Identification and verification on prognostic index of glioblastoma immune-related lncRNAs

Background. Glioblastoma (GBM) is the most common cause of primary brain malignancy. Recently, many immune-related long noncoding ribonucleic acids (ir-lncRNAs) are indicated to be closely related to the regulation of the immune microenvironment and immune cell infiltration of GBM. Objectives. Through the joint analysis of multiple public databases, key ir-lncRNAs in GBM were screened. The ir-lncRNAs were used to construct risk-scoring models and promote the development of novel GBM biomarkers. Materials and methods. In this study, we performed a three-way Venn analysis combined with a least absolute shrinkage and selection operator (LASSO) regression analysis on all lncRNAs in The Cancer Genome Atlas (TCGA), the Chinese Glioma Genome Atlas (CGGA) and Imm-Lnc datasets, and identified 10 ir-lncRNAs. Multivariate Cox analysis was used to calculate the coefficient and construct a risk-scoring model. Results. By plotting calibration curves and receiver operating characteristic (ROC) curves, the model showed excellent prediction results. Based on the Tumor Immune Estimation Resource (TIMER) database, the correlation analysis showed that 10 ir-lncRNAs risk scores were related to immune cell infiltration. The enrichment analysis was subsequently performed, which showed that these ir-lncRNAs played an important role in the progression of GBM. Among the 10 lncRNAs, we found that AL354993.1 was highly expressed in GBM, had not been reported, and was shown to be closely related to GBM progression. Conclusions. In conclusion, the 10 ir-lncRNAs have the potential to predict the prognosis of GBM patients and may play a vital role in the progression of the disease.


Background
The most aggressive primary brain malignancy originates from oligodendrocyte or astrocyte precursor cells and is known as glioblastoma (GBM).Although accurate surgical resection, radiation and adjuvant chemotherapy are now the conventional treatments for GBM, the prognosis is still poor, and the median survival is just 8-15 months. 1 Immunomodulatory therapy is a new and effective treatment option. 2 The stemness features of GBM are strictly connected to immune infiltration, 3 meaning neoadjuvant anti-programmed cell death protein 1 (PD-1) checkpoint blocking immunotherapy might improve the prognosis of properly selected GBM patients. 4However, GBM often exhibits severe local immunosuppression, which limits the efficacy of immunotherapy strategies. 5To further explain the mechanisms of immune regulation in GBM and offer a theoretical basis for GBM immunological treatment, we evaluated effective immune-related prognostic factors and constructed a prognostic model for GBM patients.
Long noncoding ribonucleic acids (lncRNAs) are a group of transcripts with a length of more than 200 nt that primarily function as regulators rather than protein-coding genes. 68][9] The lncRNAs are also crucial for GBM progression.The lncRNA HNF1A-AS1 was shown to drive GBM progression through the microRNA (miR)-22-3p/alpha-enolase 1 (ENO1) axis. 10Indeed, lncRNA miR155HG has been shown to promote GBM progression by upregulating annexin A2 (ANXA2) as a competing endogenous RNA (ceRNA) of the tumor suppressor miR-185. 11mmune-related lncRNAs (ir-lncRNAs) are involved in regulating the GBM immune microenvironment and have unique prognostic value.According to reports, lncRNA AC003092.1 is connected to the immunosuppressive environment in GBM. 12 Moreover, maternally expressed 3 (MEG3) levels are negatively associated with dendritic cell infiltration and positively correlated with infiltrating CD8 + T cells.The survival of GBM patients was also significantly correlated with the degree of MEG3 variation in copies. 13he heat shock protein family A member 7 (HSPA7) lncRNA was found to promote macrophage recruitment to the GBM tumor microenvironment and had a great prognostic value. 14owever, few investigations have established prognostic models based on the identification of ir-lncRNAs in GBM.

Objectives
The ir-lncRNAs obtained from the Chinese Glioma Genome Atlas (CGGA; http://www.cgga.org.cn/) and The Cancer Genome Atlas (TCGA; https://www.cancer.gov/ccg/research/genome-sequencing/tcga)were examined.The clinical prognostic model of GBM was developed after the least absolute shrinkage and selection operator (LASSO) algorithm identified the most critical lncRNAs.Additionally, the underlying pathway of ir-lncRNAs in GBM was investigated.

