Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

A mechanistic framework for cardiometabolic and coronary artery diseases


Coronary atherosclerosis results from the delicate interplay of genetic and exogenous risk factors, principally taking place in metabolic organs and the arterial wall. Here we show that 224 gene-regulatory coexpression networks (GRNs) identified by integrating genetic and clinical data from patients with (n = 600) and without (n = 250) coronary artery disease (CAD) with RNA-seq data from seven disease-relevant tissues in the Stockholm–Tartu Atherosclerosis Reverse Network Engineering Task (STARNET) study largely capture this delicate interplay, explaining >54% of CAD heritability. Within 89 cross-tissue GRNs associated with clinical severity of CAD, 374 endocrine factors facilitated inter-organ interactions, primarily along an axis from adipose tissue to the liver (n = 152). This axis was independently replicated in genetically diverse mouse strains and by injection of recombinant forms of adipose endocrine factors (EPDR1, FCN2, FSTL3 and LBP) that markedly altered blood lipid and glucose levels in mice. Altogether, the STARNET database and the associated GRN browser ( provide a multiorgan framework for exploration of the molecular interplay between cardiometabolic disorders and CAD.

This is a preview of subscription content, access via your institution

Relevant articles

Open Access articles citing this article.

Access options

Rent or buy this article

Get just this article for as long as you need it


Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Schematic overview of study flow, including main data-analysis steps, result summary and conclusions.
Fig. 2: Cross-tissue and tissue-specific coexpression network modules capture phenotypic variation associated with CMD and CAD traits.
Fig. 3: Major heritability contributions of GRNs.
Fig. 4: A high-hierarchy network organization of GRNs reveals that topologically central cross-tissue networks are important for CAD development.
Fig. 5: Endocrine factors in cross-tissue networks mediate signaling between GRNs in metabolic tissues and the arterial wall.
Fig. 6: Adipose endocrine factors in the cross-tissue GRN78 targeting liver-specific GRN98 are associated with liver and plasma lipid and glucose levels.
Fig. 7: Injection of mice with GRN78 endocrine factors affects GRN98 and levels of liver and plasma lipids and blood glucose.

Data availability

Case and control STARNET data are available at the dbGAP site (dbGaP study accession phs001203.v1.p1). Validation data are provided by the HMDP25, GTEx32 and morbid obesity33 studies. Liver RNA-seq data generated in response to injecting mice with recombinant endocrine factors are available under GEO entry GSE189001. Source data are provided with this paper.

Code availability

Computer code used for data analysis in this study is available at


  1. Kumar, V., Hsueh, W. A. & Raman, S. V. Multiorgan, multimodality imaging in cardiometabolic disease. Circ. Cardiovasc. Imaging 10, e005447 (2017).

  2. Rask-Madsen, C. & Kahn, C. R. Tissue-specific insulin signaling, metabolic syndrome, and cardiovascular disease. Arterioscler. Thromb. Vasc. Biol. 32, 2052–2059 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  3. Beverly, J. K. & Budoff, M. J. Atherosclerosis: pathophysiology of insulin resistance, hyperglycemia, hyperlipidemia, and inflammation. J. Diabetes 12, 102–104 (2020).

    Article  PubMed  Google Scholar 

  4. Schadt, E. E. et al. An integrative genomics approach to infer causal associations between gene expression and disease. Nat. Genet. 37, 710–717 (2005).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  5. Schadt, E. E. Molecular networks as sensors and drivers of common human diseases. Nature 461, 218–223 (2009).

    Article  CAS  PubMed  Google Scholar 

  6. Schadt, E. E. & Bjorkegren, J. L. NEW: network-enabled wisdom in biology, medicine, and health care. Sci. Transl. Med. 4, 115rv111 (2012).

