Skip to main content

Network-based inference of master regulators in epithelial membrane protein 2-treated human RPE cells



The application of cell-specific construction of transcription regulatory networks (TRNs) to identify their master regulators (MRs) in EMP2 induced vascular proliferation disorders has been largely unexplored.


Different expression gene (DEGs) analyses was processed with DESeq2 R package, for public RNA-seq transcriptome data of EMP2-treated hRPECs versus vector control (VC) or wild type (WT) hRPECs. Virtual Inference of protein activity by Enriched Regulon analysis (VIPER) was used for inferring regulator activity and ARACNE algorithm was conducted to construct TRNs and identify some MRs with DEGs from comparisons.


Functional analysis of DEGs and the module analysis of TRNs demonstrated that over-expressed EMP2 leads to a significant induction in the activity of regulators next to transcription factors and other genes implicated in vasculature development, cell proliferation, and protein kinase B signaling, whereas regulators near several genes of platelet activation vascular proliferation were repressed. Among these, PDGFA, ALDH1L2, BA1AP3, ANGPT1 and ST3GAL5 were found differentially expressed and significantly activitve in EMP2-over-expressed hRPECs versus vector control under hypoxia and may thus identified as MRs for EMP2-induced lesion under hypoxia.


MRs obtained in this study might serve as potential biomarkers for EMP2 induced lesion under hypoxia, illustrating gene expression landscapes which might be specific for diabetic retinopathy and might provide improved understanding of the disease.

Peer Review reports


Pathologic retinal neovascularization [1] is a potentially blinding consequence seen in many common diseases including diabetic retinopathy [2], retinopathy of prematurity [3], retinal vascular occlusive diseases [4], and neovascular age-related macular degeneration (AMD) [5,6,7] among others. Epithelial membrane protein 2 (EMP2) has been shown to reasonably modulate activity through neovascularization network in the retinal pigment epithelial cell line ARPE-19 [8, 9] which is reported to be important factors in progress of DR. To date, several studies have applied this approach to compare transcriptomes in vascular proliferation process within various cell types in humans [10, 11], mice [12]. However, to the best of our knowledge, there has been no genome-wide gene expression profiling study specifically comparing the transcriptom within EMP2-treated hRPECs.

Gene expression profiling by RNA sequencing (RNA-Seq) provides an opportunity to better understand the underlying mechanisms of vascular proliferation. Therefore, given the large number of genes potentially involved, it has been challenging to identify which are the core set of genes that play major roles in the pathways or networks. It becomes even more challenging if the recent hypothesis of omnigenic model is true for DR. Thus, it is imperative to identify disease-relevant core gene networks, and possibly the network master regulators (MRs), which, if exist, are more likely to be targeted for therapeutic interventions [13, 14]. The term MR were defined for transcription factors (TFs) and genes inflicting regulatory effects on their targets [15].

ARACNe is an unbiased algorithm that infers direct transcriptional interactions based on the mutual information between each transcriptional regulator and its potential targets [16,17,18]. Transcriptional regulatory networks (TRNs) conducted by ARACNe summarize the connections between transcription factors and the genes that they regulate, known as their “target genes” [19], in which case, almost all of the genes expressed in a disease-relevant cell type may confer risk for the disease through widespread network interactions with a core set of master regulators (MRs) [9]. With the constructed regulatory networks, MRs could be identified by computing the activity of each regulator (i.e., TFs, TGs, and DEGs) with the R package of VIPER (GitHub commit 0170c27).

Hence, in this study we compared EMP2-treated hRPECs and health control hRPECs at the transcriptome level using RNA-Seq analysis and core gene networks construction technology to identify the determinant MRs as well as the potential underlying pathyway which regulatoring vascular proliferation in hRPECs (Figure S1).


All samples had high-quality RNA-seq reads with an average of 96% reads per sample. Reads with mapping quality score (MAPQ) < 10 or those with > 5 mismatches in 100 bp aligned region were discarded (Fig. 1A-B).

