The Crosstalk Between CTNNB1 Mutation and M2 Macrophages Contribute to Hepatocellular Carcinoma Suppressive Immune Microenvironment

DOI: https://doi.org/10.21203/rs.3.rs-2176946/v1

Abstract

Objective:Mutations in the CTNNB1 gene was the second most common mutation after TP53 in HCC. However, the CTNNB1 mutation and tumor immune microenvironment of HCC have not been clarified.

Materials and Methods: We compared the CTNNB1 mutation frequency and hotspot site in China Pan-cancer (OrigiMed2020) and TCGA PanCancer Atlas cohort via cBioPortal database. The differentially expressed genes and corresponding function enrichment analysis between CTNNB1 mutation and non-mutation was detected by DESeq2 and MetaScape database, respectively. We also analyzed the association between CTNNB1 mutation status and drug sensitivity based on the RNAactDrug and DREIMT database.

Furthermore, we explored the genetic alteration score, infiltration of immune cell, and response to immune checkpoint inhibitor therapy under CTNNB1 mutation status by means of IPS and TIDE methods. Besides, gene module associated with CTNNB1 mutation and M2 immune cell were identified by weighted gene co-expression network analysis (WGCNA). Besides, we integrated differently expressed genes and gene modules associated crosstalk CTNNB1 mutation and M2 immune cell to seek targeted genes for CTNNB1-mutated HCC.

Results:There are obvious differences in CTNNB1 mutation frequency and mutation hotspots between European-American and Chinese patients with HCC. CTNNB1 mutation significantly altered Wnt signaling pathway score and he sensitivity to drugs, such as Nutlin-3 and PHA-665752. High TMB, microsatellite instability, neoantigen loads, intratumor heterogeneity score, number of segments, and homologous recombination defects score were significantly increased in CTNNB1 mutations group. Besides, Cibersort, EPIC, quantiseq, and xcell immune method suggested M2-type macrophages are significantly enriched in CTNNB1-mutated HCC. Interestingly, CTNNB1-mutated HCC showed a low level in immune checkpoint signature score.

11 gene modules were identified by WGCNA. Of them, we focused on MEmagenta (Gene modules positively correlated to CTNNB1 mutation and M2 macrophage) and MEbrown gene module (Gene modules negatively correlated to CTNNB1 mutation and M2 macrophage). Targeting pathways such as Wnt signaling and leukocyte activation were promising therapeutic strategy for CTNNB1-mutant HCC.

Conclusion:CTNNB1 plays an important role in the initiation and progression of HCC. Our results may provide novel insights for the selection of immunotherapeutic targets and prognostic biomarkers for CTNNB1-mutant HCC.

Introduction

When the incidence of other cancer types has trended downward globally, the global incidence of primary hepatocellular carcinoma[1] (HCC) has progressively increased over the last decades. It was reported that 70–80% of HCC patients are already in the middle and late stages at the time of diagnosis[2, 3]. Since then, some clinical studies of chemical drugs have been carried out, but there has been no breakthrough in OS benefit[4]. Recently, many researchers have explored the therapeutic targets of HCC, especially kinase and immune checkpoint inhibitors, and some progress has been made; however, this is far from sufficient, and more therapeutic targets and prognostic biomarkers must be identified.

CTNNB1 is one of the most significantly mutated genes in HCC (as high as 25%)[5, 6], encoding β-catenin protein, involved in the regulation of Wnt signaling pathway, which involved in occurrence and development of HCC. Hepatocellular adenoma (HCA) with CTNNB1[7] exon 3 mutations are at risk of malignant transformation in HCC and mutations in the TERT promoter is the second hit leading to HCC occurrence. Fan J et al., [8]suggested that the protein and phosphorylation differences between CTNNB1 mutant and wild-type HCC mainly clustered in metabolic pathways, especially, phosphorylation of fructose bisphosphate aldolase A (ALDOA) Besides, Llovet JM et al[9]., suggested that Wnt/CTNNB1-mutated HCC is an immune-privileged tumor, which is often referred to as “cold” tumor, and HCC patients with Wnt/CTNNB1 gene mutation are more sensitive to PD-1/PD -L1 monoclonal antibody is inherently resistant. Correspondingly, our study shows that HCC carrying CTNNB1 mutations was more prone to M2-type macrophage infiltration, resulting in an immunosuppressive microenvironment.

