Access the full text.
Sign up today, get DeepDyve free for 14 days.
Hindawi Journal of Oncology Volume 2022, Article ID 1317990, 13 pages https://doi.org/10.1155/2022/1317990 Research Article Developing a 5-Gene Signature Related to Pyroptosis for Osteosarcoma Patients 1 1 1 1 1 1 1 Zhe Li, Chi Jin, Xinchang Lu, Yan Zhang, Yi Zhang, Jia Wen, Yongkui Liu, 2 1 Xiaoting Liu, and Jiazhen Li Orthopaedics, e First A liated Hospital of Zhengzhou University, Zhengzhou 410100, China Orthopaedics, e A liated Hospital of Hangzhou Normal University, Hangzhou 310015, China Correspondence should be addressed to Jiazhen Li; firstname.lastname@example.org Received 23 May 2022; Revised 27 June 2022; Accepted 4 July 2022; Published 5 August 2022 Academic Editor: Zhiqian Zhang Copyright © 2022 Zhe Li et al. is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Although the incidence of osteosarcoma (OS) is relatively low compared with other cancer types, the overall survival of metastatic OS was less than 30%. is study aimed to reveal the role of pyroptosis in osteosarcoma and develop a prognostic model related to pyroptosis. Weighted correlation network analysis (WGCNA) was applied to identify key gene modules related to pyroptosis. Univariate Cox regression analysis was used to screen prognostic genes related to pyroptosis. e least absolute shrinkage and selection operator (LASSO) and stepwise Akaike information criterion (stepAIC) were employed to optimize and construct a prognostic model. Five prognostic genes (COL13A1, TNFRSF1A, LILRA6, CTNNBIP1, and CD180) related to pyroptosis were identi•ed. According to the 5-gene signature, OS samples were divided into high- and low-PPRS groups with dišerential prognosis. Immune-related pathways were more activated in the low-PPRS group. e 5-gene signature was ešective and robust to predict OS prognosis. ese •ve prognostic genes were involved in OS development and may serve as new targets for de- veloping therapeutic drugs. Evidence supports that immune checkpoint inhibitors can 1. Introduction also be potential therapeutic drugs for osteosarcoma. Tawbi Sarcomas are a group of rare malignant tumors deriving et al. observed that one of 22 OS patients had an objective from mesenchymal cells, with an incidence of about 2–4 per response to an anti-PD-1 antibody . However, in a phase 2 100,000 . Osteosarcoma (OS) is one of the types of sar- trial of pembrolizumab for treating advanced osteosarcoma, comas with an incidence of about 3.4 per million worldwide, no signi•cant antitumor activity of pembrolizumab was which commonly occurs in children and adolescents . presented in 12 patients . On the one hand, more ešective Two peaks of its incidence rate are shown at the ages of immunotherapy is needed and on the other hand, reaching a 15–19 and 75–79 . Osteosarcoma mostly localizes in long more personalized therapy for patients is also important. at is, to identify patients who are more suitable or sen- bones, especially in arms, legs, knees, and shoulders. Al- though localized osteosarcoma patients have a relatively sitive to immunotherapy is bene•cial. high overall survival, reaching a 5-year survival of 70–75%, Currently, a number of prognostic signatures or mo- metastatic patients have no more than 30% largely due to the lecular subtypes have been developed for dišerent cancer resistance to chemotherapy or radiotherapy . types based on gene expression data. Pyroptosis, as an Immunotherapy is developed as a potential therapy for emerging hot spot in cancer, plays an important role in cancers and has achieved satisfactory e£ciency in some cancer cell proliferation and migration . A number of cancer types. Some immune checkpoint inhibitors such as studies reported that pyroptosis suppresses cancer cell antiprogrammed cell death protein 1 (PD-1) inhibitors have growth in most the cancers such as glioma, ovarian cancer, been approved by Food and Drug Administration (FDA) . gastric cancer, and colon cancer . However, the tumor- 2 Journal of Oncology promotive roleof pyroptosis is shown in cervical cancer and with conditions of height �0.25, deepSplit �2, and min- esophageal adenocarcinoma . +e role of pyroptosis in ModuleSize �30. +e correlation between each module and pyroptosis score was assessed. +e most signiﬁcant gene osteosarcoma remains unclearly. Bioinformatics analysis helps a lot in exploring the mechanism of cancer develop- module was chosen to be the key module and used to ment and the biomarkers for predicting cancer prognosis construct a prognostic model. [9–11]. Inthisstudy,werevealedtherelationbetweenpyroptosis 2.3. Constructing a Prognostic Model Based on the Key Gene and osteosarcoma survival. By using weighted correlation Module. Univariate Cox regression analysis was performed networkanalysis(WGCNA)andCoxregressionanalysis,we on genes within the purple gene module for screening identiﬁed key prognostic genes of osteosarcoma. A 5-gene prognosis-related genes in the TARGET-OS dataset signature related to pyroptosis was constructed, and its (P<0.05). +e least absolute shrinkage and selection op- prognostic value was veriﬁed in three independent datasets. erator(LASSO)Coxregressionanalysisinglmnet Rpackage In addition, we evaluated the relation between the signature wasusedtodecreasethenumberofprognosticgenesby and tumor microenvironment (TME). +e 5-gene signature constructing a penalty function-based model. +e coeﬃ- was demonstrated to have an ability to identify individuals cients of each gene were compressed with changing lambda who were more sensitive to immunotherapy. values. Coeﬃcients closed to zero with the increasing lambda values. 