Fig. 1
figure 1

PCA and box plot of all samples. PCA scatter plot and box plot of all 24 RNA-seq samples, which were obtained based on Hisat2. PCA and box plot graphics illustrate variance of gene expression data of each sample. The percentages on each axis represent the percentages of variation explained by the principal components. Cutoff value was absolute log2FC > 1 and P value < 0.05. A PCA plot of RNA-Seq mapped reads of each sample. B Box plot of the log2 normalized expression values. PCA, Principal Component Analysis

Differential Expression of DEGs and functional enrichment analysis of EMP2-treated comparisons

After low-level processing of 24 EMP2-treated samples and control samples (Fig. 1A-B), we performed differential expression analysis for each treated group comparing with control samples. According to the PCA plot (Fig. 1A), we compared EMP2-OE hRPECs with VC and EMP2-KD hRPECs with WT group. Totally we identified 1239 DEGs in all comparisons of EMP2-treated hRPECs versus VC or WT (Table S1). Figure 2A and B were Venn diagram of DEGs in comparisons under hypoxia and normoxia, respectively (Fig. 2A-B). Figure 2C-D showed up regulated or down regulated DEGs in comparisons under hypoxia and normoxia, respectively (Fig. 2C-D). Then we performed functional enrichment analysis and found that DEGs of EMP2-OE hRPECs vs VC under hypoxia commonly enriched in positive regulation of cell proliferation, vasculature development, protein kinase B signaling, and negative regulation of platelet activation, and enriched in metabolism and folate biosynthesis pathway(Figure S2A-S2B), which did not show in either EMP2-OE hRPECs vs VC under normoxia, or in EMP2-KD hRPECs vs WT under hypoxia or normoxia (Figs. 3 and 4).

Fig. 2
figure 2

Identification of differential expressed genes (DEGs) in EMP2-regulated RPE cells. Differentially Expressed Genes (DEGs) Enrichment Venn Diagram and its bar graph. Venn diagrams were created using the web-based tool Identifying DEGs and up- and down- regulated DEGs terms in deferentially expressed sequences and regulated situation in treated groups vs control groups. DESeq2 R package was used in different expression gene counting. Cutoff value was absolute log2FC > 1 and P value < 0.05. A DEGs in four comparisons under hypoxic condition, B DEGs in four comparisons under normoxic condition, C-D Up- and down- regulated DEGs in comparisons under hypoxic and normoxic condition, respectively. Colors display the expression levels of each DEGs, as pink and purple indicate down- and up-regulated genes in treated group comparing with control group, respectively. DEGs: differentially expressed genes, kd_vc: EMP2-knocked down hRPECs vs vector control hRPECs, kd_wt: EMP2-knocked-down hRPECs vs wild type hRPECs, oe_vc: EMP2-over expressed hRPECs vs vector control hRPECs, oe_wt: EMP2-over expressed hRPECs vs wild type hRPECs

Fig. 3
figure 3

DEGs GO analysis in EMP2-regulated RPE cells. Bar graph of Gene Ontology (GO) enrichment analysis of DEGs. GO enrichment analysis terms in significant DEGs for treated group vs control. (p < 0.01, FDR q < 0.05, overlap cutoff > 0.5) (A) DEGs Go analysis under hypoxic condition, (B) DEGs Go analysis under normoxic condition

Fig. 4
figure 4

KEGG pathway analysis of the DEGs. Bar graph of Kyoto Encyclopedia of Genes, Genomes (KEGG) Pathway analysis. Enrichment of KEGG pathway terms in differentially expressed sequences for EMP2-treated group vs control. The RichR ( package was used in enrichment analysis. (p < 0.01, FDR q < 0.05, overlap cutoff > 0.5) (A) KEGG Pathway analysis of DEGs under hypoxic condition, (B) KEGG Pathway analysis of DEGs under normoxic condition

Transcriptome regulatory network (TRN) construction of comparisons of EMP2 treated hRPECs with control

ARACNe was run independently on the four comparisons datasets of DEGs using a conservative mutual information threshold (p ≤ 1.0 × 10–9, i.e., p ≤ 0.05 Bonferroni corrected for all candidate interactions). This resulted in highly robust TRNs of transcriptional factors (TFs) and potential target genes (TGs). To further elucidate the regulation and interaction relationship, TRNs were visualized using Cytoscape 3.7.1. In TRNs, a node is either a TG or a TF, an edge is a TF-TF, TF-TG or TG-TG relationship. Totally, 199,093 interactions were counted between 974 TFs and their inferred TGs among four comparisons (Fig. 6A-B, Tables S3-S6).