What motivates us to describe immune infiltration based on gene expression profiles? Coincidentally, it is thanks to bioinformatics algorithms such as Cibersort[10] and xcell[11].

In conclusion, our findings may provide novel insights for the selection of immunotherapeutic targets and prognostic biomarkers for CTNNB1-mutant HCC.

Materials And Methods

Study Cohort

The Cancer Genome Atlas (TCGA) GDC TCGA Liver Cancer (LIHC) gene expression profile “HTSeq – Counts”, “HTSeq – FPKM” format data and mutation dataset were download from UCSC Xena browser[12] (Supplementary 1) (https://xenabrowser.net/). Subsequently, the count data of RNA sequencing was transformed into transcript per million (TPM) using a custom R script. Besides, the probe ids was re-annotated to the gene symbols according to “gencode.v22.annotation.gene.probeMap”(https://gdc-hub.s3.us-east-1.amazonaws.com/download/gencode.v22.annotation.gene.probeMap). Of them, HTseq_counts, HTseq-FPKM and TPM format data were used to perform differential expression, gene set variation analysis (GSVA), weighted co-expression network (WGCNA). Besides, normal samples labeled with 11A and cancer samples labeled with 01A was included in our study.

Extracting CTNNB1 mutation samples and analyzing genetic alteration landscape

According to the studies published[13], we extracting CTNNB1 mutation and nonmutation samples. The R package maftools [14]was used to import the maf files of somatic mutations, perform mutual exclusivity and co-occurrence and draw the somatic mutation spectrum. We compared the CTNNB1 mutation frequency and hotspot site in China Pan-cancer (OrigiMed2020) and TCGA PanCancer Atlas cohort[15] with the help of cBioPortal database[16, 17]. The association between CTNNB1 mutation status and survival event (DFI, DSS, OS, PFS), which was performed by the cBioportal database.

Differentially expressed genes (DEGs) analysis

DESeq2 package[18] was utilized to identify differentially expressed genes for HCC vs. para-cancer tissues, HCC vs. paired para-cancer tissue, CTNNB1 mutation vs. non-mutation. P value < 0.05 and |log2FC|≥1 was considered to have a significant difference.

GO, KEGG enrichment and Gene set enrichment analysis (GSEA)

Gene Ontology (GO) annotation and pathway enrichment analyses of DEGs were performed using Metascape [19](http://metascape.org). Gene Set Enrichment Analysis (GSEA) [20]was performed with standard parameters (1000 permutations for gene sets, Signal2Noise metric for ranking genes) and significantly enriched pathways (H: hallmark gene sets) were illustrated. The intersection of differential genes was visualized by Venn diagrams[21].Besides, we performed gene set variation analysis (GSVA) to subtle 50 hallmark pathway activity changes over the sample population and to estimate variation of gene set enrichment across each dataset, using the “GSVA”R[22] package.

Exploring the association between drug sensitivity and CTNNB1 mutation

DREIMT[23](http://www.dreimt.org) and RNAactDrug[24] (http://bio-bigdata.hrbmu.edu.cn/RNAactDrug) database were used to exploring the association between drug sensitivity and CTNNB1 mutation. The input file of DREIMT database and RNAactDrug was a list of top 200 differentially expressed genes and single genes, respectively.

The Genetic score (tumor immunogenicity related genetic score) and CTNNB1 mutation

We compared some potential factors determining tumor immunogenicity between CTNNB1 mutation or not: Tumor Mutation Burden (TMB), MSI status (MANTIS score), neoantigen loads, intra-tumoral heterogeneity, Number of Segments, Fraction Altered, Aneuploidy Score, Homologous Recombination Defects by means of CAMOIP[25]. Differences between groups were detected by Wilcoxon rank-sum methods.

The immune cell landscape and CTNNB1 mutation

We used the CIBERSORT[10] deconvolution algorithm (https://cibersort.stanford.edu/) to estimate the abundance of 22 immune cell types in each cytolytic subgroup's tissue and to evaluate the corresponding intratumoral immune cell composition. Also, we used the QUANTISEQ[26], XCELL[11],and Estimating the Proportion of Immune and Cancer cells (EPIC)[27] to comparatively assess cell immune responses or cellular components between CTNNB1 mutation or not.