5-fold cross-validation was used to validate 2. Materials and Methods the model. Stepwise Akaike information criterion (stepAIC) in MASS R package  was further introduced to decrease 2.1. Data Information. +e ﬂow chart of this study is shown the number of prognostic genes. Finally, the prognostic in Supplementary Figure S1. +e TARGET-OS dataset model was established as follows: pyroptosis-related prog- containing RNA-seq data was downloaded from National nostic risk score (PPRS) �coeﬃcient 1 gene 1+ coeﬃcient Cancer Institute Genomic Data Commons Data Portal ∗ ∗ 2 gene 2+, . . . + coeﬃcient n gene n. (https://portal.gdc.cancer.gov/). GSE21257 and GSE39055 datasetscontainingexpressiondataofosteosarcomasamples were downloaded from Gene Expression Omnibus (GEO) 2.4. Validating the Prognostic Model. PPRS was calculated database (https://www.ncbi.nlm.nih.gov/geo/). For the for each sample in the TARGET-OS dataset. Samples were TARGET-OS dataset, samples without survival time and divided into high-PPRS and low-PPRS groups according to status were removed. Ensembl ID was converted to gene the optimal cut-oﬀ determined by survminer R package symbol by using hgu133plus2. db R package. +e median (http://www.sthda.com/english/rpkgs/survminer/).Receiver expressionvalueofonegenewasselectedwhenthegenehad operating characteristic (ROC) analysis in the timeROC R multiple gene symbols. For GSE cohorts, probes were package  was used to evaluate the eﬀectiveness of the matched to gene symbols by using hgu133plus2. db R prognostic model to predict 1-year, 3-year, and 5-year package. +e median expression value of one gene was overall survival. +e area under ROC curve (AUC) was selected whenthe gene matchedmultiple probes. Oneprobe calculated.Kaplan–Meiersurvivalanalysiswasperformedto was excluded when they had multiple genes. Finally, 86, 53, assess overall survival between high- and low-PPRS groups. and 37 osteosarcoma samples remained in TARGET-OS, By using the same methods, we validated the model in GSE21257, and GSE39055 datasets, respectively. GSE21257 and GSE39055 datasets. 2.2. Identifying Key Genes Related to Prognosis and Pyroptosis. 2.5. Gene Set Enrichment Analysis (GSEA). GSEA is a A gene set of “REACTOME_PYROPTOSIS” was down- popular methodology that allows calculation of the en- loaded from Molecular Signature Database (MSigDB, richment score based on the expression of a gene set . https://www.gsea-msigdb.org/gsea/msigdb/). Single sample SsGSEAis anextendedmethodologybasedon GSEA,which gene set enrichment analysis (ssGSEA) in GSVA R package enables the calculation of the enrichment score for each  was performed to calculate ssGSEA score of pyroptosis sample . We used ssGSEA to evaluate the enrichment of based on “REACTOME_PYROPTOSIS” gene set in the the pyroptosis pathway and hallmark pathways. Cluster- TARGET-OSdataset.Samplesweredivided intotwogroups Proﬁler R package was applied to annotate Kyoto Ency- with high and low scores of pyroptosis according to the clopediaofGenesandGenomes(KEGG)pathwaysandgene medianvalue.WGCNAwasappliedtoidentifykeygene ontology (GO) terms . +e top 10 signiﬁcantly enriched modules related to pyroptosis. Firstly, samples were clus- pathways and GO terms were visualized (P<0.05). tered to screen the coexpression network. To meet the standard of a scale-free network, a correlation coeﬃcient >0.85 was determined. +en, the expression matrix was transferred to the topology matrix. Based on the topological 2.6. Evaluation of Tumor Microenvironment. CIBERSORT overlap matrix, average-linkage clustering was used to (http://cibersort.stanford.edu/) was employed to assess the cluster genes with each gene module containing at least 30 estimated proportion of 22 immune cells in high- and low- genes. Next, eigengenes of each gene module were calcu- PPRS groups . Estimation of STromal and Immune cells lated, and modules were further clustered and combined in MAlignant Tumours using Expression data (ESTIMATE) Journal of Oncology 3 was applied to calculate the stromal score, immune score, pyroptosis (R �0.87, P � 4.7e − 151, Figure 2(g)). +erefore, and ESTIMATE score . Immune checkpoints obtained the purple module was considered as a pyroptosis-related moduleforthefollowinganalysis.FunctionalanalysisofKEGG from HisgAtlas database were analyzed . Tumor Im- mune Dysfunction and Exclusion (TIDE, http://tide.dfci. pathways and GO terms for genes within the purple module harvard.edu/) was implemented to predict the response of showed that some immune-related terms were signiﬁcantly high- and low-PPRS groups to immunotherapy . A enriched, such as neutrophil activation involved in immune higher TIDE score represents a higher immune escape from response, MHC protein complex, immunoglobulin binding, immunotherapy. and MHC protein complex binding (Supplementary Figure S2). 2.7. Statistical Analysis. All statistical analysis was con- 3.3. Constructing a Prognostic Model Based on Pyroptosis- ducted in R software (v4.1.1). Parameters of methodologies Related Genes. Next, we utilized univariate Cox regression notindicatingweredefault.+elog-ranktestwasconducted analysis and screened 187 prognostic genes within pyrop- in Kaplan–Meier survival analysis and univariate and tosis-related genes in the TARGET-OS dataset, with 10 multivariate Cox regression analysis. +e Wilcoxon test was positively (risk) correlating with prognosis and 177 nega- conductedtotestthediﬀerencebetweenthetwogroups.+e tively (protective) correlating with prognosis (P<0.05, Kruskal–Walls test was performed among four groups. P<0.05 was considered signiﬁcant (Ns, not signiﬁcant. Supplementary Figure S3A). To construct a model using the ∗ ∗∗ ∗∗∗ ∗∗∗∗ minimum prognostic genes, we introduced LASSO Cox P<0.05, P<0.01, P<0.001, and P<0.0001). regression and stepAIC to reduce the number of prognostic genes. In LASSO analysis, the coeﬃcients of prognostic 3. Results genes were close to zero with the increasing lambda value 3.1. Pyroptosis Is Associated with Overall Survival of OS. (Supplementary Figure S3B). 