Regulator activity inference and master regulators (MRs) identification

To further analyze the activity of identified regulators (i.e., DEGs, TFs and TGs), then inferred top MRs, the R package of VIPER was used to compute the activity of each regulator with the constructed regulatory networks. Of these, 1395 regulators were identified as high activity and their differential expression among four groups under hypoxia were compared by using Venn diagram (Table S2, Fig. 5). In present study, TRNs modules for comparisons of EMP2-treated hRPECs versus VC or WT hRPECs depicted MRs with high intramodular connectivity, i.e., genes most connected with all other genes within the module. Thus, we got PDGFA, ALDH1L2, BA1AP3, ANGPT1 and ST3GAL5 as MRs in the TRN of hypoxia EMP2-OE vs VC hRPECs (Fig. 6A, Table S3), PRSS33 as a MR in the TRN of hypoxia KD vs WT hRPECs (Fig. 6B, Table S4),

Fig. 5
figure 5

Activity of regulators of 4 comparisons under hypoxia. Venn diagram of different activity of regulators in four comparisons under hypoxia. Venn diagrams were created using the web-based tool Differently expressed activity of regulators in four comparisons under hypoxia were identified. (p < 0.01, FDR q < 0.05, overlap cutoff > 0.5). hy kd_vc: EMP2-knocked down hRPECs vs vector control hRPECs under hypoxia, hy kd_wt: EMP2-knock-down hRPECs vs wild type hRPECs under hypoxia, hy oe_vc: EMP2-over expressed hRPECs vs vector control hRPECs under hypoxia, hy oe_wt: EMP2-over expressed hRPECs vs wild type hRPECs under hypoxia

Fig. 6
figure 6

Transcription regulator networks (TRNs) of activity of regulators and DEGs. TRNs diagram of regulator activities and DEGs based on ARACNe and VIPER algorithm. TRNs diagram illustrated transcriptional factors (TFs) and potential target genes (TGs). VIPER was conducted to identify activity of TFs and TGs, then inferred potential master regulators (MRs) and mechanisms by regulating of EMP2 in different conditions. Cytoscape software (version: 3.8.0) was used for visualizing TRNs. In this study, we defined “node” as either a TF or a TG, “edge” as a TF-TF, TF-TG or TG-TG relationship. “MR” was defined as a TF or a TG having many connections with others, which were marked with red color in the diagram. (p < 0.01, FDR q < 0.05, overlap cutoff > 0). A TRN of comparisons of EMP2-OE hRPECs vs VC under hypoxic condition, B TRN of comparisons of EMP2-KD vs WT under hypoxic condition. TRNs, transcription regulatory networks, MRs, master regulators


Since early 2010s, RNA-sequencing (RNAseq) has been used to profile transcriptome of retinal neovascularization disease. Differential expression genes (DEGs) analysis can straightforwardly analyze a gene expression dataset. Through DEGs analysis, a set of highly up-regulated and down-regulated genes can be identified by comparing two groups of samples (e.g., hypoxia vs. normoxia). However, gene expression is highly dynamic, and the expression quantification may depend on the techniques (e.g., different platforms of microarray, or RNAseq), making the cross-dataset comparison difficult. The activities of regulators cannot be directly measured by microarray or RNA-seq because these techniques only measure RNA expression level and do not consider gene activity changes by post-translational modifications. Fortunately, activity of regulators can be inferred by its regulons through VIPER, by taking RNA data and a context-specific gene regulatory network (interactome) as inputs [20,21,22].

In present study, transcriptome analysis of EMP2-over-expressed and knocked down hRPECs under hypoxia or normoxia elucidated eight sets of DEGs which were then used to construct TRNs along with variant analysis. TRNs analyses attempted to simplify the interactions of activity of regulators and emphasized their role in EMP2-treated RPE cells, then illustrated top MRs. MRs are often obscured in standard differential expression analyses (such as those including environmental stressors) and indeed many of MRs identified in our network analysis were not significant in the DEGs analysis, highlighting the strength of the network approach (Fig. 5).