    Article  Google Scholar 

  7. Bjorkegren, J. L., Kovacic, J. C., Dudley, J. T. & Schadt, E. E. Genome-wide significant loci: how important are they? Systems genetics to understand heritability of coronary artery disease and other common complex disorders. J. Am. Coll. Cardiol. 65, 830–845 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  8. Zhang, B. et al. Integrated systems approach identifies genetic nodes and networks in late-onset Alzheimer’s disease. Cell 153, 707–720 (2013).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  9. Talukdar, H. A. et al. Cross-tissue regulatory gene networks in coronary artery disease. Cell Syst. 2, 196–208 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. Cohain, A. T. et al. An integrative multiomic network model links lipid metabolism to glucose regulation in coronary artery disease. Nat. Commun. 12, 547 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Marbach, D. et al. Wisdom of crowds for robust gene network inference. Nat. Methods 9, 796–804 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. van der Wijst, M. G. P., de Vries, D. H., Brugge, H., Westra, H. J. & Franke, L. An integrative approach for building personalized gene regulatory networks for precision medicine. Genome Med. 10, 96 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

  13. Zhang, B. & Horvath, S. A general framework for weighted gene co-expression network analysis. Stat. Appl. Genet. Mol. Biol. 4, Article17 (2005).

    Article  PubMed  Google Scholar 

  14. Zhang, B. & Zhu, J. Identification of key causal regulators in gene networks. Proc. World Congr. Eng. 2013 II, 1309–1312 (2013).

    Google Scholar 

  15. Shang, M. M. et al. Lim domain binding 2: a key driver of transendothelial migration of leukocytes and atherosclerosis. Arterioscler. Thromb. Vasc. Biol. 34, 2068–2077 (2014).

    Article  CAS  PubMed  Google Scholar 

  16. Wang, Y. et al. Clonally expanding smooth muscle cells promote atherosclerosis by escaping efferocytosis and activating the complement cascade. Proc. Natl Acad. Sci. USA 117, 15818–15826 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Marbach, D. et al. Tissue-specific regulatory circuits reveal variable modular perturbations across complex diseases. Nat. Methods 13, 366–370 (2016).

    Article  PubMed  PubMed Central  Google Scholar 

  18. Wang, I. M. et al. Systems analysis of eleven rodent disease models reveals an inflammatome signature and key drivers. Mol. Syst. Biol. 8, 594 (2012).

    Article  PubMed  PubMed Central  Google Scholar 

  19. Franzen, O. et al. Cardiometabolic risk loci share downstream cis- and trans-gene regulation across tissues and diseases. Science 353, 827–830 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. Dobrin, R. et al. Multi-tissue coexpression networks reveal unexpected subnetworks associated with disease. Genome Biol. 10, R55 (2009).

    Article  PubMed  PubMed Central  Google Scholar 

  21. Ritchie, S. C. et al. A scalable permutation approach reveals replication and preservation patterns of network modules in large datasets. Cell Syst. 3, 71–82 (2016).

    Article  CAS  PubMed  Google Scholar 

  22. Aibar, S. et al. SCENIC: single-cell regulatory network inference and clustering. Nat. Methods 14, 1083–1086 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Huynh-Thu, V. A., Irrthum, A., Wehenkel, L. & Geurts, P. Inferring regulatory networks from expression data using tree-based methods. PLoS ONE 5, e12776 (2010).

  24. Nelson, C. P. et al. Association analyses based on false discovery rate implicate new loci for coronary artery disease. Nat. Genet. 49, 1385–1391 (2017).

    Article  CAS  PubMed  Google Scholar 

  25. Ghazalpour, A. et al. Hybrid mouse diversity panel: a panel of inbred mouse strains suitable for analysis of complex genetic traits. Mamm. Genome 23, 680–692 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Zeng, L. et al. Contribution of gene regulatory networks to heritability of coronary artery disease. J. Am. Coll. Cardiol. 73, 2946–2957 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Yao, D. W., O’Connor, L. J., Price, A. L. & Gusev, A. Quantifying genetic effects on disease mediated by assayed gene expression levels. Nat. Genet. 52, 626–633 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Buniello, A. et al. The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Res. 47, D1005–D1012 (2019).