5-Fold cross-validation To understand the relation between pyroptosis and OS revealed the conﬁdence interval for each lambda value prognosis, we calculated the ssGSEA score of pyroptosis for (Supplementary Figure S3C). When lambda �0.1395, the each sample in the TARGET-OS dataset. We found that the model was optimal. Moreover, stepAIC was conducted to distribution of clinical features including age, gender, and further optimize the model. Finally, 5 prognostic genes metastasis was signiﬁcantly associated with the enrichment remained,withoneriskgene(COL13A1)andfourprotective score of pyroptosis (P<0.0001, Figure 1(a)). Samples were genes (TNFRSF1A, LILRA6, CTNNBIP1, and CD180) divided into two groups with high and low ssGSEA scores. (Supplementary Figure S3D). Kaplan–Meier survival analysis showed that two groups had For each sample in the TARGET-OS dataset, pyroptosis- diﬀerential overall survival (Figure 1(b)), suggesting that related prognosticrisk score(PPRS) wascalculated.According pyroptosis played an important role in tumor progression. to the optimal cut-oﬀ analyzed by survminer, samples were However, no signiﬁcant diﬀerence was observed between divided into high-PPRS and low-PPRS groups. We observed diﬀerent groups of clinical features including age, gender, that dead samples were signiﬁcantly enriched in high-PPRS metastasis, and survival status (Figure 1(c)). group compared to low-PPRS group (Figure 3(a)). +e ex- pression of COL13A1 was higher in high-PPRS group, while the other four genes were lower expressed (Figure 3(a)). ROC 3.2. Identifying Key Genes Related to Pyroptosis by WGCNA. analysis presented that the model was eﬀective to predict 1- We conﬁrmed that the pyroptosis score was signiﬁcantly as- year, 3-year, and 5-year overall survival, with a high AUC of sociated with OS prognosis. To identify key genes related to 0.87, 0.87, and 0.88 (Figure 3(b)). Kaplan–Meier survival pyroptosis, we applied WGCNA to cluster samples and screen analysis showed that two groups had diﬀerential overall sur- co-expressionmodules(Figure2).SamplesintheTARGET-OS vival in the TARGET-OS dataset (P<0.0001, Figure 3(c)). To dataset were clustered (Figure 2(a)). To reach a scale-free verify the robustness of the prognostic model, we examined it network, a negative correlation over 0.85 between log (k) and in another two independent datasets (GSE21257 and log (P(k)) was selected, where k indicated connectivity degree. GSE39055).Similarresultsweregeneratedthatsampleswereall +erefore, the power of soft threshold (β) �3 was conﬁrmed classiﬁed into two groups with distinct prognoses (P � 0.017 (Figures 2(b) and 2(c)). By using average linkage clustering and P � 0.00035, respectively, Figures 3(d)–3(g)). +e above based on the topological overlap matrix (TOM), samples were results demonstrated that the 5-gene prognostic model was further clustered. According to the dynamic cut method, valid to predict prognosis for OS patients, and pyroptosis- modules were primarily distributed with each module con- related genes played an important role in OS development. taining at least 30 genes (Figure 2(d)). +en, adjacent modules were merged based on the eigengenes of each module, and ﬁnally,41 modules remained (Figure 2(d)).+e gene counts of 3.4. e Relation between PPRS Score and Clinical Features. each module were visualized (Figure 2(e)). Furthermore, we WeveriﬁedthatPPRSscorewassigniﬁcantlyassociatedwith assessed the correlation between each module and pyroptosis. overallsurvivalinbothtestandvalidationdatasets.Toknow the relation between PPRS score and other clinical infor- As a result, we observed that the purple module was signiﬁ- cantly correlated with pyroptosis (R �0.62, P � 2.92e − 10, mation suchasages, genders,metastasis, andrecurrence,we compared their PPRS score between high- and low-PPRS Figure 2(f)). In addition, the expression of genes in the purple module was positively correlated with the enrichment of groups. No signiﬁcant diﬀerences in ages, genders, and 4 Journal of Oncology 1.00 0.75 0.50 −2 0.25 p = 0.025 Age p.value=2.61e-01 Gender p.value=6.34e-02 Metastatic p.value=2.05e-01 0.00 Age Gender Metastatic groups= 45 25 4 0 0 Pyroptosis−high Female No groups= Male Yes 41 17 7 1 1 Pyroptosis−low 0 4 8 12 16 Time Groups groups=Pyroptosis−high groups=Pyroptosis−low (a) (b) wilcox.tests p=0.3 wilcox.tests p=0.064 wilcox.tests p=0.21 wilcox.tests p=0.34 ns ns ns ns 2 2 2 2 1 1 1 1 0 0 0 0 −1 −1 −1 −1 −2 −2 −2 −2 <=14 >14 Female Male No Yes Alive Dead Age Gender Metastatic Status Group Group Group Group <=14 Female No Alive >14 Male Yes Dead (c) Figure 1: e relation between pyroptosis and clinical features of osteosarcoma in TARGET-OS dataset. (a) e distribution of dišerent clinical features ranked by the z-score of ssGSEA score of pyroptosis. (b) Kaplan–Meier survival analysis of high- and low-score groups according to the cut-oš of z-score 0. e log-rank test was conducted. (c) e distribution of pyroptosis scores in dišerent clinical features. e Wilcoxon test was conducted. ns, no signi•cance. grades were observed between the two groups in all three PPRS groups, cell cycle-related pathways such as Myc tar- datasets (Figures 4(a)–4(c)). Noteworthy, PPRS scores gets, E2F targets, and G2M checkpoint were more enriched varied signi•cantly between metastatic and nonmetastatic, in the high-PPRS group, while immune-related pathways alive and dead, recurrent and nonrecurrent samples such as interferon-c response, inªammatory response, and (P < 0.05). It could be speculated that pyroptosis-related IL6-JAK-STAT3 signaling pathway were more enriched in genes were involved in the cancer cell metastasis. the low-PPRS group (Figure 5(b)). e above results sug- gested that the high-PPRS group had higher activity in the cell cycle and may thus contribute to cancer cell invasion and 3.5. DiŠerentially Enriched Pathways between High- and Low- migration. e activation of immune-related pathways in PPRS Groups. To understand the enriched pathways of the low-PPRS group possibly served as protective factors for high- and low-PPRS groups, we calculated ssGSEA score of inhibiting cancer cell progression. hallmark pathways for each sample in TARGET-OS dataset. Pearson correlation analysis