Platelet derived growth factor (PDGF) is one of the leading vascular proliferation risk genes identified by GWAS [40]. However, the molecular mechanism underlying the genetic association remains elusive. Here, we have shown computationally that PDGFA is one of the top MRs of vascular proliferation factors in hRPECs under hypoxia. By examining the PDGFA-perturbed gene networks in EMP2- over-expressed hRPECs, we found that PDGFA-altered genes in hRPECs are more enriched for gene pathways that are related to proliferation and migration activities. PDGF families were proved to induce the proliferation and migration effects on RPE cells in proliferative retinopathy (PVR), and has been reported to participate in pericyte regulated vascular proliferation, vessel stabilization, and contribute to the formation of both the blood–brain and blood-retina barriers by regulating pericyte-endothelial cell communication [23,24,25,26]. Furthermore, we have shown that PDGFA-altered genes in EMP2-over-expressed hRPECs under hypoxia are more enriched for credible inflammation, oxidative stress, apoptosis GWAS risk genes [27]. Besides, we inferred that when EMP2 was over expressed under hypoxia, PDGFA was upregulated and enhanced the interactions with other genes. Thus, multiple lines of evidence from our study suggest that the PDGFA gene network expression activity in the early stages of vascular proliferation may be very important for pathogenesis.

Although our computationally identification of MR has focused on PDGFA, other MRs that we identified here suggest that additional gene networks are relevant to vascular proliferation. The identified MRs contain many of the TFs previously reported in the literature that may potentially be involved in the vascular proliferation pathogenesis of DR, such as ALDH1L2, BA1AP3, ANGPT1 and ST3GAL5. ALDH1L2 (Aldehyde Dehydrogenase 1 Family Member L2) is a protein coding gene which functions as mitochondrial folate enzyme [28], the maintenance of mitochondrial integrity and energy balance of the cell [29]. The related pathways of LDH1L2 are one carbon pool by folate and metabolism of water-soluble vitamins and cofactors (Figure S2). ALDH1L2 has been previously proven to have correlation with EMP2 in breast cancer cells [30]. but it was the first time to be identified in EMP2-treated hRPECs. We also noticed that PDGFA connected with ALDH1L2 by PLSCR4. In this way, PDGFA may mediate accelerated ATP-independent bidirectional trans-bilayer migration of phospholipids upon binding calcium ions, and it will result in a loss of phospholipid asymmetry in the plasma membrane. This process may play a central role in the initiation of fibrin clot formation, in the activation of mast cells as well as in the recognition of apoptotic and injured cells by the reticuloendothelial system. Thus, we thought that over-expression of EMP2 would activate PDGFA through ALDH1L2-PLSCR4 pathway by initializing fibrin clot and recognizing apoptosis.

We also identified that BAI1 Associated Protein 3 (BAIAP3) had correlation with MMP16 [31]. BAIAP3, functioning in endosome to Golgi retrograde transport, may regulate behavior and food intake by controlling calcium-stimulated exocytosis of neurotransmitters including NPY and serotonin and hormones like insulin. Such neurotransmitters have been proven to have the ability to degrade various components of the extracellular matrix, such as collagen type III and fibronectin in the pathogenesis of diabetic cataract [17].

This investigation was mainly based on ARACNe algorithm analysis of transcriptome information of hRPECs treated by hypoxia and exogenous EMP2. The TRN constructed in this paper provided clues of interaction mechanism between these genes, which not only provided theoretical support for revealing the regulatory relationship among them, but also provided candidate genes, i.e., up-and down-regulated pathway genes, transcriptional factors, and MRs, which could be studied for future SNP discovery work to establish association with different phenotype or traits of interest. We also established a scientific exploration mode for exploring effective targets of anti-neovascularization, which made the follow-up basic verification and clinical research more reliable. However, the results of bioinformatics analysis ultimately need to be verified by many experiments and combined with clinical practice to carry out gene function and mechanism verification for substantive exploration.