Data and resources
Both TCGA-GBM (n = 166) and CGGA cohorts (n = 140) were used as public transcriptome datasets in our analysis. 15Any case with a survival information null value was eliminated.The UCSC Xena database was used to retrieve the clinical information and fragments per kilobase per million (FPKM) data for the TCGA-GBM cohort (https:// xenabrowser.net/).Transcripts per kilobase million (TPM) values were obtained from all FPKM data.The RNA-sequencing (RNA-seq) data of 140 specimens were retrieved from the CGGA data collection in addition to the clinical data for use as a validation set.The TCGA database provided gene mutation data (MAF files) for the TCGA-GBM group.The proportional hazards assumption test, linearity assumption test and multicollinearity test assessed the TCGA and CGCA cohorts (Supplementary Fig. 1-3).

Detection of immune-related lncRNA prognostic signature
The ImmLnc database (http://bio-bigdata.hrbmu.edu.cn/ImmReg/index.jsp)has collected 3115 GBM ir-lncRNAs. 16y evaluating the intersection of lncRNAs among the TCGA, CGGA and ImmLNC datasets, we selected ir-lncRNAs.To prevent overfitting and examine the ideal ir-lncRNA signature for estimating the overall survival of GBM individuals, the LASSO was selected.The LASSO regression analysis was performed using the "glmnet" R program.Every sample's risk score was determined from the formula: risk score = expression value of lncRNA 1 × coefficient + expression value of lncRNA 2 × coefficient + … + expression value of lncRNA n × coefficient.Then, depending on the middle threshold of the risk score, GBM patients were allocated into elevated-and reduced-risk cohorts.The "Survival" program of R software's area under the curve (AUC) function was employed to verify the specificity and sensitivity of the immune-related signature.

Nomogram
To anticipate the 1-, 2-and 3-year survival rates, a nomogram was developed after the independent prognostic parameters were identified.Receiver operating characteristic (ROC) curves were employed to assess the effectiveness of the model.Additionally, calibration plots were shown utilizing the rms tool to compare the model-predicted survival with the actual survival probability.

Gene set enrichment analysis
Kyoto Encyclopedia of Genes and Genomes (KEGG; https://www.genome.jp/kegg/)and Gene Ontologygy (GO; https://geneontology.org/) mechanisms that positively related to elevated-or reduced-risk scores were investigated using gene set enrichment analysis (GSEA; https://www.gsea-msigdb.org/gsea/index.jsp).Molecular Signatures Database gene sets were obtained.Typically, 1000 permutations were used in the analysis, and pathways with a false discovery rate (FDR) of less than 0.25 were detected.

Estimation of cancer immune microenvironment
The Tumor Immune Estimation Resource (TIMER) (timer.cistrome.org/)platform 17 was employed to investigate the connections between risk score and immune infiltrates, such as B cells, CD8 + T cells, CD4 + T cells, macrophages, neutrophils, and dendritic cells, as demonstrated by the purity-corrected partial Spearman approach.Depending on gene expression patterns, the Estimation of STromal and Immune Cells in MAlignant Tumor tissues using Expression data (ESTIMATE) program evaluated the stromal scores and immune scores. 18In addition, we utilized the tumor immune dysfunction and exclusion (TIDE algorithm; http://tide.dfci.harvard.edu/)to evaluate each participant's potential reaction to immune checkpoint inhibitor (ICI) treatment. 19