    Article  CAS  PubMed  Google Scholar 

  29. Langfelder, P. & Horvath, S. Eigengene networks for studying the relationships between co-expression modules. BMC Syst. Biol. 1, 54 (2007).

    Article  PubMed  PubMed Central  Google Scholar 

  30. Chickering, D. M. Optimal structure identification with greedy search. J. Mach. Learn. Res. 3, 507–554 (2002).

    Google Scholar 

  31. Seldin, M. M. et al. A strategy for discovery of endocrine interactions with application to whole-body metabolism. Cell Metab. 27, 1138–1155 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. GTEx Consortium. Human genomics. The Genotype–Tissue Expression (GTEx) pilot analysis: multitissue gene regulation in humans. Science 348, 648–660 (2015).

    Article  PubMed Central  Google Scholar 

  33. Greenawalt, D. M. et al. A survey of the genetics of stomach, liver, and adipose gene expression from a morbidly obese cohort. Genome Res. 21, 1008–1016 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Sabatine, M. S. PCSK9 inhibitors: clinical evidence and implementation. Nat. Rev. Cardiol. 16, 155–165 (2019).

    Article  CAS  PubMed  Google Scholar 

  35. von Scheidt, M. et al. Applications and limitations of mouse models for understanding human atherosclerosis. Cell Metab. 25, 248–261 (2017).

    Article  Google Scholar 

  36. Deshmukh, A. S. et al. Proteomics-based comparative mapping of the secretomes of human brown and white adipocytes reveals EPDR1 as a novel batokine. Cell Metab. 30, 963–975 (2019).

    Article  CAS  PubMed  Google Scholar 

  37. Jones, P. D. et al. JCAD, a gene at the 10p11 coronary artery disease locus, regulates hippo signaling in endothelial cells. Arterioscler. Thromb. Vasc. Biol. 38, 1711–1722 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Davidson, E. H. Emerging properties of animal gene regulatory networks. Nature 468, 911–920 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Kauffman, S. Gene regulation networks: a theory for their global structure and behaviors. Curr. Top. Dev. Biol. 6, 145–182 (1971).

    Article  CAS  PubMed  Google Scholar 

  40. Hastie, T., Tibshirani, R. & Friedman, J. H. The Elements of Statistical Learning: Data Mining, Inference, and Prediction 2nd edn. (Springer, 2009).

  41. Leek, J. T., Johnson, W. E., Parker, H. S., Jaffe, A. E. & Storey, J. D. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics 28, 882–883 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Buuren, S. V. & Groothuis-Oudshoorn, K. mice: multivariate imputation by chained equations in R. J. Stat. Softw. 45, 1–67 (2011).

  43. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).

    Article  PubMed  PubMed Central  Google Scholar 

  44. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013).

    Article  CAS  PubMed  Google Scholar 

  45. Langfelder, P. & Horvath, S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9, 559 (2008).

    Article  PubMed  PubMed Central  Google Scholar 

  46. Mi, H., Muruganujan, A., Ebert, D., Huang, X. & Thomas, P. D. PANTHER version 14: more genomes, a new PANTHER GO-slim and improvements in enrichment analysis tools. Nucleic Acids Res. 47, D419–D426 (2019).

    Article  CAS  PubMed  Google Scholar 

  47. Lambert, S. A. et al. The human transcription factors. Cell 172, 650–665 (2018).

    Article  CAS  PubMed  Google Scholar 

  48. Shu, L. et al. Mergeomics: multidimensional data integration to identify pathogenic perturbations to biological systems. BMC Genomics 17, 874 (2016).

    Article  PubMed  PubMed Central  Google Scholar 

  49. Nikpay, M. et al. A comprehensive 1,000 Genomes-based genome-wide association meta-analysis of coronary artery disease. Nat. Genet. 47, 1121–1130 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. CARDIoGRAMplusC4D Consortium et al. Large-scale association analysis identifies new risk loci for coronary artery disease. Nat. Genet. 45, 25–33 (2013).