In summary, by using a computational approach, we identified PDGFA, ALDH1L2, BA1AP3, ANGPT1 and ST3GAL5 as MRs that likely contributed to EMP2-induced vascular proliferation in hRPECs under hypoxia. Although powerful, we acknowledged the limitation of our approach in identifying top MRs, because selecting a top MR was not purely “data-driven”; for instance, PDGFA ALDH1L2, BA1AP3, ANGPT1 and ST3GAL5 were identified as top MRs among many other MRs was not only data-driven but also based on prior knowledge about their association with vascular proliferation. Nonetheless, our study suggested that MRs in vascular proliferation in EMP2 and hypoxia treated hRPECs could be identified by transcriptional network construction and that MRs such as PDGFA as well as other MRs might constitute convergent gene networks that confer disease risk in a spatial and temporal manner. Therefore, PDGFA and other identified MRs might be a building part of vascular proliferation architecture, which collectively drived DR onset and progression. Transcriptional network construction in larger and more DR-relevant cell types/stages, combined with empirical network perturbation, would further pave the way to deepen our understanding of the genetic contribution to the complex biology of vascular proliferation during DR disease biology.


Data availability

The RNA-seq data profiles were obtained from the NCBI GEO database ( GSE151610. The dataset contained 24 samples, which contained mRNA profiles of hRPECs panel consisting of EMP2-knocked-down (KD) hRPECs, EMP2-over-expressed (OE) hRPECs, vector control (VC) hRPECs, and wild type (WT) hRPECs under hypoxia or normoxia. This project was approved by the Institutional Review Board of Sir Run Run Hospital, Nanjing Medical University. All procedures of this study complied with the protocol.

Identification of DEGs

The data set is available at the NCBI Short-Read Archive (SRA), under accession SRA265560. The RNA-Seq data set was mapped in the Homo sapiens (human) genome assembly GRCh38 (hg38) by Hisat2. Feature Counts was used in read counts. After filtering all 0 expression genes, DESeq2 R package was used in different expression gene counting. Cutoff value was absolute log2FC > 1 and P value < 0.05.

GO and KEGG pathway enrichment analysis

We performed functional enrichment analysis using Kyoto Encyclopedia of Genes, Genomes (KEGG) pathway and Gene Ontology (GO) with a p value < 0.05 as the significance cutoff. The RichR ( package was used in enrichment analysis (Figure S1).

Transcription regulatory network construction

Regulatory networks were constructed by ARACNe (Algorithm for the Reconstruction of Accurate Cellular Networks) [32,33,34] using significant DEGs. First, mutual interaction between a candidate TF(x) and its potential target (y) was computed by pairwise MI, MI(x, y), using a Gaussian kernel estimator. A threshold was applied on MI based on the null hypothesis of statistical independence (P < 0.05, Bonferroni-corrected for the number of tested pairs). Secondly, the constructed network was trimmed by removing indirect interactions using the data processing inequality (DPI), a property of the MI. Parameters was set to zero DPI (Data Processing Inequality) tolerance and MI (Mutual Information) p-value threshold of 10 − 8. Therefore, for each (x, y) pair, a path through another TF(z) was considered, and every path pertaining to the following constraint was removed: MI (x, y) < min (MI (x, z),MI (z, y)). A P value threshold of 1 × 10 − 8 using DPI = 0.1 (as recommended [35]) was used when running ARACNe. Cytoscape software (version: 3.8.0) was used for visualizing the regulatory networks.

Regulator activity inference

The VIPER algorithm had proven to be a useful tool for estimating regulator activity from gene expression data using enriched regulon analysis. With the constructed regulatory networks, we computed the activity of each regulator (i.e., TFs, TGs, and DEGs) using the R package of VIPER (GitHub commit 0170c27). The output of VIPER is a list of highly active MRs as well as their activity scores and their enrichment P values. A set of regulators with significant activity difference were defined at adjusted p-value < 0.01.

Statistical test

All statistical tests and plotting were performed in R 3.5.2 ( unless otherwise mentioned.

Availability of data and materials

The datasets analyzed during the current study are available in the NCBI GEO database ( GSE151610.