Statistical analyses
The TCGA-GBM cohort contains 166 tumor samples from patients with GMB, while the CGGA cohort includes 140 tumor samples from patients with GMB and normal tissue from 20 patients.For the TCGA-GBM cohort, we divided patients into 2 groups based on the risk score: highrisk group (n = 83) and low-risk group (n = 83).In addition, we used the TIDE algorithm to predict the responsiveness of patients in the TCGA-GBM cohort to immunotherapy, with 48 patients evaluated as responders and 120 evaluated as non-responders.According to isocitrate dehydrogenase 1 (IDH1) mutation status, patients in the CGGA cohort were divided into 2 groups: wild-type (wt) group (n = 100) and mutant (mut) group (n = 40).Based on the co-deletion status of x1p19q, patients in the CGGA cohort were divided into 2 groups: Non_codel group (n = 128) and codel group (n = 12).Additionally, patients in the CGGA cohort were divided into 2 groups based on the expression level of AL354993.1:low-expression group (n = 70) and highexpression group (n = 70).
Continuous variables were tested for normality using Kolmogorov-Smirnov or Shapiro-Wilk tests.When the sample size was ≤50, the Shapiro-Wilk test was employed.Otherwise, the Kolmogorov-Smirnov test was used.The variables were considered to conform to a normal distribution when p > 0.05.When performing a difference analysis for 2 sets of variables, an F-test was used to evaluate the homogeneity of variance between the 2 groups.The variance between the 2 variables was considered equal when p > 0.05.The results of the tests assessing the assumptions are provided in Supplementary Tables.
When the variables conformed to a normal distribution and the variance was equal, Student's t-test was used to compare between-group differences.If at least one of the assumptions was violated, the Mann-Whitney U test was used to compare the differences between the groups.For Student's t-test, we also calculated the test values and degrees of freedom (df).For the Mann-Whitney U test, we calculated the U and Z values.The χ 2 or Fisher's exact tests were used to compare the differences between the groups for categorical variables.When the total sample size was >40 and the minimum theoretical frequency was >5, the χ 2 test was used.When the total sample size was >40 and 5>, and the minimum theoretical frequency was >1, the corrected χ 2 test was used.If the total sample size was <40 or the minimum theoretical frequency was <1, Fisher's exact test was used.The results of tests assessing the differences between the groups are presented in tabular form and illustrated using box-and-whisker plots, which contain 5 lines representing the estimated maximum upper quartile (Q U ), median lower quartile (Q L ), and estimated minimum of the data from top to bottom.Outliers were defined as a value greater than Q U + 1.5 × QR or less than Q L -1.5QR, where QR = Q U -Q L .
The Cox proportional hazards model was used to evaluate the impact of clinical parameters on patient survival time.In terms of parameter selection, we evaluated the clinical parameters common to both TCGA and CGGA cohorts, and age and gender were included in both.Therefore, we included 3 parameters: patient age, gender and the risk score calculated using the Cox regression model.First, we performed the proportional hazards assumption test on all 3 parameters (Supplementary Fig. 1).When the Schoenfeld individual test p-value was less than 0.05, the proportional hazards assumption was considered valid.For the continuous parameters, age and risk score, we performed a linearity assumption test (Supplementary Fig. 2) and a multicollinearity test (Supplementary Fig. 3).When the fitted curve was approximately linear, the linearity assumption was considered valid.When the p-value of the correlation was less than 0.05, these 2 parameters were considered to have no multicollinearity.For each Cox regression result, we calculated Harrell's compliance index as the goodness-of-fit.
The sample size is labeled in the figure legends.The Kaplan-Meier technique (R package survival) was employed to create overall survival (OS) curves, and the log-rank test was used to assess alterations between the curves.This study used p < 0.05 as a statistically significant criterion.Statistical analyses employed R 3.6.2(R Foundation for Statistical Computing, Vienna, Austria) and IBM SPSS v. 26.0 for Windows (IBM Corp., Armonk, USA).
The Kaplan-Meier examination indicated that elevated expression of each ir-lncRNA was substantially related to poorer OS in the TCGA database, except for NDUFB2-AS1 and LINC00571 (Fig. 1C).

Progression and validation of the ten survival-related ir-lncRNA signatures for survival anticipation
Based on the multivariate Cox regression model, the above 10 ir-LncRNAs were integrated to create a risk score model in the TCGA database.In the TCGA, the Kaplan-Meier examination revealed that OS was considerably poorer in raised-risk participants than in decreasedrisk individuals (p < 0.0001, Fig. 2A).The risk scores and survival status of every GBM specimen were represented by the risk curve and scatterplot, respectively.The samples in the elevated-risk group had greater risk ratings and mortality rates than those in the reduced-risk cohort (Fig. 2C).
The ROC analysis revealed that the AUC for 1-, 2-and 3-year survival were 0.716, 0.803 and 0.852, respectively (Fig. 2E).Therefore, the risk score model constructed using the 10 ir-lncRNAs was effective in predicting GBM prognosis.Additionally, we developed and assessed the risk score model using the CGGA database as the validation set and acquired comparable outcomes (Fig. 2B,D,F).