  51. Speed, D., Hemani, G., Johnson, M. R. & Balding, D. J. Improved heritability estimation from genome-wide SNPs. Am. J. Hum. Genet. 91, 1011–1021 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  52. Yang, J. et al. Common SNPs explain a large proportion of the heritability for human height. Nat. Genet. 42, 565–569 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  53. Wang, L. & Michoel, T. Efficient and accurate causal inference with hidden confounders from genome–transcriptome variation data. PLoS Comput. Biol. 13, e1005703 (2017).

    Article  PubMed  PubMed Central  Google Scholar 

  54. Chen, L. S., Emmert-Streib, F. & Storey, J. D. Harnessing naturally randomized transcription to infer regulatory relationships among genes. Genome Biol. 8, R219 (2007).

    Article  PubMed  PubMed Central  Google Scholar 

  55. Yavorska, O. O. & Burgess, S. MendelianRandomization: an R package for performing Mendelian randomization analyses using summarized data. Int. J. Epidemiol. 46, 1734–1739 (2017).

    Article  PubMed  PubMed Central  Google Scholar 

  56. Cooper, G. F. et al. The center for causal discovery of biomedical knowledge from big data. J. Am. Med. Inform. Assoc. 22, 1132–1136 (2015).

    Article  PubMed  PubMed Central  Google Scholar 

  57. Bostrom, P. et al. A PGC1-α-dependent myokine that drives brown-fat-like development of white fat and thermogenesis. Nature 481, 463–468 (2012).

    Article  PubMed  PubMed Central  Google Scholar 

  58. Lusis, A. J. et al. The Hybrid Mouse Diversity Panel: a resource for systems genetics analyses of metabolic and cardiovascular traits. J. Lipid Res. 57, 925–942 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  59. Serruys, P. W. et al. Assessment of the SYNTAX score in the Syntax study. EuroIntervention 5, 50–56 (2009).

    Article  PubMed  Google Scholar 

  60. Mark, D. B. et al. Continuing evolution of therapy for coronary artery disease. Initial results from the era of coronary angioplasty. Circulation 89, 2015–2025 (1994).

    Article  CAS  PubMed  Google Scholar 

  61. Phillips, P. C. Epistasis—the essential role of gene interactions in the structure and evolution of genetic systems. Nat. Rev. Genet. 9, 855–886 (2008).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references


J.L.M.B. acknowledges research support from NIH R01HL125863, the American Heart Association (A14SFRN20840000), the Swedish Research Council (2018-02529) and the Heart Lung Foundation (20170265) and by AstraZeneca through ICMC, Karolinska Institutet, Sweden. J.L.M.B. and H.S. acknowledge support from the Foundation Leducq (‘PlaqueOmics: Novel Roles of Smooth Muscle and Other Matrix Producing Cells in Atherosclerotic Plaque Stability and Rupture’, 18CVD02; ‘CADgenomics: Understanding CAD Genes’, 12CVD02) and the European Union under grant agreement HEALTH-F2-2013-601456 (‘CVgenes-at-target’). H.S. received grants for a British Heart Foundation (BHF)–German Center of Cardiovascular Research (DZHK) collaboration and from the ERA-NET (‘Druggable-MI-Genes’, 01KL1802) and was supported by grants from the Federal German Ministries (AbCD-Net grant 01ZX1706C, BLOCK-CAD grant 16GW0198K and ModulMax grant ZF4590201BA8) and the DFG as part of the Sonderforschungsbereich CRC 1123 (B2) and the Transregio TRR 267 (B05) as well as the DigiMed Bayern project (DBM-1805-0001). J.C.K. acknowledges research support from the National Institutes of Health (R01HL130423, R01HL135093 and R01HL148167-01A1) and New South Wales health grant RG194194. A.J.L. acknowledges the Foundation Leducq (‘CADgenomics: Understanding CAD Genes’, 12CVD02) and NIH grants R01 HL144651, R01 HL147883, PO1 HL28481 and R01 DK117850. M.S. acknowledges NIH grant HL138193.