Master regulators

EMP2 :

Epithelial membrane protein 2


Human retinal pigment epithelial cells


Transcription factors


Target genes


Differential expression genes


Victor control


Wild type


Algorithm for the Reconstruction of Accurate Cellular Networks


Virtual Inference of Protein activity by Enriched Regulon analysis


Transcription regulatory networks


Age-related macular degeneration






Short-Read Archive


Kyoto Encyclopedia of Genes, Genomes


Gene Ontology


  1. Plastino F, Pesce NA, André H. MicroRNAs and the HIF/VEGF axis in ocular neovascular diseases. Acta Ophthalmol. 2021;99(8):e1255–62.

    Article  CAS  PubMed  Google Scholar 

  2. Yu G, Aaberg MT, Patel TP, Iyengar RS, Powell C, Tran A, et al. Quantification of Retinal Nonperfusion and Neovascularization With Ultrawidefield Fluorescein Angiography in Patients With Diabetes and Associated Characteristics of Advanced Disease. JAMA Ophthalmol. 2020;138:680.

    Article  Google Scholar 

  3. Aouiss A, Anka Idrissi D, Kabine M, Zaid Y. Update of inflammatory proliferative retinopathy: Ischemia, hypoxia and angiogenesis. Current Research in Translational Medicine. 2019;67:62–71.

    Article  CAS  Google Scholar 

  4. Esmaili DD, Boyer DS. Recent advances in understanding and managing retinal vein occlusions. F1000Res. 2018;7:467.

    Article  Google Scholar 

  5. Kiss S, Malangone-Monaco E, Wilson K, Varker H, Stetsovsky D, Smith D, et al. Real-World Injection Frequency and Cost of Ranibizumab and Aflibercept for the Treatment of Neovascular Age-Related Macular Degeneration and Diabetic Macular Edema. JMCP. 2020;26:253–66.

    Article  Google Scholar 

  6. Campochiaro PA, Akhlaq A. Sustained suppression of VEGF for treatment of retinal/choroidal vascular diseases. Prog Retin Eye Res. 2020;83:100921.

    Article  Google Scholar 

  7. Ngo Ntjam N, Thulliez M, Paintaud G, Salvo F, Angoulvant D, Pisella P-J, et al. Cardiovascular Adverse Events With Intravitreal Anti-Vascular Endothelial Growth Factor Drugs: A Systematic Review and Meta-analysis of Randomized Clinical Trials. JAMA Ophthalmol. 2021.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Sun M, Wadehra M, Casero D, Lin M-C, Aguirre B, Parikh S, et al. Epithelial Membrane Protein 2 (EMP2) Promotes VEGF-Induced Pathological Neovascularization in Murine Oxygen-Induced Retinopathy. Invest Ophthalmol Vis Sci. 2020;61:3.

    Article  CAS  Google Scholar 

  9. Telander DG, Yu AK, Forward KI, Morales SA, Morse LS, Park SS, et al. Epithelial Membrane Protein-2 in Human Proliferative Vitreoretinopathy and Epiretinal Membranes. Invest Ophthalmol Vis Sci. 2016;57:3112.

    Article  CAS  Google Scholar 

  10. Liu Y, Dakou E, Meng Y, Leyns L. Loss of Emp2 compromises cardiogenic differentiation in mouse embryonic stem cells. Biochem Biophys Res Commun. 2019;511:173–8.

    Article  CAS  Google Scholar 

  11. Yan Y, Jiang R, Yuan P, Wen L, Pang X, Jing Z, et al. Implication of proliferation gene biomarkers in pulmonary hypertension. Anim Models and Exp Med. 2021;4:369–80.

    Article  CAS  Google Scholar 

  12. Villacampa P, Liyanage SE, Klaska IP, Cristante E, Menger KE, Sampson RD, et al. Stabilization of myeloid-derived HIFs promotes vascular regeneration in retinal ischemia. Angiogenesis. 2020;23:83–90.

    Article  CAS  Google Scholar 

  13. Corbett BF, You JC, Zhang X, Pyfer MS, Tosi U, Iascone DM, et al. ΔFosB Regulates Gene Expression and Cognitive Dysfunction in a Mouse Model of Alzheimer’s Disease. Cell Rep. 2017;20:344–55.