Evaluation of the immunotherapy efficacy and CTNNB1 mutation

We used two approaches to evaluation CTNNB1 mutation responsiveness to immunotherapy Immunophenoscore (IPS) algorithm[28], as a scoring tool for evaluating the tumor immunogenicity, was applied to analyze the correlation between the new immune signature and intratumor immune response. The TIDE[29] (http://tide.dfci.harvard.edu/) algorithm is used to model tumor immune evasion, which gathers two primary mechanisms of immune evasions: T cell dysfunction and T cell exclusion. Besides, it was reported [30]that the expression levels of immune checkpoint genes such as PD-L1 can be used as an independent predictor of ICI response.

Gene Network Construction and Visualization

To explore which genes are involved in the association between CTNNB1 mutation and M2 cell, weighted gene co-expression network analysis (WGCNA) was constructed using the WGCNA [31] in R. WGCNA data preprocessing: the top 8,000 genes after the mean absolute deviation (MAD) sorting and import to co-expression network, the integrity of the data was checked with the “goodSampleGenes” function. Soft Threshold selection: the “pickSoftThreshold” function was used to choose the satisfactory soft threshold power beta. Network construction: we used one-step network construction method to identify co-expression modules through the “blockwiseModules” function. WGCNA visualization: Plot of the soft threshold with “plot” function; the “plotDendroAndColors” function was applied to draw a dendrogram and to color each module for visualization; The heatmap of module-trait associations was generated using the "labeledHeatmap" function.

Interesting gene modules functional analysis from WGCNA network

We integrated differentially expressed genes profile data and interesting gene modules functional analysis from WGCNA network to identify the gene linker between CTNNB1 mutation and M2 macrophage infiltration. We performed gene pathway enrichment and sub-PPI network analyses for interesting modules with the help of Metascape [19].

Statistical analysis

Most of the statistical analysis were performed by online bioinformatic databases and tools as mentioned. Differentially expressed mRNAs were calculated by DESeq2 R package. P-values < 0.05 were considered statistically significant. The visualization of the data was done by ggplot2 R package.

Results

CTNNB1 mutation frequently occurred in HCC

A flow diagram summarizing the study selection process is shown in Fig. 1. Summarizing the study selection process, including CTNNB1 mutation frequency, differential expression analysis, survival analysis, gene set enrichment analysis (GSEA), drug priority analysis, immune filtrating analysis, weighted gene co-expression network analysis (WGCNA), and integrative analysis.

As was showed in Fig. 2A, 2B, 2C, and 2E, mutations in the CTNNB1 gene are frequent genetic alterations in pan-cancer. Of them, CTNNB1 mutation was detected in 21% of Chinese HCC patients, and D32G/Y/V was a hot spot (Fig. 2C-2D). Findings from TCGA pan-cancer atlas showed that CTNNB1 mutation was screened in 26% of European and American crow, and S45P/F/Y was a hot spot (Fig. 2E-2F). The difference of mutation frequency and mutation site in different populations may provide new insights for patients with CTNNB1 gene mutation. And then, genome integrity (43%) and Wnt signaling pathway (37%) were significantly altered in HCC (Fig. 2A).

CTNNB1 and TP53, AXIN1 mutations are frequently mutually exclusive, while it had the synergistic phenomenon with OBSCN (Supplementary 2A-2B). Compared with the CTNNB1 mutations, patients with CTNNB1 wild-type carried more TP53, WNT pathway (AXIN1), BRCA1 associated protein 1(BAP1), and PI3K-mTOR (TSC2) (Supplementary 2C). Interestingly, there was no significant survival event difference between CTNNB1 mutation and non-mutation in in hepatocellular carcinoma (Supplementary 3).