revealed that 29 pathways were signi•cantly correlated with PPRS score (|R| ≥ 0.4, 3.6. TME Features and Immunotherapy/Chemotherapy Re- Figure 5(a)). e majority of pathways were related to sponse of High- and Low-PPRS Groups. Next, we evaluated immunity such as primary immunode•ciency, complement whether there was a dišerence in TME features between the two groups. CIBERSORT analysis revealed a similar dis- and coagulation cascades, cytokine-cytokine receptor in- teraction, B cell receptor signaling pathway, T cell receptor tribution of 22 immune cells in two groups (Supplementary signaling pathway, and chemokine signaling pathway. In the Figure S4A). However, ESTIMATE analysis showed that the comparison of enriched pathways between high- and low- low-PPRS group had higher stromal and immune Pyroptosis (ssGSEA score) Pyroptosis Pyroptosis (ssGSEA score) Pyroptosis (ssGSEA score) Groups Survival probability Pyroptosis (ssGSEA score) Journal of Oncology 5 Sample clustering to detect outliers Scale independence 4 5 6 7 8 910 3 12 14 16 18 20 22 24 26 28 30 0.8 0.6 0.4 0.2 0.0 1 0 5 10 15 20 25 30 Soft Threshold (power) (a) (b) Mean connectivity Cluster Dendrogram 1.0 0.9 0.8 0.7 0.6 5 6 0 7 8 910 12 14 16 18 20 22 24 26 28 30 0.5 0 5 10 15 20 25 30 Soft Threshold (power) Dynamic Module Merged Module (c) (d) yellowgreen Clustering of module eigengenes yellow white violet steelblue 0.9 skyblue3 skyblue salmon saddlebrown royalblue 0.7 red purple plum1 pink paleturquoise 0.5 orangered4 orange midnightblue mediumpurple3 magenta 0.3 lightyellow lightsteelblue1 lightgreen lightcyan1 lightcyan ivory grey60 grey greenyellow green darkturquoise darkred darkorange darkolivegreen darkmagenta darkgrey darkgreen cyan brown 0.5 blue black 0 1000 2000 3000 4000 −0.5 Number of Genes −1 (e) (f) Figure 2: Continued. Height Mean Connectivity Height Height lightyellow −0.037 (7.33e−01) MElightyellow paleturquoise −0.019 (8.61e−01) MEpaleturquoise skyblue3 0.086 (4.34e−01) MEskyblue3 ivory 0.02 (8.57e−01) MEivory salmon 0.49 (1.31e−06) MEsalmon steelblue 0.32 (3.09e−03) MEsteelblue saddlebrown 0.11 (3.09e−01) MEsaddlebrown darkgreen 0.4 (1.17e−04) MEdarkgreen darkmagenta 0.48 (3.65e−06) MEdarkmagenta purple 0.62 (2.92e−10) MEpurple lightgreen 0.019 (8.59e−01) MElightgreen Scale Free Topology Model Fit,signed R^2 violet 0.12 (2.57e−01) MEviolet black −0.16 (1.43e−01) MEblack yellow 0.2 (6.52e−02) MEyellow lightsteelblue1 0.1 (3.52e−01) MElightsteelblue1 MEgrey grey 0.091 (4.06e−01) MEorange orange −0.15 (1.80e−01) MEdarkturquoise darkturquoise −0.18 (9.76e−02) MElightcyan lightcyan −0.22 (4.58e−02) MEwhite white 0.16 (1.47e−01) MEyellowgreen yellowgreen −0.12 (2.55e−01) MEdarkorange darkorange 0.027 (8.02e−01) MElightcyan1 lightcyan1 −0.034 (7.56e−01) MEroyalblue royalblue −0.21 (5.39e−02) MEplum1 plum1 −0.35 (1.08e−03) MEgreenyellow greenyellow −0.37 (4.48e−04) MEmidnightblue midnightblue −0.35 (8.95e−04) MEdarkred darkred −0.3 (5.09e−03) MEskyblue skyblue −0.25 (1.92e−02) MEcyan cyan 0.017 (8.75e−01) MEpink pink −0.31 (3.96e−03) MEmagenta magenta −0.21 (5.28e−02) MEblue blue −0.44 (1.94e−05) MEgreen green −0.36 (5.72e−04) MEorangered4 orangered4 −0.033 (7.63e−01) MEgrey60 grey60 −0.011 (9.20e−01) MEred red −0.21 (5.04e−02) MEbrown brown −0.2 (5.94e−02) MEmediumpurple3 mediumpurple3 −0.15 (1.56e−01) MEdarkgrey darkgrey −0.31 (3.89e−03) MEdarkolivegreen darkolivegreen −0.048 (6.58e−01) Pyroptosis 6 Journal of Oncology Module membership vs. gene signiﬁcance cor=0.87, p=4.7e−151 0.6 0.4 0.2 0.0 −0.2 −0.4 −0.2 0.0 0.2 0.4 0.6 0.8 Module Membership in purple module (g) Figure 2: WGCNA for identifying key gene modules related to pyroptosis. (a) Clustering for TARGET-OS samples. (b, c) Analysis of scale independence and mean connectivity for dišerent soft thresholds (power). (d) Cluster dendrogram based on topology and identi•cation of gene modules. (e) Gene numbers of each gene module. (f) Pearson correlation analysis between gene modules and pyroptosis. Correlation coe£cients were indicated by colors. Blue indicates negative correlation, and red indicates positive correlation. (g) Pearson correlation analysis between pyroptosis score (gene signi•cance) and gene expression in the purple module (module membership). −1 −2 RiskType High Low Status Alive Dead TNFRSF1A LILRA6 CTNNBIP1 COL13A1 CD180 Samples z−score of expression (a) Figure 3: Continued. Time RiskScore Gene signiﬁcance for Pyroptosis −2 −1 2 Journal of Oncology 7 1.00 1.00 0.75 0.75 0.50 0.50 0.25 p < 0.0001 0.25 0.00 HR=9.21 95CI%(4.33−19.61) groups=Low 62 39 11 1 1 0.00 groups=High 24 3000 048 12 16 0.00 0.25 0.50 0.75 1.00 Time False positive fraction Type Groups 1−Years,AUC=0.87,95%CI(0.75−0.98) groups=Low 3−Years,AUC=0.87,95%CI(0.76−0.98) groups=High 5−Years,AUC=0.88,95%CI(0.78−0.98) (b) (c) 1.00 1.00 0.75 0.75 0.50 0.50 0.25 p = 0.017 0.25 0.00 HR=1.64 95CI%(1.1−2.45) groups=Low 24 12 5 4 1 0.00 groups=High 29 8320 0.00 0.25 0.50 0.75 1.00 0 5 10 15 20 Time False positive fraction Type Groups 1−Years,AUC=0.83,95%CI groups=Low (0.69−0.97) groups=High 3−Years,AUC=0.69,95%CI (0.54−0.85) 5−Years,AUC=0.73,95%CI (0.58−0.89) (d) (e) 1.00 1.00 0.75 0.75 0.50 0.50 0.25 p = 0.00035 0.25 0.00 HR=3.71 95CI%(1.01−13.58) groups=Low 27 15 3 3 2 0.00 groups=High 10 2200 0 4 8 12 16 0.00 0.25 0.50 0.75 1.00 False positive fraction Time Type Groups 2−Years,AUC=0.87,95%CI groups=Low (0.71−1.02) groups=High 3−Years,AUC=0.78,95%CI (0.58−0.97) 5−Years,AUC=0.8,95%CI (0.61−0.98) (f) (g) Figure 3: Validating the performance of the 5-gene prognostic model. (a) Risk score, survival status, and z-score expression of prognostic genes of each sample in the TARGET-OS dataset. (b) ROC analysis of the model for predicting 1-year, 3-year, and 5-year overall survival in the TARGET-OS dataset. (c) Kaplan–Meier survival plot of high- and low-PPRS groups in TARGET-OS dataset. (d, e) ROC analysis and survival analysis in GSE21257 dataset. (f, g) ROC analysis and survival analysis in the GSE39055 