    Article  CAS  Google Scholar 

  14. Mondragón E, Maher LJ. Anti-Transcription Factor RNA Aptamers as Potential Therapeutics. Nucleic Acid Ther. 2016;26:29–43.

    Article  Google Scholar 

  15. Singhania A, Graham CM, Gabryšová L, Moreira-Teixeira L, Stavropoulos E, Pitt JM, et al. Transcriptional profiling unveils type I and II interferon networks in blood and tissues across diseases. Nat Commun. 2019;10:2887.

    Article  Google Scholar 

  16. Ruggiero AD, Davis A, Sherrill C, Westwood B, Hawkins GA, Palmer ND, et al. Skeletal muscle extracellular matrix remodeling with worsening glycemic control in nonhuman primates. Ame J Physiol Regul Integr Comp Physiol. 2021;320:R226–35.

    Article  CAS  Google Scholar 

  17. Li Y, Jiang S-H, Liu S, Wang Q. Role of lncRNA NEAT1 mediated by YY1 in the development of diabetic cataract via targeting the microRNA-205-3p/MMP16 axis. Eur Rev Med Pharmacol Sci. 2020;24:5863–70.

    CAS  PubMed  Google Scholar 

  18. Li Y, Chen D, Sun L, Wu Y, Zou Y, Liang C, et al. Induced Expression of VEGFC, ANGPT, and EFNB2 and Their Receptors Characterizes Neovascularization in Proliferative Diabetic Retinopathy. Invest Ophthalmol Vis Sci. 2019;60:4084.

    Article  CAS  Google Scholar 

  19. Yang B, Wittkopp PJ. Structure of the Transcriptional Regulatory Network Correlates with Regulatory Divergence in Drosophila. Mol Biol Evol. 2017;34:1352–62.

    Article  CAS  Google Scholar 

  20. Sánchez-Luis E, Joaquín-García A, Campos-Laborie FJ, Sánchez-Guijo F, De Las Rivas J. Deciphering Master Gene Regulators and Associated Networks of Human Mesenchymal Stromal Cells. Biomolecules. 2020;10:557.

    Article  Google Scholar 

  21. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 2008;9:559.

    Article  Google Scholar 

  22. DoostparastTorshizi A, Armoskus C, Zhang H, Forrest MP, Zhang S, Souaiaia T, et al. Deconvolution of transcriptional networks identifies TCF4 as a master regulator in schizophrenia. Sci Adv. 2019;5:eaau4139.

    Article  Google Scholar 

  23. Alvarez MJ, Shen Y, Giorgi FM, Lachmann A, Ding BB, Ye BH, et al. Functional characterization of somatic mutations in cancer using network-based inference of protein activity. Nat Genet. 2016;48:838–47.

    Article  CAS  Google Scholar 

  24. Yu B, Chen L, Zhang W, Li Y, Zhang Y, Gao Y, et al. TOP2A and CENPF are synergistic master regulators activated in cervical cancer. BMC Med Genomics. 2020;13:145.

    Article  CAS  Google Scholar 

  25. Yang Y, Huang X, Ma G, Cui J, Matsubara JA, Kazlauskas A, et al. PDGFRβ plays an essential role in patient vitreous-stimulated contraction of retinal pigment epithelial cells from epiretinal membranes. Exp Eye Res. 2020;197:108116.

    Article  CAS  Google Scholar 

  26. Ma J, Tang W, Gu R, Hu F, Zhang L, Wu J, et al. SHP-2-Induced Activation of c-Myc Is Involved in PDGF-B-Regulated Cell Proliferation and Angiogenesis in RMECs. Front Physiol. 2020;11:555006.

    Article  Google Scholar 

  27. Cai R, Jiang J. LncRNA ANRIL Silencing Alleviates High Glucose-Induced Inflammation, Oxidative Stress, and Apoptosis via Upregulation of MME in Podocytes. Inflammation. 2020;43:2147–55.