The dysfunction of cell-cell adhesion and Wnt pathway was enriched in presence of CTNNB1 mutation

Principal component analysis for HCC vs. para-cancer tissues, HCC vs. paired para-cancer tissue, CTNNB1 mutation vs. non-mutation were showed in Fig. 3A-3C. A total of 4507 differentially expressed genes were selected; 3285 were up-regulated and 1222 were down-regulated in HCC relative to para-cancer tissue (Fig. 3D). A total of 4047 differentially expressed genes were selected; 2854 were up-regulated and 1193 were down-regulated in HCC relative to paired para-cancer tissue (Fig. 3E). A total of 2283 differentially expressed genes were selected; 500 were up-regulated and 2854 were down-regulated in CTNNB1 mutation relative to non-mutation (Fig. 3E).

Considering that the number of differential genes was too large, it was difficult to accurately make GO and pathway enrichment analysis directly. We intersect differential expression from three collection above, 181 co-genes were up-regulated and 191 co-genes were down-regulated (Fig. 3G and 3H). Functional enrichment analysis showed the dysfunction of cell-cell adhesion and Wnt pathway was enriched in presence of CTNNB1 mutation (Fig. 3G and 3H).

As was illustrated in Supplementary 4, hallmark_E2F_target, and hallmark_G2M_checkpoint was significantly enriched in HCC vs. para-cancer tissues and HCC vs. paired para-cancer tissue. Hallmark_oxidative_phosphorylation and hallmark_wnt_beta_catenin_signaling was enriched in CTNNB1 mutation vs. non-mutation.

Kinase protein OBSCN was significantly altered in Supplementary 5A, the differential expression of the top 30 kinase protein was showed in Supplementary 5B. 50 cancer hallmark pathway and Wnt pathway related gene set were illustrated in Supplementary 5C-5D.

Association between drug sensitivity and CTNNB1 mutation

15 kinds of small molecules associated with CTNNB1 mutation were found in the DREIMT database, and they mainly related to glucocorticoid receptor agonist, acetylcholine receptor antagonist/agonist, cyclooxygenase inhibitors, tubulin polymerization inhibitors, and protein synthesis inhibitors, which may be involved in the potential mechanism for the treatment of CTNNB1 mutation HCC (Fig. 4A).

Subsequently, RNAactDrug was used to predict anticancer drugs associated with CTNNB1 mutation (Fig. 4B-4D). T-test showed that drug sensitivity of Nutlin-3 and PHA-665752 remained significantly difference between CTNNB1 mutation and non-mutation (Fig. 4B). Pearson and Spearman correlation coefficient from CCLE (Fig. 4B)., celllminer (Fig. 4C), and GDSC (Fig. 4D) drug database showed that CTNNB1 expression was negatively correlated with PLX4720, Trametinib, 6-benzylthioinsone, and PD0325901, while CTNNB1 expression was positively correlated with L-685458, PD-0332991, and sb-682330-a. The findings above suggested that these anti-tumor drugs may become potential therapeutic drugs for CTNNB1 mutation HCC.

High genetic score and TMB in CTNNB1 mutation HCC

As was illustrated in Fig. 5, high tumor mutation burden (TMB), neoantigen loads, intra-tumoral heterogeneity, number of segments, and homologous recombination defects were elevated in the presence of CTNNB1 mutation HCC.

M2 macrophages was significant enriched in CTNNB1 mutation

Findings from Cibersort (Fig. 6A), CTNNB1 mutation had higher levels of M2 macrophages and mast resting cell, while CTNNB1 wildtype HCC tissues had higher levels of CD4 T resting memory cell. Figure 6B showed that macrophage was enriched in CTNNB1 mutation HCC. Furthermore, Fig. 6C suggested that M1 macrophage was enriched in CTNNB1 wildtype HCC. M2 macrophage cell was increased in the presence of CTNNB1 mutation HCC (Fig. 6D)

Collectively, our results suggested M2-type macrophages are significantly enriched in CTNNB1-mutated HCC.

Evaluation of the immunotherapy efficacy and CTNNB1 mutation

The results from IPS methods (Fig. 7A) suggested that decreased effector cells and increased suppressor cells was exhibited in the CTNNB1 mutation HCC, while CTNNB1 mutation HCC remained high immunophenoscore. As was showed in Fig. 7B-7C, CTNNB1 mutation had higher levels of M2 macrophages, while CTNNB1 wildtype HCC tissues had higher levels of TIDE score, CAF, PDCD1, LAG3, HAVCR2 TIGIT and CD8.