Establishment and evaluation of nomograms by the risk scores and the prognostic value of medical variables
A univariate Cox regression examination was conducted to determine whether the risk score model for the 10 ir-lncRNAs and GBM-related clinical variables were prognostic factors.The results indicated that risk scores (p < 0.001; hazard ratio (HR) = 13.68) and patient age (p < 0.001; HR = 1.02) were strongly correlated with OS in the TCGA database (Fig. 3A).The CGGA database was also used as a validation set for univariate Cox regression analysis, and the ir-lncRNA model was found to be a substantial risk factor for GBM individuals (Fig. 3B).
The nomogram with age, gender and risk score was developed for the prediction of patient prognosis in the TCGA dataset.We collected survival information from all patients to anticipate the 1-, 2-and 3-year OS (Fig. 3C).The calibration curve of the 1-, 2-and 3-year OS showed that the nomogram had excellent prognostic value (Fig. 3E).The AUC of the 1-, 2-and 3-year OS were 0.79, 0.82 and 0.91, respectively, according to the ROC curve analysis (Fig. 3F).Similar prediction outcomes were achieved in the CGGA validation set, supporting the nomogram's effective prediction ability over the risk score model (Fig. 3D,G,H).

Enrichment analysis identified ir-lncRNA-related biological functions and signaling pathways in GBM
Based on the risk score, GBM participants in the TCGA were separated into elevated-and reduced-risk cohorts.Gene ontology and KEGG enrichment analysis were then performed on the genes that were differently expressed between the 2 groups.The outcomes of KEGG enrichment investigations revealed that the differentially expressed genes were primarily related to the activation of innate immune response, cellular responses to chemokines, collagen binding, cytokine secretion, humoral immune responses, the Janus kinase (JAK)-signal transducer and activator transcription (STAT) cascade, leukocyte migration, macrophage activation, modulation of lymphocyte activity, and T cell activation (Fig. 4A).According to the outcomes of the GO enrichment analysis, these genes primarily participated in the signaling pathways for cell adhesion molecules, chemokine signaling, cytokine-cytokine

The ir-lncRNAs-related risk score was correlated with the GBM immune microenvironment
To additionally explore the function of the 10 ir-ln-cRNAs in the immune microenvironment, the TIMER database was used to examine the connection between the risk score and various infiltrating immune cell subpopulations in GBM.It was shown that the risk score was closely related to the infiltration of CD8 + T cells, neutrophils, macrophages, and myeloid dendritic cells, with correlation coefficients of 0.24, 0.25, 0.27, and 0.24, respectively (Fig. 5A).It is well-known that chemokines efficiently control immune cell infiltration in malignancies.The expression of 16 immune-related chemokines recognized to be linked with GBM development was compared in elevated-and low-risk tissues to further examine the relationship between risk scores and chemokine secretion.The outcomes demonstrated that 14 chemokines were differentially expressed (chemokine ligand (CCL)2, CCL5, CCL17, CCL20, CCL22, CCR2, CCR4, CCR5, CCR6, CCR7, CXC motif chemokine ligand (CXCL)12, CXCL16, CXC motif chemokine receptor (CXCR)4, and CXCR6) (Fig. 5B, Table 1 and Supplementary Table 1).
Further calculations of the GBM patients' immunological, stromal and ESTIMATE scores in the TCGA dataset indicated that the immune score of the reduced-risk cohort was much less than that of the elevated-risk cohort (Fig. 5C-E, Table 2 and Supplementary Table 2).A strong association between the risk score and commonly employed immunological checkpoints was determined using Pearson's correlation analysis, indicating that the risk score may have implications for immunotherapy (Fig. 5F).To further anticipate the reaction of GBM to immunotherapy, we used the TIDE algorithm.A high TIDE score means that the likelihood of responding to immunotherapy is low.According to the findings, individuals with elevated-risk GBM had reduced TIDE scores and were more likely to react to immunotherapy (Fig. 5G,H, Table 2,3 and Supplementary Table 2).The previous findings demonstrated that the risk score associated with ir-lncRNAs was strongly connected to the GBM immune microenvironment, and it was anticipated that this relationship might be helpful in targeting future GBM immunotherapy.