    Article  CAS  Google Scholar 

  28. Krupenko NI, Sharma J, Pediaditakis P, Helke KL, Hall MS, Du X, et al. Aldh1l2 knockout mouse metabolomics links the loss of the mitochondrial folate enzyme to deregulation of a lipid metabolism observed in rare human disorder. Hum Genomics. 2020;14:41.

    Article  CAS  Google Scholar 

  29. Sarret C, Ashkavand Z, Paules E, Dorboz I, Pediaditakis P, Sumner S, et al. Deleterious mutations in ALDH1L2 suggest a novel cause for neuro-ichthyotic syndrome. NPJ Genom Med. 2019;4:17.

    Article  Google Scholar 

  30. Dillard C, Kiyohara M, Mah V, McDermott SP, Bazzoun D, Tsui J, et al. EMP2 Is a Novel Regulator of Stemness in Breast Cancer Cells. Mol Cancer Ther. 2020;19:1682–95.

    Article  CAS  Google Scholar 

  31. Caporarello N, D’Angeli F, Cambria MT, Candido S, Giallongo C, Salmeri M, et al. Pericytes in Microvessels: From “Mural” Function to Brain and Retina Regeneration. IJMS. 2019;20:6351.

    Article  CAS  Google Scholar 

  32. Hwang S-G, Kim K-H, Lee B-M, Moon J-C. Transcriptome analysis for identifying possible gene regulations during maize root emergence and formation at the initial growth stage. Genes Genom. 2018;40:755–66.

    Article  CAS  Google Scholar 

  33. Doostparast Torshizi A, Wang K. Deconvolution of Transcriptional Networks in Post-Traumatic Stress Disorder Uncovers Master Regulators Driving Innate Immune System Function. Sci Rep. 2017;7:14486.

    Article  Google Scholar 

  34. Margolin AA, Nemenman I, Basso K, Wiggins C, Stolovitzky G, Favera RD, et al. ARACNE: An Algorithm for the Reconstruction of Gene Regulatory Networks in a Mammalian Cellular Context. BMC Bioinformatics. 2006;7:S7.

    Article  Google Scholar 

  35. Margolin AA, Wang K, Lim WK, Kustagi M, Nemenman I, Califano A. Reverse engineering cellular networks. Nat Protoc. 2006;1:662–71.

    Article  CAS  Google Scholar 

Download references


The authors acknowledge all the members of their respective teams, from Sir Run Run Hospital, Nanjing Medical University, as well as their collaborators (fundamentalist and clinicians), for their support and interaction during many years of involvement with genetics. We thank the developers of the ARACNe and VIPER software for comments on the software usage. We also thank the KEGG copyright permission for these images cited by us and obtained from KEGG in


This work was supported by grants from the National Key Research and Development Plan of China (No. 2020YFC2008505 to Xiang Lu), the National Natural Science Foundation of China (No. 81970218 to Xiang Lu and No. 81970217 to Wei Gao), the Six Talent Peaks Project of Jiangsu Province (No. WSN-175 to Wei Gao), Jiangsu Geriatrics Society (JGS2019ZXYY06 to Xiang Lu), Project of Jiangsu Provincial Health Science and Technology Commission(LGY2019093 to Jian Qin) and the Jiangsu Government Scholarship for Overseas Studies (JS-2018–082 to Hua Wan).

Author information

Authors and Affiliations



Author HW, FC and WG have given substantial contributions to the conception or the design of the manuscript, author WZ and author ZT to acquisition, analysis and interpretation of the data. All authors have participated to drafting the manuscript, author JQ and XL revised it critically. All authors read and approved the final version of the manuscript.

Corresponding authors

Correspondence to Xiang Lu or Jian Qin.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Wan, H., Gao, W., Zhang, W. et al. Network-based inference of master regulators in epithelial membrane protein 2-treated human RPE cells. BMC Genom Data 23, 52 (2022).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI:


  • Differentially expressed genes (DEGs)
  • Epithelial membrane protein 2 (EMP2)
  • Retinal pigment epithelial cells (hRPECs)
  • Computational inference
  • Gene expression
  • Master regulators (MRs)
  • Transcription regulatory networks (TRNs)
  • Systems biology