8^+ T lymphocytes or T regulatory cells among the two cohorts
   ([112]Fig. 6C). Furthermore, the variances in immune function scores
   between the high and low risk groups were primarily focused on APC
   co-inhibition, B cells, checkpoint, cytolytic activity, DCs, HLA, mast
   cells, T cell co-inhibition, T cell co-stimulation, Tfh, Th1, and
   TIL.In all of these aspects, the high-risk group outperformed the
   low-risk group ([113]Fig. 6D).

Fig. 6.

   [114]Fig. 6
   [115]Open in a new tab

   The Immune Microenvironment of the High and Low Risk Groups.
   (A)ESTIMATE algorithm in high- and low-risk groups. (B)The percentages
   of various immune cells in high- and low-risk groups. (C)The immune
   cell infiltration in high- and low-risk groups. (D)The immune function
   scores in high- and low-risk groups. *p < 0.05, **p < 0.01,
   ***p < 0.001.

4. Discussion

   For the first time, a thorough examination of genes associated with
   disulfidptosis was conducted in this study to investigate TAMG,
   yielding noteworthy findings that establish a notable connection
   between disulfidptosis-related genes and TAMG. Furthermore, an
   exceptional risk prediction model was developed with outstanding
   predictive capabilities. Using our risk prediction model, we examined
   the TIME in high- and low-risk categories and observed noteworthy
   variances in B cells, T cells, and APC between these two groups. The
   results indicate that the genes linked to increased risk in MG might
   engage with immune cells, resulting in the disruption of muscle actin
   and ultimately contributing to the emergence of TAMG.

   Currently, there is still a lack of complete comprehension regarding
   the pathological and physiological mechanisms that underpin TAMG.
   Further research is needed to explore the specific mechanisms behind
   thymoma, as numerous studies have indicated its association with immune
   system abnormalities and their impact on the immune system's
   recognition of self-antigens [[116]10,[117]32]. Furthermore, TAMG could
   potentially be linked to other elements like irregularities in
   neurotransmitters and harm to the neuromuscular junction. Hence,
   additional investigation is required to acquire a more profound
   comprehension of the etiology and therapeutic approaches for this
   ailment. With the advancement of gene sequencing technology in recent
   years, it is possible that specific transcriptome genes play a role in
   the progression of TAMG. An instance of thymoma exhibits high levels of
   expression of a gene for mid-sized neurofilament, which contains
   similar genetic codes for acetylcholine receptors and titin epitopes,
   and is frequently associated with MG [[118]12]. For example, NEFM
   displays restricted sequence resemblance to primary autoimmune targets,
   while RYR3 demonstrates significant similarity [[119]12]. Furthermore,
   prior research has shown that alterations in the expression of lncRNAs
   can play a crucial role in differentiating TAMG and uncovering the
   molecular mechanisms that contribute to its development
   [[120]33,[121]34]. Nevertheless, there has been no inquiry into the
   possible participation of DRLs. Hence, considering the proof provided
   in this document, it seems that the expression of genes associated with
   disulfidptosis can control multiple signaling pathways that play a role
   in the formation of TAMG.

   Initially, we employed correlation analysis to identify genes
   associated with disulfidptosis. Furthermore, LASSO regression
   ([122]Fig. 2F) enabled the identification of 11 genes associated with
   TAMG. The involvement of these genes encompasses a range of cellular
   processes, including cellular transportation and the synthesis of
   proteins. Crucially, all 11 genes were discovered to have a connection
   with disulfidptosis. Moreover, the well-established risk assessment
   model demonstrated excellent diagnostic effectiveness for TAMG,
   suggesting a strong correlation between TAMG and disulfidptosis in
   tumors. Nevertheless, there have been no reports linking TAMG with
   disulfidptosis-related genes like SLC7A11, SLC3A2, and GYS1. Apoptosis
   in animal experiments previously entailed caspase-mediated gelsolin
   cleavage, resulting in heightened actin filament fragmentation and
   cleavage of the myosin-activating kinase ROCK. As a consequence,
   myosin-induced contractile activity was hyperactivated, leading to the
   generation of blebs in the process of apoptosis [[123]35]. Identifying
   a novel mechanism underlying disulfidptosis, characterized by the
   collapse of the cytoskeleton initiated by disulfide bond formation,
   opens up exciting avenues for exploring potential targets to modulate
   this specific type of cell death [[124]36]. TAMG may occur due to the
   overexpression of the mentioned risk genes, which can potentially cause
   malfunctioning of myosin and disruption of the cellular skeleton,
   ultimately resulting in the progression of TAMG.

   The TETs in TIME is composed of various groups of innate and adaptive
   immune cells with different characteristics that display a varied
   distribution throughout the tumor [[125]37]. Thymic cancers are linked
   to disturbed structure of the thymus and decreased or missing presence
   of MHC class II and AIRE factor. The outcome of this is expected to
   lead to the unsuccessful process of selecting both positive and
   negative thymocytes or a decrease in the generation of regulatory T
   cells [[126]10]. In this research, we discovered that there were
   variations in both the quantity and performance of B lymphocytes among
   the high- and low-risk categories. Recent studies have focused on
   exploring and understanding the role of B cells in tumor immunity and
   their connection to the presence of thymic germinal centers (GCs) in
   the neighboring tissues of thymomas. These thymomas are enclosed by
   high endothelial venules (HEVs), which are recognized for their ability
   to attract peripheral cells. Furthermore, individuals with TAMG exhibit
   elevated concentrations of IFN-γ, IL-1β, and sCD40L in their serum
   [[127]38].