Author information

Authors and Affiliations



S.K. and J.L.M.B. conceived the study. S.K. performed computational analysis and created STARNET software. M.S. and A.J.L. performed mouse validation and provided data from the HMDP. R.E., A.R. and J.L.M.B. are principal investigators of the STARNET study. R.E. is responsible for the STARNET phenotype database. K.S. is responsible for and L.A., A.J. and R.K.J. assisted with the STARNET biobank. S.P., L.Z. and H.S. performed the CAD heritability analysis. S.B., A.D.N., H.C., K.H. and T.M. performed MR analyses. A.M., V.M., H.T., A.C., O.F. and T.M. assisted with the network analysis. O.A.M., K.I. and J.C.K. analyzed angiograms to obtain angiographic scoring. J.L.M.B., S.K., H.S., M.S., A.J.L., K.S., R.E., E.E.S., J.C.K., A.R., C.G., L.-M.G., C.A., S.M.H., C.W., P.R. and G.M. prepared the manuscript.

Corresponding author

Correspondence to Johan L. M. Björkegren.

Ethics declarations

Competing interests

J.L.M.B. is the founder of Clinical Gene Networks (CGN). J.L.M.B. (chair) and A.R. are on CGN’s board of directors. J.L.M.B., A.R. and T.M. are shareholders in CGN. J.L.M.B. receives financial compensation as a consultant for CGN. CGN has an invested interest in STARNET that is regulated in an agreement with the Icahn School of Medicine at Mount Sinai. Neither the Icahn School of Medicine at Mount Sinai nor CGN have made claims to results presented in this study. E.E.S. is the CEO of Sema4. C.W. and L.-M.G. are employees of AstraZeneca. No funding for this study was received from Sema4. AstraZeneca supported this study through independent grants to J.L.M.B. at the Karolinska Institutet (ICMC). The remaining authors declare no competing interests.

Peer review information

Nature Cardiovascular Research thanks Thomas Quertermous and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

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

Extended data

Extended Data Fig. 1 Overview of matching STARNET CAD cases to controls without obstructive CAD according to age, gender and BMI.

The 600 CAD cases underwent coronary artery by-pass grafting (CABG) surgery due to obstructive CAD. The 250 controls were patients eligible to open heart surgery other than CABG (mainly valve replacements) who had no signs of obstructive CAD in pre-operative angiograms (SYNTAX score =0). a, Overview of steps used to select RNA samples from the 600 CAD cases for sequencing by matching them to RNA samples available from the 250 controls according to age, gender and body mass index (BMI). b, Evolutionary algorithm optimizing an objective function quantifying the discrepancy of univariate distribution of age, BMI, and gender between selected cases and controls. c, Distribution of age and BMI (kernel probability density estimates, upper panel) and bar plots of gender for all cases, matched cases and controls (lower panel). d, Bar plots showing per-tissue sample sizes based on tissue sample availability (left panel). Only STARNET subjects in whom at least 3 tissues were sampled were included (right panel). e, Two box plots and one bar plot showing the final balances in age (upper panel) and BMI (middle panel), and gender (lower panel), respectively, between the selected CAD cases and the CAD-negative controls. Median center, lower and upper quartile box, and 1.5 interquartile range whiskers. Two-tailed t-tests.

Extended Data Fig. 2 Case-control differential gene expression statistics adjusted for metabolic phenotypes, drug treatments and genotype using DESeq2.

The metabolic featured adjusted for comprised: plasma triglycerides, HbA1c, Waist/Hip ratio, plasma low-density lipoprotein (LDL) cholesterol, plasma high-density lipoprotein (HDL) cholesterol, total cholesterol, and C-reactive protein (CRP). The drug treatments adjusted for comprised as discrete categories: β-blocker, Thrombin inhibitors, Long-acting nitrate, Short-acting nitrates, ACE inhibitors or ARB, Loop diuretic, Thiazide diuretic, Lipid lowerer, Ca-channel blocker, Carvedilol, Anticoagulant, Oral antidiabetics, and Insulin. Representing population genotype, we used the first 4 components from MDS analysis of the SNP array data.