dataset. e log-rank test was conducted in Kaplan–Meier survival analysis. in•ltration than the high-PPRS group in the TARGET-OS macrophages and resting dendritic cells were positively dataset (P < 0.01, Supplementary Figure S4B). In GSE21257 correlated with PPRS score. and GSE39055 datasets, we observed similar results (Sup- Of the immune checkpoints, we found that 6 of 21 were plementary Figure S5). In addition, we assessed the corre- dišerentially expressed between high- and low-PPRS lation between the PPRS score and the enrichment of 22 groups, including CD27, CD47, GEM, HAVCR2, LAG3, and immune cells. CD8 Tcells and activated memory CD4 Tcells TNFSF4 (P < 0.05, Supplementary Figure S6A). Further- were negatively correlated with PPRS score. M0 more, we assessed the enrichment of three True positive fraction True positive fraction True positive fraction Groups Survival probability Groups Groups Survival probability Survival probability 8 Journal of Oncology wilcox.tests p=0.33 wilcox.tests p=0.84 wilcox.tests p=0.013 wilcox.tests p=1.2e−07 wilcox.tests p=0.016 ns ns * **** * 2 2 2 2 2 1 1 1 1 1 0 0 0 0 0 −1 −1 −1 −1 −1 −2 −2 −2 −2 −2 <=14 >14 Female Male No Yes Alive Dead Pyroptosis Pyroptosis −high −low Age Gender Metastatic Status Cluster (a) wilcox.tests p=0.72 wilcox.tests p=0.37 Kruskal−Wallis test p=0.48 wilcox.tests p=0.0083 wilcox.tests p=0.034 ns ns ns 6 ** * 3 3 ns 3 3 ns ns 2 2 2 2 4 ns ns 1 1 1 1 0 0 0 0 −1 −1 −1 −1 −2 −2 −2 −2 −2 <=14 >14 Female Male 1234 No Yes Alive Dead Age Gender Huvos grade Metastatic status (b) wilcox.tests p=0.053 wilcox.tests p=0.5 wilcox.tests p=0.036 wilcox.tests p=0.052 ns ns ns 0 0 0 0 −2 −2 −2 −2 −4 −4 −4 −4 <=14 >14 Female Male No Yes Alive Dead Age Gender Recurrence Status (c) Figure 4: e relation between PPRS score and clinical features in TARGET-OS (a), GSE21257 (b), and GSE39055 (c) datasets. the ∗ ∗∗ ∗∗∗∗ Wilcoxon test was conducted. ns, no signi•cance. P < 0.05, P < 0.01, and P < 0.0001. KEGG_PATHOGENIC_ESCHERICHIA_COLI_INFECTION KEGG_JAK_STAT_SIGNALING_PATHWAY *** KEGG_FC_EPSILON_RI_SIGNALING_PATHWAY ** *** KEGG_PRIMARY_IMMUNODEFICIENCY *** *** *** KEGG_COMPLEMENT_AND_COAGULATION_CASCADES *** *** *** *** KEGG_NOD_LIKE_RECEPTOR_SIGNALING_PATHWAY *** *** *** *** *** KEGG_CELL_ADHESION_MOLECULES_CAMS *** *** *** *** *** *** KEGG_CYTOKINE_CYTOKINE_RECEPTOR_INTERACTION *** *** *** *** *** *** *** KEGG_LEUKOCYTE_TRANSENDOTHELIAL_MIGRATION *** *** *** *** *** *** *** *** KEGG_FC_GAMMA_R_MEDIATED_PHAGOCYTOSIS *** *** *** *** *** *** *** *** *** KEGG_B_CELL_RECEPTOR_SIGNALING_PATHWAY *** *** *** *** *** *** *** *** *** *** KEGG_ASTHMA *** *** *** *** *** *** *** *** *** *** *** KEGG_ANTIGEN_PROCESSING_AND_PRESENTATION *** *** *** *** *** *** *** *** *** *** *** *** KEGG_INTESTINAL_IMMUNE_NETWORK_FOR_IGA_PRODUCTION ** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_TYPE_I_DIABETES_MELLITUS *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_AUTOIMMUNE_THYROID_DISEASE *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_GRAFT_VERSUS_HOST_DISEASE *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_ALLOGRAFT_REJECTION *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_T_CELL_RECEPTOR_SIGNALING_PATHWAY *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_LEISHMANIA_INFECTION *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_TOLL_LIKE_RECEPTOR_SIGNALING_PATHWAY *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_HEMATOPOIETIC_CELL_LINEAGE *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_NATURAL_KILLER_CELL_MEDIATED_CYTOTOXICITY *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_CHEMOKINE_SIGNALING_PATHWAY *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_NEUROACTIVE_LIGAND_RECEPTOR_INTERACTION *** *** ** . *** *** ** *** *** ** ** ** ** ** ** ** ** * ** ** *** KEGG_ACUTE_MYELOID_LEUKEMIA *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** KEGG_NEUROTROPHIN_SIGNALING_PATHWAY *** *** ** * ** ** *** *** *** *** * ** ** *** *** *** *** PPRS *** *** ** *** *** *** ** *** *** *** *** ** ** *** ** ** ** ** *** *** *** *** *** *** *** *** ** KEGG_HEDGEHOG_SIGNALING_PATHWAY * ** * ** ** *** KEGG_STEROID_BIOSYNTHESIS * *** −log10(p value) correlation 5 5.2 5.5 5.8 6 −1 −0.5 0 0.5 1 (a) Figure 5: Continued. PPRS PPRS PPRS PPRS PPRS PPRS PPRS PPRS PPRS PPRS PPRS PPRS PPRS PPRS Journal of Oncology 9 TARGET-OS NES GSE21257 −2 −4 GSE39055 (b) Figure 5: KEGG pathway analysis of high- and low-PPRS groups in the TARGET-OS dataset. (a): Pearson correlation analysis between KEGG pathways and PPRS score. Pathways with |correlation coe£cient| ≥ 0.4 were visualized. Blue indicates negative correlation, and red indicates positive correlation. (b) Comparison of enriched pathways between high- and low-PPRS groups in three datasets (FDR <0.05). Yellow indicates higher enrichment in the high-PPRS group, and purple is the reverse. NES, normalized enrichment score. P < 0.05, ∗∗ ∗∗∗ P < 0.01, and P < 0.001. immunosuppressive cells in two groups. Myeloid-derived and 6(c)). Dead samples were the most distributed in the C4 suppressor cells (MDSCs) and tumor-associated macro- subgroup and alive samples composed the most in the C1 phages (TAMs) were more enriched in the high-PPRS subgroup, which was consistent with the survival analysis (Figures 6(b) and 6(d)). Univariate and multivariate Cox group, while cancer-associated •broblasts (CAFs) were more enriched in the low-PPRS group (Supplementary regression analysis showed that metastasis and PPRS score Figure S6B). TIDE analysis showed that the low-PPRS group were independent risk factors (Figures 6(e) and 6(f)). had more serious T cell exclusion and dysfunction than the According to PPRS score and metastasis, we established a high-PPRS group. A higher TIDE score was shown in the nomogram to predict 1-year, 3-year, and 5-year prognosis low-PPRS group, indicating a higher possibility to escape for osteosarcoma patients (Figure 6(g)). e predicted 1- from immune checkpoint blockade therapy, although there year, 3-year, and 5-year overall survival were corrected was no signi•cance between the two groups (Supplementary (Figure 6(h)). Compared with metastasis and PPRS score, Figure S6B). Moreover, in the response to chemotherapy, the the nomogram was optimal to assist decision-making and PPRS-high group was more sensitive to doxorubicin than create the maximum bene•t for patients (Figure 6(i)). ROC the PPRS-low group (Supplementary Figure S7). However, analysis showed that the PPRS score and the nomogram had the highest AUC (Figure 6(j)), which further proved the the estimated IC50 of the other three drugs (cisplatin, methotrexate, and paclitaxel) showed no signi•cant dišer- ešectiveness and practicability of the nomogram for its ence between the two groups. application in the clinic. 