Gene module associated with CTNNB1 mutation and M2 macrophage cell

We constructed co-expression network to identify gene module associated with CTNNB1 mutation and M2 macrophage cell, with soft thresholds (power) equal 14 based on the scale-free topology fit indices (Fig. 8A). Then, 11 gene functional modules were identified and exhibited as a hierarchical clustering dendrogram (Fig. 8B). The heatmap of module-trait relationships also suggested that the different functional modules with associated with CTNNB1 mutation and M2 macrophage cell (Fig. 8C). Importantly, we identified two interesting gene modules, including MEmagenta (gene modules positively correlated to CTNNB1 mutation and M2), and MEbrown (gene modules negatively correlated to CTNNB1 mutation and M2).

Enrichment analysis and sub-hub-PPI for interesting gene module from WGCNA

As was showed in Fig. 9A and 9B, genes positively correlated to CTNNB1 mutation and M2 was involved in Wnt signaling pathway and fatty acid derivative biosynthetic process, which consist of Bossard P et al. findings[32], while genes negatively correlated to CTNNB1 mutation and M2 was enriched in leukocyte activation and inflammatory response.

We integrated differentially expressed genes and co-expression gene, Fig. 9C-9F showed that 21 genes and 3 gene was obtained. PPI network constructed by STRING database showed there are close association among RNF43, LGR5, BMP4, AXIN2, ASPSCR1, GLUL, TNFRSF19 (Fig. 9G), CCL2 and FCGR2B remained high connetity (Fig. 9H).

.

Discussion

CTNNB1 mutation HCC was a subtype of liver cancer with a high degree of malignancy and rapid growth[1]. Our research seek for the treatment of CTNNB1 mutant HCC from the perspective of immune cell infiltration, which has great application value. Our findings highly highlighted that M2 macrophage abnormalities are the most significant immune cell changes in CTNNB1-mutant HCC and the WGCNA algorithm was used to characterize the target genes that crosstalk CTNN1B mutations and M2 macrophages (Fig. 10).

There were differences in mutation frequency and hot spot mutation regions in different populations, which also suggests that the clinical use of drugs targeting CTNNB1 should consider population differences. In this study, we applied public databases to analyze the frequency of CTNNB1 mutations and hotspot mutations on the TCGA of European and American populations and the China dataset of Chinese populations, which was also confirmed. Subsequently, we used the TCGA database for analysis, but did not make an in-depth analysis of the Chinese population, which is also the shortcoming of this study.,

Analysis of differentially expressed genes in cancer and para-cancerous tissues, liver cancer and paired samples, and CTNNB1-mutated and non-mutated liver cancers indicated that WNT signaling pathway activation and cell adhesion inhibition were significantly altered signaling pathways in CTNNB1-mutated liver cancer. In addition, the kinase protein OBSCN was significantly mutated in CTNNB1-mutated HCC.

Remarkably, WNT signaling pathway-related proteins LGR5, AXIN2, and DKK4 were significantly up-regulated, while SFRP5 gene was significantly down-regulated in CTNNB1-mutated HCC). Similarityly, Chen X et al., suggested[33] high expression of tumor suppressor transcription factor TBX3 in CTNNB1 HCC and downregulation of TBX3 in non-CTNNB1 mutant tumors. These results suggest that we should pay more attention to the role of certain tumor suppressors genes in CTNNB1-mutant HCC/

CTNNB1 mutations may cause changes in sensitivity to certain drugs, such as Nutlin-3 (effectively restores p53 function and induces cell cycle arrest and apoptosis in MDM2 expression human rhabdomyosarcoma cells with wild-type p53.) and PHA-665752 (MET inhibitor). Application of these drugs in CTNNB1-mutated HCC may improve the prognosis of liver cancer. Besides, the development of targeted drugs based on the principle of SYNTHETIC death of CTNNB1 is also our future research direction.