Extended Data Fig. 3 PCA plots of STARNET case and control RNA-seq samples.

a, PCA plot of RNA sequence data from five tissues sampled in both CAD cases and controls. Cases are represented by circles and controls by triangles. Read counts were normalized by DESeq size factors. b, Zoom in of PCA plots for individual tissues.

Extended Data Fig. 4 Schematics of the multiscale network modeling and normalization of RNA-seq data from the STARNET cases.

a, Schematic illustration of multiscale network modeling based on RNA-seq from multiple patient tissue samples. b, Analytic flow chart of the RNA-seq analysis. Green, blue and red boxes represent input data, steps of data normalization and network inferences, respectively (see also Methods). c, Typical effects of normalization on Pearson’s correlations between 1,000 randomly selected AOR transcripts. The grey lines indicate equivalence with different β-values used in WGCNA. d, Boxplots of n = 499,500 pairwise correlation coefficients among randomly selected genes. Median center, lower and upper quartile box, and 1.5 interquartile range whiskers. e, Density plots demonstrating the quenching effect on AOR transcript correlations following normalization. f, Representative example of the resolution of a partly spurious correlation most likely due to batch effects from different laboratory. The point shape corresponds to laboratory, colors to RNA-seq flow cell, and size to patient age. g, Surrogate variable used for adjustment that correlated to patient ID and not to other known covariates.

Extended Data Fig. 5 Optimization of scale-free properties of co-expression networks across tissues.

To achieve scale-free properties for both tissue-specific and cross-tissue co-expression networks using weighted gene co-expression network analysis (WGCNA), different β-values (exponentiation parameter of network weights) must be used for gene-gene interactions within and between tissues (Methods). a, β-value calibration curves per tissue and between all tissue combinations, showing the TOM-adjusted scale-free network properties of tissue-specific and cross-tissue networks at different β-values. b, The sizes of co-expression modules (left) and tissue specificity (right) detected using a single (upper panel, green), dual (middle, red) and complete set of pairwise β-values (lower, blue). WGCNA with a single, average β-value for all networks both within and across tissues resulted in 336 predominantly tissue-specific modules. A dual approach with two different β-values: average β-values for networks within tissues (the diagonal in panel c) and for networks across different tissues (off-diagonal β-values in panel c). This resulted in 224 co-expression modules whereof 158 contained genes in ≥2 tissues (70.5%). A complete use of the optimal β-value for each specific pair of tissues (β-values in panel c) resulted in 183 co-expression modules. c, Scatterplot showing the cross-tissue fraction of module transcripts at different module sizes (number of genes). The line indicates the 5% threshold used to define cross-tissue modules. d, Number of detected cross-tissue modules (y-axis) using single, dual and complete β values and the cross-tissue fraction thresholds of 5% (left) and 0.5% (right). e, Detection of a previously characterized and replicated liver-specific co-expression module10 identified from tissue-specific WGCNA of STARNET using a single (top, green), dual (middle, red) and complete (bottom, blue) β-values in the WGCNA across tissues. f, Overlapping genes with liver co-expression module for the most similar module inferred using single, dual and complete β-values with WGCNA across tissues. g, Frequency of co-expression modules with increasing numbers of cross-tissue genes.

Extended Data Fig. 6 Cross-tissue co-expression module inference using block-wise WGCNA.