4. Discussion 3.7. Optimizing the Prognostic Model for Clinical Use. To is study demonstrated that pyroptosis was associated with make the prognostic model more accurate for clinical use, we constructed a decision tree based on ages, genders, the overall survival of osteosarcoma, suggesting that pyroptosis played an important role in OS progression. By metastasis, and the model. Finally, only metastasis and the model remained, and four subgroups were generated (C1 to using the WGCNA methodology, we identi•ed a gene module signi•cantly correlated with pyroptosis. Within the C4, Figure 6(a)). Four subgroups varied massively in overall survival, where C1 had the longest survival and C4 had the gene module, we screened 187 genes related to pyroptosis and OS prognosis. To construct a prognostic model, LASSO worst prognosis (Figure 6(b)). Low-PPRS samples were only displayed in C1 and C2 subgroups, and high-PPRS samples and stepAIC were applied to decrease the number of these were only included in C3 and C4 subgroups (Figures 6(a) genes. Finally, a 5-gene prognostic model consisting of MYC_TARGETS_V1 E2F_TARGETS G2M_CHECKPOINT MTORC1_SIGNALING UNFOLDED_PROTEIN_RESPONSE OXIDATIVE_PHOSPHORYLATION UV_RESPONSE_UP CHOLESTEROL_HOMEOSTASIS ESTROGEN_RESPONSE_LATE MYC_TARGETS_V2 ESTROGEN_RESPONSE_EARLY ADIPOGENESIS TGF_BETA_SIGNALING GLYCOLYSIS DNA_REPAIR FATTY_ACID_METABOLISM MYOGENESIS HEME_METABOLISM REACTIVE_OXYGEN_SPECIES_PATHWAY HYPOXIA WNT_BETA_CATENIN_SIGNALING SPERMATOGENESIS MITOTIC_SPINDLE ANGIOGENESIS UV_RESPONSE_DN PI3K_AKT_MTOR_SIGNALING HEDGEHOG_SIGNALING APICAL_JUNCTION APOPTOSIS EPITHELIAL_MESENCHYMAL_TRANSITION INTERFERON_ALPHA_RESPONSE P53_PATHWAY XENOBIOTIC_METABOLISM IL2_STAT5_SIGNALING APICAL_SURFACE KRAS_SIGNALING_UP TNFA_SIGNALING_VIA_NFKB COMPLEMENT COAGULATION IL6_JAK_STAT3_SIGNALING ALLOGRAFT_REJECTION INFLAMMATORY_RESPONSE INTERFERON_GAMMA_RESPONSE 10 Journal of Oncology 29 / 85 100% PPRS = PPRS−low 75 yes no 0.43 3.8 10 / 62 19 / 23 73% 27% 25 Metastatic = No Metastatic = No p < 0.0001 0 4 8 12 16 0.32 1 2.3 5.5 Time (months) 6 / 52 4 / 10 9 / 12 10 / 11 Number at risk 61% 12% 14% 13% C1 52 33 10 1 1 C1 C2 C3 C4 C2 10 6100 C3 12 3000 C4 11 0000 0 4 8 12 16 Time (months) PPRS C1 C3 C2 C4 (a) (b) −log10(anova p value) −log10(anova p value) C4 13.2(*) 4.47(*) NaNNA 0 C4 6.5(*) 1.35(*) 0.19 0 C3 13.49(*) 4.69(*) 0 NaNNA C3 4.77(*) 0.66 0 0.19 C2 NaNNA 0 4.69(*) 4.47(*) C2 1.12 0 0.66 1.35(*) C1 0 NaNNA 13.49(*) 13.2(*) C1 0 1.12 4.77(*) 6.5(*) 1.00 1.00 0.75 0.75 0.50 0.50 0.25 0.25 0.00 0.00 C1 C2 C3 C4 C1 C2 C3 C4 Group Group PPRS−low Alive PPRS−high Dead (c) (d) Features p.value Hazard Ratio(95%CI) Features p.value Hazard Ratio(95%CI) Age 0.77 0.988(0.91-1.07) Age 0.312 1.05(0.952-1.17) Gender 1.47(0.706-3.05) Gender 0.342 1.49(0.654-3.39) 0.304 Metastatic 4.77(2.29-9.95) Metastatic 0.00166 4.34(1.74-10.8) 3.16E-05 PPRS 8.46E-09 9.21(4.33-19.6) PPRS 1.46E-07 7.02(3.39-14.5) Hazard Ratio(95%CI) Hazard Ratio(95%CI) (e) (f) Figure 6: Continued. Survival probability (%) PPRS 14 Journal of Oncology 11 Nomogram Points PPRS*** −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 No Metastatic*** Yes Total points 0.0 0.2 0.4 0.6 0.8 1.0 40 60 80 100 120 140 160 180 Nomogram−prediced OS (%) Pr( time < 5 year ) 1−year 0.006 0.015 0.04 0.1 0.3 0.7 0.94 0.998 3−year 5−year Pr( time < 3 year ) 0.003 0.006 0.015 0.04 0.1 0.3 0.7 0.94 0.998 Pr( time < 1 year ) 0.004 0.01 0.025 0.06 0.2 0.5 0.9 (g) (h) 1.0 1.0 0.8 0.9 0.6 0.8 0.4 0.2 0.7 0.0 0.6 0.0 0.2 0.4 0.6 0.8 1.0 High Risk reshold 0.5 1:100 1:4 2:3 3:2 4:1 100:1 Cost:Beneﬁt Ratio time t Nomogram All Metastatic None Nomogram Metastatic PPRS Age PPRS Gender (i) (j) Figure 6: Application of the 5-gene prognostic model in the TARGET-OS dataset. (a) A decision tree based on the prognostic model and metastasis. (b) Kaplan–Meier survival curve of C1 to C4 subgroups. (c, d) e distribution in high- and low-PPRS groups and alive and dead groups in C1 to C4 subgroups. ANOVA was conducted. (e, f) Univariate and multivariate Cox regression analyses on ages, genders, metastasis, and PPRS score. (g) A nomogram based on PPRS score and metastasis. (h) Correction for predicted 1-year, 3-year, and 5-year overall survival based on the observed survival. (i) Decision curve analysis (DCA) of the nomogram, metastasis, and PPRS score. (j): AUC values of the nomogram, ages, genders, metastasis, and PPRS score. e log-rank test was conducted in Kaplan–Meier survival analysis (b) and univariate and multivariate Cox regression analysis (e, f). ANOVA test was conducted in (c, d). P < 0.05. COL13A1, TNFRSF1A, LILRA6, CTNNBIP1, and CD180 was Except for COL13A1 more expressed in the high-PPRS established for predicting the prognosis of osteosarcoma. group, other four prognostic genes were all expressed low in e 5-gene signature could divide OS samples into high- and the high-PPRS group. For osteosarcoma, COL13A1 was a low-PPRS groups with distinct overall survival in three risk factor and the other four genes were protective factors. independent datasets. In research of identifying survival-related genes in osteo- Pyroptosis has been reported to occur with a strong sarcoma, COL13A1 and CTNNBIP1 