CTNNB1 mutations are significantly associated with various genomic scores, such as high TMB, etc. High TMB, microsatellite instability, neoantigen loads, intratumor heterogeneity score, number of segments, and homologous recombination defects score were significantly increased in CTNNB1 mutations group. Besides, Cibersort, EPIC, quantiseq, and xcell immune method suggested M2-type macrophages are significantly enriched in CTNNB1-mutated HCC. Interestingly, CTNNB1-mutated HCC showed a low level in immune checkpoint signature score .These 21 upregulated co-expression and 3 down-regulated co-expression genes significantly mediate the interaction of CTNNB1 mutations with M2 macrophages.

In conclusion, we hope our results provide novel insights to assist in the design of new immunotherapeutic drugs, to help clinicians choose appropriate drugs for CTNNB1 mutation HCC. Besides, our research was based on bioinformatic exploration and requires validation

Declarations

DATA AVAILABILITY STATEMENT

The datasets analyzed in this study were described in MATERIALS AND METHODS. 

ETHICS STATEMENT

This studymet the publication guidelines stated by TCGA (https://cancergenome.nih.Gov/publications/publicationguidelines). 

CONFLICTS OF INTEREST

The authors declare that they have no conflict of interest. 

AUTHOR CONTRIBUTIONS

Yong-gang Luo contributed to the conception of the study and wrote the manuscript. Zhong-neng Xu contributed to experimental technology and experimental design. Qi Wang performed the data analyses. Jian-qiang Zhao supervised the study. 

FUNDING

None.