To accommodate the high number of transcripts quantified in the multitissue RNA-seq data, we used block-wise WGCNA, first partitioning transcripts from multiple tissues into 5 blocks using k-means. a, Barplot showing the tissue composition of transcripts segmented into 5 k-means clusters. b, Barplot showing hierarchical clustering of the tissue composition of cross-tissue co-expression modules detected with WGCNA. Each bar corresponds to one co-expression module. Y-axis shows the number of transcripts by tissue. Only modules containing <5,000 transcripts were included. c, Line plot showing the mean fraction of Pearson’s correlation coefficients (y-axis) captured in the 93 cross-tissue modules as a function of cross-tissue correlation cutoff criteria (x-axis). The means are calculated over the 21 tissue pairs in STARNET with the blue area indicating 95% confidence intervals. The vertical dotted line represents a correlation network edge criterium of 0.5, with all supporting tissues pairs plotted as points. d, Hierarchical clustering of co-expression modules, as shown in Fig. 2 d with module IDs.

Extended Data Fig. 7 Co-expression network module replication and tissue associations with CAD traits.

a, Concordance with GTEx RNA-seq data of gene-gene correlation coefficients in tissue-specific modules and in, b, cross-tissue modules, excluding gene-gene interactions within tissue. Modules indicated with black circle showed significant similarity at FDR < 0.05, Bonferroni corrected for n = 224 tests. c, d, In the absence of normal distributions of network correlation coefficients and weights, we also used 4 one-sided permutation tests implemented in the NetRep R package21 for replication of tissue-specific (c) and cross-tissue (d) co-expression modules. The scale is -log10 P values from the permutation test for the 4 different measures of network validation, where the extreme corresponds to P < 0.001 indicating the maximum sensitivity based on 1,000 permutations. e, Violin plots comparing tissue fractions of genes in co-expression modules that are significantly associated (+) with SYNTAX score (upper panel), Duke score (middle panel), or enriched (+) in differentially expressed CAD genes (Hypergeometric test) (Fig. 2c) with the tissue fraction of module genes that are not (-) (FDR < 0.05). P-values comparing the ‘+’and “-“module groups are calculated using Wilcoxon two-sided signed-rank test. For SYNTAX and Duke scores, which estimate arterial atherosclerosis, the co-expression modules were associated by aggregating gene-level P-values using Fisher’s method. f, QQ plots comparing tissue-specific and cross-tissue co-expression module aggregated P-values (Fisher’s method) for Duke (upper panel) and SYNTAX score (middle panel) and module enrichment of CAD DEGs. In contrast to CAD DEGs, cross-tissue modules are significantly more associated with Duke and SYNTAX scores than tissue-specific modules (Wilcoxon rank-sum test).

Extended Data Fig. 8 A tutorial of the STARNET browser.

The STARNET browser ( brings the mechanistic framework of GRNs to the bench-side of the individual CMD and CAD scientist for immediate use. In the displayed example, 304 lead SNPs identified by the most recent GWAS of CAD28 are queried using the STARNET browser. Six of these lead SNPs are found as eQTLs in GRN39. Together with plasma LDL/HDL levels, these lead SNPs regulate the activity of GRN39 that in turn governs the extracellular matrix organization in the coronary arteries, thereby affecting CAD development (Fig. 3).

Extended Data Fig. 9 Mendelian Randomization to determine gene-regulatory network causality.

Stack plots showing fraction of tissue-specific GRNs considered causal in MR analysis (Methods). The hypergeom function from the SciPy stats package in Python62 was used to test overlaps with tissue-specific causal networks (blue, Extended Data Table 5) in relation to the total number of GRNs in each STARNET tissue (yellow, FDR < 5% according to Storey and Tibshirani63).

Extended Data Fig. 10 Modeling network module eigengene-eigengene interactions in a supernetwork to capture high-level tissue patterns of CAD disease progression.