Alterations in the mutation landscape between elevated-and low-risk groups
Tumor somatic mutation of commonly mutated genes was profiled in GBM elevated-risk and reduced-risk cohorts.Most mutated genes, including TP53, PTEN, TTN, EGFR, MUC16, SPTA1, NF1, RYR2, and FLG, had high rates of somatic mutations in both raised-and reducedrisk groups.In particular, somatic mutation rates for RB1, HYDIN, ATRX, and IDH1 were greater in the low-risk cohort (Fig. 6A,B).Notably, ATRX and IDH1 did not have somatic mutations in the raised-risk cohort in the TCGA dataset (Fig. 6C,D).In the CGGA dataset, the patterns of risk scores for IDH1 mutation and 1p/19q codeletion stats were identified.According to the findings, those with IDH1 mutations had considerably lower risk scores than individuals with IDH1 wild type (Fig. 6E, Table 4 and Supplementary Table 4).Participants who had a 1p/19q codeletion also had a lower risk score than patients who did not have a codeletion (Fig. 6F, Table 4 and Supplementary Table 5).

AL354993.1 was recognized as a potential marker for GBM
The expression of 10 ir-lncRNAs was compared between normal and tumor samples in the CGGA dataset.The LEF1-AS1, HOXC-AS2 and AL354993.1 were found to be significantly overexpressed in GBM (p < 0.0001), among which AL354993.1 has not been reported (Fig. 7A, Table 5 and Supplementary Table 6).The overexpression of AL354993.1 suggested a worse GBM prognosis (p = 0.029) (Fig. 7B).Furthermore, the expression of AL354993.1 was lower in GBM tissues with IDH1 but not with 1p19q mutations (Fig. 7C,D, Table 6 and Supplementary Table 7,8).To discover the biological function of AL354993.1,GBM subjects in the CGGA dataset were allocated into 2 cohorts with raised and decreased AL354993.1 expression, and a differential gene analysis combined with GSEA enrichment was conducted.The outcomes indicated that raised AL354993.1 expression might indicate the activation of the JAK-STAT pathway, the extracellular matrix (ECM) receptor interaction pathway, chemokine signaling mechanisms, and cytokine-cytokine receptor interaction mechanisms (Fig. 7E).Hence,  the relationship between AL354993.1 expression and chemokine secretion were further explored.The expression of 16 chemokines known to be associated with GBM progression was examined in reduced-and elevated-expression groups.The results showed that 13 chemokines were differentially expressed (CCL2, CCL5, CCL17, CCL20, CCR2, CCR4, CCR5, CCR6, CCR7, CXCL12, CXCL16, CXCR4, CXCR6) (Fig. 7F, Table 7 and Supplementary Table 9).These outcomes indicate that AL354993.1 is related to GBM progression and might be a potential GBM biomarker.