   The reduction in thymic macrophages hinders the removal of dying
   thymocytes and promotes the liberation of internal genetic material
   from dead thymic cells. In this inflammatory state, thymic epithelial
   cells have the potential to excessively produce IFN-β, which
   specifically triggers α-AChR expression, causing self-sensitization and
   thymic changes that ultimately lead to AChR-MG. The results emphasize
   the intricacy of the immune system [[128]39]. In our study, we observed
   variations in B cells between the high- and low-risk groups of TAMG.
   However, it is worth noting that the disparity in IFN did not reach
   statistical significance. The complexity of thymoma's TIME and the
   variety of factors contributing to TAMG are implied by this.

   Besides B cells, which are the main contributors to the production of
   autoantibodies, T lymphocyte subsets, particularly T helper cells, also
   have a vital function in TAMG. T helper cells have the ability to start
   and maintain long-lasting inflammation, enhance inflammatory signals,
   influence class switching and somatic hypermutation, and transform B
   cells into plasma cells that generate autoantibodies [[129]11].
   Ultimately, these processes result in the growth and advancement of MG.

   Due to the absence of any immunological signature or lymphocyte
   activation in TAMG, it seems more likely that the prevailing mechanism
   underlying targeted anti-muscle autoimmunity in TAMG is “false-positive
   selection” driven by MHC-bound autoantigen-derived peptides [[130]12],
   though there may be other underlying mechanisms. The immune
   dysfunctions witnessed in individuals with thymoma are probably a
   result of a blend of mechanisms. Additional research is required to
   clarify the fundamental processes accountable for immune dysfunction in
   individuals with thymoma.

   The removal of the thymus gland can be considered as a successful
   treatment in changing the development from ocular myasthenia gravis to
   generalized myasthenia gravis [[131]40]. TETs exhibit high expression
   of PD-1 and PD-L1, two vital immune checkpoints, which can potentially
   indicate the progression of TETs and the effectiveness of
   immunotherapy. Nonetheless, the use of ICIs in TETs is associated with
   a notably elevated occurrence of immune-related adverse events (irAEs)
   in contrast to other forms of cancer [[132]41]. Despite the ongoing
   research on various immunotherapies for TETs, most of them are still in
   the experimental or preclinical phases.

   In recent times, there has been a shift in cancer treatment towards
   therapies that target molecules. Nevertheless, the rarity and
   histological heterogeneity of TET pose a challenge to new treatment
   development [[133]42]. Individuals identified with TETs, irrespective
   of their particular histopathology, additionally exhibit an
   inexplicable tendency towards the emergence of autoimmune toxicity
   impacting the muscular or neuromuscular system [[134]43]. Hence, these
   DRLs have the potential to function as indicators for distinguishing
   thymoma with or without MG. Moreover, these molecules associated with
   the development of diseases may hold promise as targets for treatment.

   While this study has produced significant findings, it is crucial to
   acknowledge the various constraints that should be taken into account.
   Further research is needed to validate the precise molecular functions
   of the identified DRLs, as the initial sample size was relatively
   small, thus requiring additional support for our findings. Furthermore,
   the dataset employed did not contain data on AChR, which is recognized
   to have a connection with TAMG. Hence, it is necessary to conduct
   further investigations to examine the possible correlation between the
   lncRNAs analyzed in the study and AChR. This exploration could offer
   valuable additional knowledge regarding the fundamental mechanisms of
   TAMG.

   A limitation of this study is the reliance on retrospective data
   sources. Future research could benefit from the inclusion of
   prospective data. Additionally, while our findings are promising, the
   sample size could be expanded further. Incorporating a multi-center
   patient cohort would not only enhance the generalizability of our
   results but also provide a more robust validation of the identified
   lncRNAs as potential biomarkers for TAMG. Moreover, future studies
   could employ single-cell sequencing technologies to offer a more
   detailed interpretation of immune cell infiltration within the TIME.

5. Conclusion

   By conducting bioinformatic analysis on thymoma patients from the TCGA
   dataset, we discovered a group of genes related to disulfidptosis that
   were linked to TAMG. Among these, 11 DRLs stood out for their
   association with TAMG, specifically [135]AL359232.1, ZBTB20-AS4,
   [136]AC002347.1, KLF7-IT1, among others. Utilizing this information, we
   created a risk prediction model for TAMG. Functional pathways
   associated with the immune and nervous systems may be involved in
   regulating the development of TAMG through the target genes of the
   identified lncRNAs. The high-risk genes may interact with immune cells,
   causing dysregulation of muscle actin and leading to the development of
   TAMG. The results indicate a notable progress in the TAMG field and
   establish a basis for future studies on the molecular mechanisms
   involved in disulfidptosis and its impact on cellular balance.

Ethics declarations

   Review and/or approval by an ethics committee was not needed for this
   study because the information was obtained from the TCGA repository,
   and it was utilized in a morally sound manner.

Agreement to make information public

   Does not apply.

Data availability statement

   Data from TCGA database ([137]https://portal.gdc.cancer.gov/) including
   transcriptome RNA-seq data and relevant clinical data were obtained.
   The analyzed datasets generated during the study are available from the
   corresponding author upon reasonable request.

Funding

   The writers affirm that they did not receive any financial assistance,
   grants, or other forms of support while preparing this manuscript.

CRediT authorship contribution statement

   Yue Pan: Formal analysis, Data curation, Conceptualization. Hongsheng
   Deng: Validation, Supervision. Chao Yang: Validation, Software. Lixuan
   Lin: Software, Methodology. Qi Cai: Supervision, Software. Jianxing He:
   Writing – review & editing.

Declaration of competing interest

   The authors declare that they have no known competing financial
   interests or personal relationships that could have appeared to
   influence the work reported in this paper.

Acknowledgements