were also included as inªammatory response, which is activated by inªamma- prognostic biomarkers . COL13A1 was upregulated and somes such as NLR family pyrin domain containing 1 CTNNBIP1 was downregulated in dead OS patients, which (NLRP1) , NLRP3 , and NOD-like receptor con- was consistent with our result that COL13A1 overexpression taining a caspase activating and recruitment domain 4 and CTNNBIP1 downregulation were associated with un- favorable prognosis. (NLRC4) . By comparing the high-PPRS group to low- PPRS group, we observed dišerential enrichment of KEGG In bladder cancer, Miyake et al. found that a high uterine pathways between them. Noteworthy, a number of immune- level of COL13A1 was associated with a poor prognosis . related pathways were identi•ed to be negatively associated ey discovered that COL4A1+COL13A1 was an indepen- with PPRS, such as T cell receptor signaling, Toll-like sig- dent predictor for intravesical recurrence of bladder cancer naling, JAK-STAT signaling, NOD-like receptor signaling, . High expression of COL13A1 was observed in breast chemokine signaling, and cytokine-cytokine receptor sig- cancer cells, correlated with invasive tumor growth, and naling pathways. Our results further demonstrated that the induced anoikis resistance . CTNNBIP1 was reported as •ve pyroptosis-related genes were possibly involved in a suppressor in lung cancer that high expression of immune-related pathways and the modulation of immunity. CTNNBIP1 could inhibit the progression of lung cancer Standardized Net Beneﬁt AUC (t) 12 Journal of Oncology .Lowexpressionof CTNNBIP1wasariskfactorforlung model. (A) Identifying genes in the purple module was cancer (hazard ratio � 1.85) , which is accordant with signiﬁcantly associated with prognosis by univariate Cox our observation that the low-PPRS group had lower ex- regression analysis. Log-rank test was conducted. (B–C) pression of CTNNBIP1. CTNNBIP1 downregulation was LASSOCoxregressionanalysisfordecreasingthenumberof also discovered in human gastric adenocarcinoma tissues genes.+edottedredlineindicatestheoptimallambdavalue . of 0.1395. (D) +e LASSO coeﬃcients of ﬁve prognostic TNFRSF1A encodes a transmembrane receptor for tu- genes. Supplementary Figure S3. Comparison of TME be- mornecrosisfactor(TNF)-α.Highexpressionof TNFRSF1A tween high- and low-PPRS groups in TARGET-OS dataset. wasdemonstratedtobeassociatedwithSTAT3activationin (A) +e proportion of 22 immune cells in two groups. breastcancercells,whereSTAT3isknownasacriticalfactor Student’s t-test was conducted. (B) Comparison of the in tumorigenesis . CD180 was identiﬁed as a pharma- stromal score, immune score, and ESTIMATE score be- codynamic biomarker for tumors especially in lymphocytic tween high- and low-PPRS groups. Student’s t-test was leukemia . LILRA6 has not been reported to be signif- conducted. (C) Pearson correlation analysis between PPRS icantlyassociatedwithcancerdevelopment,servingasanew scoreandenrichmentofimmunecells.Blueandredindicate potential biomarker for predicting OS prognosis. negative and positive correlations, respectively. ns, not ∗ ∗∗ ∗∗∗ Besides diﬀerentially enriched pathways, the high-PPRS signiﬁcant. P<0.05, P<0.01, and P<0.001. Supple- and low-PPRS groups also had a diﬀerence in immune mentary Figure S4. Comparison of TME in GSE21257 (A-B) inﬁltration. Higher immune inﬁltration was shown in the and GSE39055 (C-D) datasets. ns, not signiﬁcant. P<0.05, ∗∗ ∗∗∗ low-PPRS group due to a more activated immune response P<0.01, and P<0.001. Supplementary Figure S5. (A) in the low-PPRS group. In addition, we constructed a de- Expression of immune checkpoints in high- and low-PPRS cision tree based on PPRS and metastasis. Four subgroups groups. (B) Enrichment of immunosuppressive cells (C1–C4) were classiﬁed by the decision tree with diﬀerential (MDSC, CAF, and M2 TAM), T cell exclusion, T cell dys- prognoses. For the application of the 5-gene signature, we function, and TIDE score in high- and low-PPRS groups. established a nomogram presenting superior performance Supplementary Figure S6. (A) Expression of immune than PPRS only for predicting OS prognosis. checkpoints in high- and low-PPRS groups. (B) Enrichment In conclusion, this study identiﬁed ﬁve prognostic genes of immunosuppressive cells (MDSC, CAF, and M2 TAM), related to pyroptosis and constructed a 5-gene signature Tcell exclusion, Tcell dysfunction, and TIDE score in high- with robust performance in three independent datasets. We and low-PPRS groups. Supplementary Figure S7. +e esti- further demonstrated the important role of pyroptosis in matedIC50offourchemotherapeuticdrugsinTARGET-OS osteosarcoma development and the relation between (A), GSE21257 (B), and GSE39055 (C) datasets. (Supple- pyroptosis and immunity. +e ﬁve prognostic pyroptosis- mentary Materials) associatedgenesmayplayanimportantroleinosteosarcoma pyroptosis. +e 5-gene signature could serve as a new tool References for predicting OS prognosis. In addition, the ﬁve prognostic genes may be potential targets for exploring new molecular  B. Pingping, Z. Yuhong, L. Weiqi et al., “Incidence and therapies for OS patients. mortality of sarcomas in shanghai, China, during 2002-2014,” Frontiers in Oncology, vol. 9, p. 662, 2019.  L. Mirabello, R. J. Troisi, and S. A. Savage, “International Data Availability osteosarcoma incidence patterns in children and adolescents, middle ages and elderly persons,” International Journal of +e datasets generated and/or analyzed during the current Cancer, vol. 125, no. 1, pp. 229–234, 2009. studyareavailableintheGSE21257repository(https://www.  S. A. Savage and L. Mirabello, “Using epidemiology and ncbi.nlm.nih.gov/geo/query/acc.cgi?acc�GSE), in GSE39055 genomics to understand osteosarcoma etiology,” Sarcoma, repository (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? vol. 2011, Article ID 548151, 13 pages, 2011. acc�), and in TARGETrepository (https://portal.gdc.cancer.  