Discussion
Glioblastoma multiforme is one of the most common primary brain tumors, and survival rates are low despite extensive treatment options (surgical resection, radiotherapy and adjuvant chemotherapy). 20In recent years, many studies carried out on immunotherapy for GBM have failed to achieve ideal results due to the heterogeneity of GBM and the cancer immunosuppressive microenvironment. 21,22Numerous investigations have verified that lncRNAs have a function in the modulation of GBM.Several ir-lncRNAs have also been shown to be strongly linked to the modulation of the immunological milieu in GBM and the infiltration of immune cells. 7,16Therefore, we created a medical prediction model using the TCGA and CGGA datasets and evaluated the lncRNAs most relevant to GBM to further examine the function of lncRNAs in GBM prognosis.
In this investigation, we conducted a three-way Venn examination of all lncRNAs in the TCGA, CGGA and Im-mLnc datasets and found that 276 lncRNAs coexisted.The LASSO regression analysis further identified 10 key ir-lncRNAs, with the Kaplan-Meier analysis suggesting that 8 of the 10 were prognostic risk factors for GBM (CHRM3-AS2, LEF1-AS1, AL354993.1,AC034243.1,LINC00452, UNC5B-AS1, AC009093.1,and HOXC-AS2).Among these, LEF1-AS1 was found to be related to the malignant growth of GBM through increased EN2 by mir-543 via sponge absorption. 23Besides, CHRM3-AS2, an immuneassociated lncRNA, has been recognized as a prognostic risk factor for ovarian cancer and cholangiocarcinoma. 24,25he LINC00452 has been shown to promote ovarian cancer by inhibiting ubiquitin-mediated degradation through sponge absorption of mir-501-3p, thereby increasing ROCK1. 26Meanwhile, UNC5B-AS1 is involved in liver malignancy, prostate tumors and colorectal malignancy. 27,28y mixing with HOXC13, HOXC-AS2 regulates the growth, cell death and migration of non-small cell lung tumors. 29he roles of these ir-lncRNAs in GBM are worthy of further investigation.
The coefficient was determined by employing multivariate Cox analysis, and a risk score model was developed to categorize GBM individuals into high-and reducedrisk cohorts.We discovered that the elevated-risk cohort had a lower OS rate and that the risk score model was beneficial in anticipating GBM prognosis.We also developed a nomogram that included sex, age and risk score.The nomogram demonstrated excellent predictive performance when calibration and ROC curves were plotted.Since these ir-lncRNAs have independent prognostic value in GBM OS, the risk prediction model constructed by these ir-lncRNAs also showed strong prognostic value.Therefore, we believe that monitoring these ir-lncRNAs by liquid biopsy or tumor tissue collection for GBM risk and prognosis assessment is a promising strategy.
Functional enrichment analysis was conducted to investigate the probable function and role of the 10 lncRNAs in GBM formation.They were shown to be connected with immune-related activities, including chemokinerelated mechanisms, the JAK-STAT signaling pathway, cytokine-related mechanisms, and immune cell activation-related pathways, according to GO and KEGG enrichment assessment.Additionally, they were connected to the stimulation of immune-related mechanisms, such as the Toll-like receptor pathway, the TNF signaling pathway and NF-ĸB signaling.To further explore the potential impact of the 10 ir-lncRNAs on GBM immunity, we examined the relationship between risk score and immune cell infiltration in GBM tissues.The outcomes demonstrated that CD8 + T cells, neutrophils, macrophages and myeloid dendritic cells were substantially positively correlated with the risk score.The relationship between risk scores and chemokines was also investigated, with the raised-risk cohort related to a higher chemokine expression.
Immunological checkpoint blocking has become one of the frontiers of cancer immunotherapy, 30 and we discovered that the risk score was substantially associated with immunological checkpoint expression.The ESTI-MATE and TIDE scores also verified the relationship between risk score and immune microenvironment.Mutations in the IDH1 gene are commonly found in GBM, with the product of this mutated enzyme having a novel ability to catalyze the production of 2-hydroxyglutarate, while IDH1 mutations are shown to be associated with more favorable survival outcomes. 31In addition, 1p/19q codeletion suggested a better prognosis. 32We found higher rates of IDH1 mutation and 1p/19q codeletion in GBM individuals with reduced risk scores, while no mutations were found in the raised-risk cohort.These results may be the reason underlying the prognostic value of risk scores.Furthermore, investigation is still required to fully understand the precise pathway of the 10 ir-lncRNAs in GBM.
Among the 10 lncRNAs, we found that only 2 were significantly highly expressed in GBM compared with normal controls, namely LEF1-AS1 and AL354993.1, of which AL354993.1 has not been reported.We found that AL354993.1 had significant prognostic value in GBM and was also associated with the tumor immune microenvironment.In addition, AL354993.1 was closely related to IDH1 mutation and other clinicopathological factors.Gene enrichment results suggested that AL354993.1 may promote GBM progression through the JAK-STAT pathway, the ECM interaction pathway, chemokine pathways, and cytokine receptor binding pathways.

Limitations
There were some restrictions to this analysis since it was retrospective and dependent on open-access databases.Thus, more prospective clinical data are required for validation and future clinical applications.On the other hand, the effect of AL354993.1 was only analyzed using a database and should be verified by a series of wet experiments in the future.

Conclusions
Through the joint analysis of multiple public databases, we screened 10 key ir-lncRNAs in GBM and used them to construct risk-scoring models and promote the development of novel GBM biomarkers.In conclusion, the 10 ir-lncRNAs have the potential to predict the GBM prognosis and may play a vital role in the progression of GBM.

Fig. 1 .
Fig. 1.Identification of 10 survival-related immune long noncoding ribonucleic acids (lncRNAs) in glioblastoma.A. Venn diagram illustrates that 276 lncRNAs were shared among 3 datasets; B. Distribution of least absolute shrinkage and selection operator (LASSO) coefficients for 276 immune-related lncRNAs; C. Partial likelihood deviation of the LASSO coefficient distribution.Vertical dashed lines indicate lambda.minand lambda.lse;D. The Kaplan-Meier survival curves comparing overall (OS) survival between high-and low-expression groups of the selected immune-related lncRNAs in The Cancer Genome Atlas (TCGA) dataset CGGA -Chinese Glioma Genome Atlas.