References

  1. SIEGEL R L, MILLER K D, FUCHS H E, et al. Cancer statistics, 2022 [J]. CA: a cancer journal for clinicians, 2022, 72(1): 7-33.
  2. BRAY F, FERLAY J, SOERJOMATARAM I, et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries [J]. CA: a cancer journal for clinicians, 2018, 68(6): 394-424.
  3. FORNER A, REIG M, BRUIX J. Hepatocellular carcinoma [J]. Lancet (London, England), 2018, 391(10127): 1301-14.
  4. KOULOURIS A, TSAGKARIS C, SPYROU V, et al. Hepatocellular Carcinoma: An Overview of the Changing Landscape of Treatment Options [J]. Journal of hepatocellular carcinoma, 2021, 8(387-401.
  5. HE J, ZENG Z, WANG Y, et al. Characterization of novel CTNNB1 mutation in Craniopharyngioma by whole-genome sequencing [J]. Molecular cancer, 2021, 20(1): 168.
  6. BUGTER J M, FENDERICO N, MAURICE M M. Mutations and mechanisms of WNT pathway tumour suppressors in cancer [J]. Nature reviews Cancer, 2021, 21(1): 5-21.
  7. DESJONQUERES E, CAMPANI C, MARRA F, et al. Preneoplastic lesions in the liver: Molecular insights and relevance for clinical practice [J]. Liver international : official journal of the International Association for the Study of the Liver, 2022, 42(3): 492-506.
  8. GAO Q, ZHU H, DONG L, et al. Integrated Proteogenomic Characterization of HBV-Related Hepatocellular Carcinoma [J]. Cell, 2019, 179(2): 561-77.e22.
  9. PINYOL R, SIA D, LLOVET J M. Immune Exclusion-Wnt/CTNNB1 Class Predicts Resistance to Immunotherapies in HCC [J]. Clinical cancer research : an official journal of the American Association for Cancer Research, 2019, 25(7): 2021-3.
  10. CHEN B, KHODADOUST M S, LIU C L, et al. Profiling Tumor Infiltrating Immune Cells with CIBERSORT [J]. Methods in molecular biology (Clifton, NJ), 2018, 1711(243-59.
  11. ARAN D, HU Z, BUTTE A J. xCell: digitally portraying the tissue cellular heterogeneity landscape [J]. Genome biology, 2017, 18(1): 220.
  12. GOLDMAN M J, CRAFT B, HASTIE M, et al. Visualizing and interpreting cancer genomics data via the Xena platform [J]. Nature biotechnology, 2020, 38(6): 675-8.
  13. SANCHEZ-VEGA F, MINA M, ARMENIA J, et al. Oncogenic Signaling Pathways in The Cancer Genome Atlas [J]. Cell, 2018, 173(2): 321-37.e10.
  14. MAYAKONDA A, LIN D C, ASSENOV Y, et al. Maftools: efficient and comprehensive analysis of somatic variants in cancer [J]. Genome research, 2018, 28(11): 1747-56.
  15. HOADLEY K A, YAU C, HINOUE T, et al. Cell-of-Origin Patterns Dominate the Molecular Classification of 10,000 Tumors from 33 Types of Cancer [J]. Cell, 2018, 173(2): 291-304.e6.
  16. GAO J, AKSOY B A, DOGRUSOZ U, et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal [J]. Science signaling, 2013, 6(269): pl1.
  17. CERAMI E, GAO J, DOGRUSOZ U, et al. The cBio Cancer Genomics Portal: An Open Platform for Exploring Multidimensional Cancer Genomics Data [J]. Cancer discovery, 2012, 2(5): 401-4.
  18. LOVE M I, HUBER W, ANDERS S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2 [J]. Genome biology, 2014, 15(12): 550.
  19. ZHOU Y, ZHOU B, PACHE L, et al. Metascape provides a biologist-oriented resource for the analysis of systems-level datasets [J]. Nature communications, 2019, 10(1): 1523.
  20. DEBRABANT B. The null hypothesis of GSEA, and a novel statistical model for competitive gene set analysis [J]. Bioinformatics (Oxford, England), 2017, 33(9): 1271-7.
  21. CHEN T, ZHANG H, LIU Y, et al. EVenn: Easy to create repeatable and editable Venn diagrams and Venn networks online [J]. Journal of genetics and genomics = Yi chuan xue bao, 2021, 48(9): 863-6.
  22. HäNZELMANN S, CASTELO R, GUINNEY J. GSVA: gene set variation analysis for microarray and RNA-seq data [J]. BMC bioinformatics, 2013, 14(7.
  23. TROULé K, LóPEZ-FERNáNDEZ H, GARCíA-MARTíN S, et al. DREIMT: a drug repositioning database and prioritization tool for immunomodulation [J]. Bioinformatics (Oxford, England), 2021, 37(4): 578-9.
  24. DONG Q, LI F, XU Y, et al. RNAactDrug: a comprehensive database of RNAs associated with drug sensitivity from multi-omics data [J]. Briefings in bioinformatics, 2020, 21(6): 2167-74.
  25. LIN A, QI C, WEI T, et al. CAMOIP: a web server for comprehensive analysis on multi-omics of immunotherapy in pan-cancer [J]. Briefings in bioinformatics, 2022, 23(3):
  26. PLATTNER C, FINOTELLO F, RIEDER D. Deconvoluting tumor-infiltrating immune cells from RNA-seq data using quanTIseq [J]. Methods in enzymology, 2020, 636(261-85.
  27. RACLE J, GFELLER D. EPIC: A Tool to Estimate the Proportions of Different Cell Types from Bulk Gene Expression Data [J]. Methods in molecular biology (Clifton, NJ), 2020, 2120(233-48.
  28. CHAROENTONG P, FINOTELLO F, ANGELOVA M, et al. Pan-cancer Immunogenomic Analyses Reveal Genotype-Immunophenotype Relationships and Predictors of Response to Checkpoint Blockade [J]. Cell reports, 2017, 18(1): 248-62.
  29. FU J, LI K, ZHANG W, et al. Large-scale public data reuse to model immunotherapy response and resistance [J]. Genome medicine, 2020, 12(1): 21.
  30. PARK S, CHOI Y D, KIM J, et al. Efficacy of immune checkpoint inhibitors according to PD-L1 tumor proportion scores in non-small cell lung cancer [J]. Thoracic cancer, 2020, 11(2): 408-14.
  31. LANGFELDER P, HORVATH S. WGCNA: an R package for weighted correlation network analysis [J]. BMC bioinformatics, 2008, 9(559.
  32. SENNI N, SAVALL M, CABRERIZO GRANADOS D, et al. β-catenin-activated hepatocellular carcinomas are addicted to fatty acids [J]. Gut, 2019, 68(2): 322-34.
  33. LIANG B, ZHOU Y, QIAN M, et al. TBX3 functions as a tumor suppressor downstream of activated CTNNB1 mutants during hepatocarcinogenesis [J]. Journal of hepatology, 2021, 75(1): 120-31.