J. Zhao, D. C. Dean, F. J. Hornicek, X. Yu, and Z. Duan, gov/projects/TARGET-OS). “Emerging next-generation sequencing-based discoveries for targeted osteosarcoma therapy,” Cancer Letters, vol. 474, Conflicts of Interest pp. 158–167, 2020.  J.D.TwomeyandB.Zhang,“Cancerimmunotherapyupdate: +e authors declare that they have no conﬂicts of interest. FDA-approved checkpoint inhibitors and companion diag- nostics,” e American Association of Pharmaceutical Scien- Supplementary Materials tists Journal, vol. 23, no. 2, p. 39, 2021.  H. A. Tawbi, M. Burgess, V. Bolejack et al., “Pembrolizumab in advanced soft-tissue sarcoma and bone sarcoma Supplementary Figure S1. KEGG and GO function analysis (SARC028): a multicentre, two-cohort, single-arm, open-la- on genes within the purple module. (A) +e top 10 enriched bel, phase 2 trial,” e Lancet Oncology, vol. 18, no. 11, 2017. KEGG pathways. (B–D) +e top 10 enriched GO terms of  K. Boye, A. Longhi, T. Guren et al., “Pembrolizumab in molecular function, cellular component, and biological advanced osteosarcoma: results of a single-arm, open-label, process. Supplementary Figure S2. Identifying prognostic phase 2 trial,” Cancer Immunology, Immunotherapy, vol. 70, genes related to pyroptosis and constructing a prognostic no. 9, 2021. Journal of Oncology 13  Y. Fang, S. Tian, Y. Pan et al., “Pyroptosis: a new Frontier in osteosarcoma database,” Experimental Biology and Medicine, cancer,” Biomedicine and Pharmacotherapy, vol. 121, 2020. vol. 246, no. 19, 2021.  Z. Chen, M. Wang, R. L. De Wilde et al., “A machine learning  M. Miyake, Y. Morizawa, S. Hori et al., “Diagnostic and model to predict the triple negative breast cancer immune prognostic role of urinary collagens in primary human bladder cancer,” Cancer Science, vol. 108, no. 11, 2017. subtype,” Frontiers in Immunology, vol. 12, 2021.  H. Zhang, T. Fredericks, G. Xiong et al., “Membrane asso-  D. J. Hu, W. J. Shi, M. Yu, and L. Zhang, “High WDR34 ciatedcollagenXIIIpromotescancermetastasisandenhances mRNA expression as a potential prognostic biomarker in anoikis resistance,” Breast Cancer Research, vol. 20, no. 1, patients with breast cancer as determined by integrated p. 116, 2018. bioinformaticsanalysis,” Oncology Letters,vol.18,no.3,2019.  J.M.Chang,A.C.D.Tsai,W.R.Huang,andR.C.Tseng,“+e  W.Shi,D.Hu,S.Lin,andR.Zhuo,“Five-mRNAsignaturefor alterationof CTNNBIP1inlungcancer,” International Journal the prognosis of breast cancer based on the ceRNA network,” of Molecular Sciences, vol. 20, no. 22, 2019. BioMed Research International, vol. 2020, Article ID 9081852,  M. Kosari-Monfared, N. Nikbakhsh, S. Fattahi et al., 17 pages, 2020. “CTNNBIP1 downregulation is associated with tumor grade  S. Hanzelmann, ¨ R. Castelo, and J. Guinney, “GSVA: gene set and viral infections in gastric adenocarcinoma,” Journal of variation analysis for microarray and RNA-seq data,” BMC Cellular Physiology, vol. 234, no. 3, 2019. Bioinformatics, vol. 14, no. 1, p. 7, 2013.  S. P. Egusquiaguirre, J. E. Yeh, S. R. Walker, S. Liu, and  P. Langfelder and S. Horvath, “WGCNA: an R package for D. A. Frank, “+e STAT3 target gene TNFRSF1A modulates weighted correlation network analysis,” BMC Bioinformatics, the NF-κB pathway in breast cancer cells,” Neoplasia, vol. 20, vol. 9, no. 1, p. 559, 2008. no. 5, pp. 489–498, 2018.  J. Friedman, T. Hastie, and R. Tibshirani, “Regularization  T. C. Yeh, G. O’Connor, P. Petteruti et al., “Identiﬁcation of paths for generalized linear models via coordinate descent,” CCR2 and CD180 as robust pharmacodynamic tumor and Journal of Statistical Software, vol. 33, no. 1, pp. 1–22, 2010. blood biomarkers for clinical use with BRD4/BETinhibitors,”  Z. Zhang, “Variable selection with stepwise and best subset Clinical Cancer Research, vol. 23, no. 4, 2017. approaches,” Annals of Translational Medicine, vol. 4, no. 7, p. 136, 2016.  P. Blanche, J. F. Dartigues, and H. Jacqmin-Gadda, “Esti- mating and comparing time-dependent areas under receiver operating characteristic curves for censored event times with competing risks,” Statistics in Medicine, vol. 32, no. 30, 2013.  A. Subramanian, P. Tamayo, V. K. Mootha et al., “Gene set enrichment analysis: a knowledge-based approach for inter- preting genome-wide expression proﬁles,” Proceedings of the National Academy of Sciences, vol. 102, no. 43, 2005.  D. A. Barbie, P. Tamayo, J. S. Boehm et al., “Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1,” Nature, vol. 462, pp. 108–112, 2009.  G. Yu, L. G. Wang, Y. Han, and Q. Y. He, “ClusterProﬁler: an R package for comparing biological themes among gene clusters,” OMICS: A Journal of Integrative Biology, vol. 16, no. 5, pp. 284–287, 2012.  B. Chen, M. S. Khodadoust, C. L. Liu, A. M. Newman, and A. A. Alizadeh, “Proﬁling tumor inﬁltrating immune cells with CIBERSORT,” Methods in Molecular Biology, vol. 1711, pp. 243–259, 2018.  K.Yoshihara,M.Shahmoradgoli,E.Mart´ınezetal.,“Inferring tumour purity and stromal and immune cell admixture from expression data,” Nature Communications, vol. 4, no.1, 2013.  Y. Liu, M. He, D. Wang et al., “HisgAtlas 1.0: A human immunosuppression gene database,” 2017, https://www.ncbi. nlm.nih.gov/pmc/articles/PMC7243927/.  P. Jiang, S. Gu, D. Pan et al., “Signatures of Tcell dysfunction and exclusion predict cancer immunotherapy response,” Nature Medicine, vol. 24, no. 10, 2018.  F. Martinon, K. Burns, and J. Tschopp, “+e inﬂammasome,” Molecular Cell, vol. 10, no. 2, pp. 417–426, 2002.  E. Latz, T. S. Xiao, and A. Stutz, “Activation and regulation of the inﬂammasomes,” Nature Reviews Immunology, vol. 13, no. 6, pp. 397–411, 2013.  S. Mariathasan, K. Newton, D. M. Monack et al., “Diﬀerential activation of the inﬂammasome by caspase-1 adaptors ASC and Ipaf,” Nature, vol. 430, pp. 213–218, 2004.  E.Rothzerg,J.Xu,D.Wood,andS.Kõks,“12Survival-related diﬀerentially expressed genes based on the TARGET-
Journal of Oncology – Hindawi Publishing Corporation
Published: Aug 5, 2022
Access the full text.
Sign up today, get DeepDyve free for 14 days.