Fig. 2 .
Fig. 2. Development and validation of 10 survivalrelated immune long noncoding ribonucleic acids (lncRNA) signatures for survival prediction.The Kaplan-Meier survival analysis of high-risk and low-risk patients divided by the medium value in The Cancer Genome Atlas (TCGA) (A) and Chinese Glioma Genome Atlas (CGGA) (B) datasets.The distribution of risk scores, survival time and status of patients in TCGA (C) and CGGA (D) datasets.Receiver operating characteristic (ROC) analysis of the prognostic signatures to predict 1-, 2-and 3-year overall survival (OS) in TCGA (E) and CGGA (F) datasets

Fig. 3 .
Fig.3.Establishment and evaluation of nomograms using risk scores and the prognostic value of clinical variables.A. Univariate Cox analysis showed that risk scores and age were significantly related to overall survival (OS) in The Cancer Genome Atlas (TCGA) dataset using Harrell's compliance index (C-index) as the goodness-of-fit.For the TCGA cohort, the C-index was 0.69 (95% confidence interval (95% CI): of 0.64-0.73),and for the Chinese Glioma Genome Atlas (CGGA cohort), the C-index was 0.6 (95% CI: 0.54-0.65);B. Univariate Cox analysis showed that risk scores and age were significantly related to OS in the CGGA dataset; C. Development of the nomograms for the prediction of patient prognosis in the TCGA dataset; D. Development of the nomograms for the prediction of patient prognosis in the CGGA dataset; E. The calibration curve for 1-, 2-and 3-year OS of the nomogram in the TCGA dataset; F. Receiver operating characteristic (ROC) curves displayed the area under the curve (AUC) for 1-, 2-and 3-year OS in the TCGA dataset; G.The calibration curve for 1-, 2-and 3-year OS of the nomogram in the CGGA dataset; H. ROC curves displayed the AUC of 1-, 2-and 3-year OS in the CGGA dataset

Fig. 4 .
Fig. 4. Gene set enrichment analysis between high-risk and low-risk groups in the Kyoto Encyclopedia of Genes and Genomes (KEGG) and gene ontology (GO) datasets

Fig. 5 .
Fig. 5.The immune risk score was correlated with the tumor immune microenvironment.A. The correlation between risk score and immune cell infiltration; B. The expression of chemokines between low-(n = 83) and high-risk groups (n = 83) (parametric test and the Mann-Whitney U test).The difference in immune score (parametric test) (C), stromal score (parametric test) (D), and Estimation of STromal and Immune Cells in Malignant Tumor tissues using Expression data (ESTIMATE) score (Mann-Whitney U test) (E) between low-(n = 83) and high-risk groups (n = 83); F. Assessment of associations between risk scores and commonly used immunological checkpoints; G. Tumor immune dysfunction and exclusion (TIDE) scores between high-(n = 83) and low-risk group (n = 83); H. Comparison of the risk scores between immunotherapy responders (n = 48) and non-responders (n = 118) predicted with TIDE (parametric test)

Fig. 6 .
Fig. 6.Differences in the mutation landscape between high-and low-risk groups.A. Oncoplots of tumor somatic mutations of frequently mutated genes in the low-risk group; B. Oncoplots of tumor somatic mutations of frequently mutated genes in the high-risk group.Mutation rates of IDH1 (C) and ATRX (D) were compared between low-and high-risk groups in The Cancer Genome Atlas (TCGA) dataset; E. The distribution of risk score between IDH1-wild type (wt) (n = 100) and IDH1-mutation (Mut) (n = 40) patients in the Chinese Glioma Genome Atlas (CGGA dataset) (Mann-Whitney U test); F. The distribution of risk score between 1p/19q-codeleted (n = 12) and 1p/19q-noncodeleted (n = 128) patients in the CGGA dataset (Mann-Whitney U test)

Table 1 .
The expression of chemokines between low-and high-risk groups

Table 2 .
The difference of immune score, stromal score, ESTIMATE score, and TIDE score

Table 3 .
Comparison of the risk scores between immunotherapy responders (n = 48) and non-responders (n = 118) predicted with TIDE

Table 5 .
The expression of immune-related lncRNAs between normal and tumor samples in the CGGA dataset lncRNAs -long noncoding ribonucleic acids; CGGA -Chinese Glioma Genome Atlas.Values in bold are statistically significant.

Table 6 .
The

Table 7 .
The expression of chemokines between low-and high-AL354993.1 expression groups