a, Bayesian supernetwork inferred from eigengene values of the 224 STARNET co-expression modules. The likelihood of directed acyclic graphs was assessed using the Bayesian information criteria (BIC) and optimized using fast greedy equivalence search from Tetrad. Network layout was determined by the Fruchterman-Reingold algorithm. Left, colors indicate primary tissues, middle, green indicates cross-tissue network modules and right, yellow indicates tissue-specific network modules. b, Diagram showing bootstrap confidence estimates of supernetwork edges (Methods). Random sets of subjects were drawn and used to infer an ensemble of 1,000 Bayesian networks, from which supernetwork edges were ranked. Bootstrapped P values estimated by 1–frequency of edges in network ensemble. c, Bar plot showing the distribution of 153 statistically robust edges in the supernetwork among the four possible combinations of cross-tissue (CT) and tissue-specific (TS) module interactions. Edge robustness was assessed by bootstrap analysis (M = 10,000) of the full supernetwork of 224 module nodes and 49,952 possible edges (Methods). d, Box plots comparing centrality of cross-tissue modules (n = 89) compared to tissue-specific modules (n = 135). Median center, lower and upper quartile box, 1.5 interquartile range whiskers. The network centrality analysis was based on the betweenness measure, counting the number of shortest paths through each node, performed on a 224×224 matrix of supernetwork eigengene interaction P-values (bootstrap estimates). Comparison by two-sided t test. e, Venn diagram showing 113 GRNs significantly enriched in at least one of 3 CAD clinical measures; 32 GRNs with at least 2 CAD measures, and 8 GRNs with all 3 CAD measures. The CAD measures of the GRNs are enrichment with genes associated with SYNTAX (1) and/or Duke (2) scores (FDR < 0.01) and/or with DEGs in CAD cases versus controls (3). f,g, High-confidence directed edges in the supernetwork (see Methods) detected between 25 of the 32 CAD-associated GRNs with more than two CAD measures (a) colored according to module tissue distribution (f) and enrichment with candidate genes identified by GWAS (g and Fig. 3e). The layout was determined by the Sugiyama algorithm. Indicated below each supernetwork node is the GRN identification number. h, Scatter plot of Pearson correlations between the eigengene values of the most significant edge in the supernetwork between cross-tissue network module 78 and tissue-specific network module 98. P-values were calculated using a two-tailed t-test.

Supplementary information

Reporting Summary

Supplementary Table 1 Clinical summary statistics of individuals with CAD and controls in the STARNET study.


Supplementary Table 2 PANTHER GO enrichment for DEGs in CAD by tissue and downregulation and upregulation in disease. Fisher’s exact test, Benjamini–Hochberg-adjusted FDR estimates.

Supplementary Table 3 GRN contributions to CAD H2.


Supplementary Table 4 GWAS CAD genes and their enrichment in STARNET coexpression modules. P values represent gene set enrichment by hypergeometric tests, adjusted for multiple hypotheses (number of modules) with the Benjamini–Hochberg method.


Supplementary Table 5 MR to identify causal key drivers in GRNs. The representation of causal network genes within key driver modules was measured using a hypergeometric function to compare the overlap between causal networks and module genes. The intersection represents the number of overlapping genes between Findr causal networks and key drivers in modules. Hypergeometric P values were derived from the probability mass function of the hypergeometric distribution, and q values represent P values adjusted for multiple testing using the Storey and Tibshirani method.


Supplementary Table 6 MR to identify GRN key drivers causal for CMD and CAD traits. For the GRN key driver–CMD–CAD trait MR analysis, cis eQTL in AOR, MAM, BLOOD, LIV, SKM, SF and VAF samples were tested for causal associations with 11 STARNET traits: eight for CMDs (BMI, waist/hip ratio, CRP, total cholesterol, LDL and HDL cholesterol, triglycerides and HbA1c) and three for CAD (number of lesions and SYNTAX and Duke scores). Reported P values are nominal and adjusted by the Bonferroni and Benjamini–Hochberg methods using one-sided Fisher’s exact test.


Supplementary Table 7 Identified endocrine candidates (n = 374) and validation statistics for adipose-to-liver factors in HMDP, GTEx and morbid obesity gene expression data.

Source data

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Koplev, S., Seldin, M., Sukhavasi, K. et al. A mechanistic framework for cardiometabolic and coronary artery diseases. Nat Cardiovasc Res 1, 85–100 (2022).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:

This article is cited by


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing