A Biological-Systems-Based Analyses Using Proteomic and Metabolic Network Inference Reveals Mechanistic Insights into Hepatic Lipid Accumulation: An IMI-DIRECT Study

A Biological-Systems-Based Analyses Using Proteomic and Metabolic Network Inference Reveals Mechanistic Insights into Hepatic Lipid Accumulation: An IMI-DIRECT Study | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search A Biological-Systems-Based Analyses Using Proteomic and Metabolic Network Inference Reveals Mechanistic Insights into Hepatic Lipid Accumulation: An IMI-DIRECT Study Natalie N. Atabaki , Daniel E. Coral , Hugo Pomares-Millan , Kieran Smith , Harry H. Behjat , Robert W. Koivula , Andrea Tura , Hamish Miller , Katherine Pinnick , View ORCID Profile Leandro Agudelo , Kristine H. Allin , Andrew A. Brown , Elizaveta Chabanova , Piotr J Chmura , Ulrik P. Jacobsen , Adem Y. Dawed , Petra J.M. Elders , Juan J. Fernandez-Tajes , Ian M. Forgie , Mark Haid , Tue H. Hansen , Elizaveta L. Hansen , Angus G. Jones , Tarja Kokkola , Sebastian Kalamajski , Anubha Mahajan , Timothy J. McDonald , Donna McEvoy , Mirthe Muilwijk , Konstantinos D. Tsirigos , Jagadish Vangipurapu , Sabine van Oort , Henrik Vestergaard , Jerzy Adamski , Joline W. Beulens , View ORCID Profile Søren Brunak , Emmanouil T. Dermitzakis , Giuseppe N. Giordano , Ramneek Gupta , Torben Hansen , Leen t Hart , View ORCID Profile Andrew T. Hattersley , Leanne Hodson , Markku Laakso , Ruth J.F. Loos , Jordi Merino , Mattias Ohlsson , Oluf Pedersen , Martin Ridderstråle , Hartmut Ruetten , Femke Rutters , View ORCID Profile Jochen M. Schwenk , Jeremy Tomlinson , Mark Walker , Hanieh Yaghootkar , Fredrik Karpe , View ORCID Profile Mark I McCarthy , View ORCID Profile Elizabeth Louise Thomas , Jimmy D. Bell , Andrea Mari , Imre Pavo , View ORCID Profile Ewan R. Pearson , Ana Viñuela , Paul W. Franks doi: https://doi.org/10.1101/2025.06.02.25328773 Natalie N. Atabaki 1 Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen , Copenhagen, Denmark 2 Department of Clinical Science, Genetic and Molecular Epidemiology, Lund University Diabetes Centre , Helsingborg, Sweden 3 Oxford Centre for Diabetes, Endocrinology and Metabolism, University of Oxford , Oxford, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: paul.franks{at}med.lu.se Daniel E. Coral 2 Department of Clinical Science, Genetic and Molecular Epidemiology, Lund University Diabetes Centre , Helsingborg, Sweden Find this author on Google Scholar Find this author on PubMed Search for this author on this site Hugo Pomares-Millan 2 Department of Clinical Science, Genetic and Molecular Epidemiology, Lund University Diabetes Centre , Helsingborg, Sweden 4 Department of Epidemiology, Geisel School of Medicine at Dartmouth, Dartmouth-Hitchcock Medical Center, Lebanon, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Kieran Smith 3 Oxford Centre for Diabetes, Endocrinology and Metabolism, University of Oxford , Oxford, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Harry H. Behjat 5 Clinical Memory Research Unit, Department of Clinical Sciences, Lund University , Lund, Sweden Find this author on Google Scholar Find this author on PubMed Search for this author on this site Robert W. Koivula 3 Oxford Centre for Diabetes, Endocrinology and Metabolism, University of Oxford , Oxford, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Andrea Tura 6 Institute of Neuroscience, National Research Council , Padova, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Hamish Miller 3 Oxford Centre for Diabetes, Endocrinology and Metabolism, University of Oxford , Oxford, UK 7 Barts Liver Centre, Queen Mary University London and Barts Health NHS Trust , London, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Katherine Pinnick 3 Oxford Centre for Diabetes, Endocrinology and Metabolism, University of Oxford , Oxford, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Leandro Agudelo 8 Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology , Cambridge, USA 9 Broad Institute of MIT and Harvard , Cambridge, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Leandro Agudelo Kristine H. Allin 1 Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen , Copenhagen, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Andrew A. Brown 10 Population Health and Genomics, Ninewells Hospital and Medical School, University of Dundee , Dundee, Scotland Find this author on Google Scholar Find this author on PubMed Search for this author on this site Elizaveta Chabanova 11 Department of Diagnostic Radiology, Copenhagen University Hospital Herlev Gentofte , Herlev, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Piotr J Chmura 12 Department of Health Technology, Technical University of Denmark , Kongens Lyngby, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Ulrik P. Jacobsen 12 Department of Health Technology, Technical University of Denmark , Kongens Lyngby, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Adem Y. Dawed 10 Population Health and Genomics, Ninewells Hospital and Medical School, University of Dundee , Dundee, Scotland Find this author on Google Scholar Find this author on PubMed Search for this author on this site Petra J.M. Elders 13 Department of General Practice, Amsterdam UMC-location Vumc, Amsterdam Public Health Research Institute , Amsterdam, The Netherlands Find this author on Google Scholar Find this author on PubMed Search for this author on this site Juan J. Fernandez-Tajes 2 Department of Clinical Science, Genetic and Molecular Epidemiology, Lund University Diabetes Centre , Helsingborg, Sweden Find this author on Google Scholar Find this author on PubMed Search for this author on this site Ian M. Forgie 10 Population Health and Genomics, Ninewells Hospital and Medical School, University of Dundee , Dundee, Scotland Find this author on Google Scholar Find this author on PubMed Search for this author on this site Mark Haid 14 Metabolomics and Proteomics Core, German Research Center for Environmental Health , Helmholtz Zentrum München, Neuherberg, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Tue H. Hansen 1 Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen , Copenhagen, Denmark 15 Department of Medicine, Zealand University Hospital , Køge, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Elizaveta L. Hansen 11 Department of Diagnostic Radiology, Copenhagen University Hospital Herlev Gentofte , Herlev, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Angus G. Jones 16 Department of Clinical and Biomedical Sciences, University of Exeter College of Medicine & Health , Exeter, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Tarja Kokkola 17 Internal Medicine, Institute of Clinical Medicine, University of Eastern Finland , Kuopio, Finland Find this author on Google Scholar Find this author on PubMed Search for this author on this site Sebastian Kalamajski 2 Department of Clinical Science, Genetic and Molecular Epidemiology, Lund University Diabetes Centre , Helsingborg, Sweden Find this author on Google Scholar Find this author on PubMed Search for this author on this site Anubha Mahajan 18 Wellcome Trust Centre for Human Genetics, University of Oxford , Oxford, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Timothy J. McDonald 19 Blood Sciences, Royal Devon and Exeter NHS Foundation Trust , Exeter, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Donna McEvoy 20 Diabetes Research Network, Royal Victoria Infirmary , Newcastle upon Tyne, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Mirthe Muilwijk 21 Amsterdam UMC, Epidemiology and Data Science , Amsterdam, The Netherlands 22 Amsterdam Public Health, Health Behaviors & Chronic Diseases , Amsterdam, The Netherlands Find this author on Google Scholar Find this author on PubMed Search for this author on this site Konstantinos D. Tsirigos 23 Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark , Kongens Lyngby, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jagadish Vangipurapu 17 Internal Medicine, Institute of Clinical Medicine, University of Eastern Finland , Kuopio, Finland 24 A.I. Virtanen Institute for Molecular Sciences, University of Eastern Finland , Kuopio, Finland Find this author on Google Scholar Find this author on PubMed Search for this author on this site Sabine van Oort 13 Department of General Practice, Amsterdam UMC-location Vumc, Amsterdam Public Health Research Institute , Amsterdam, The Netherlands Find this author on Google Scholar Find this author on PubMed Search for this author on this site Henrik Vestergaard 1 Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen , Copenhagen, Denmark 25 Steno Diabetes Center , Copenhagen, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jerzy Adamski 26 Yong Loo Lin School of Medicine, National University of Singapore , Singapore, Singapore 27 Institute of Experimental Genetics, Helmholtz Zentrum München, German Research Center for Environmental Healt , Neuherberg, Germany 28 Institute of Biochemistry, Faculty of Medicine, University of Ljubljana , Ljubljana, Slovenia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Joline W. Beulens 21 Amsterdam UMC, Epidemiology and Data Science , Amsterdam, The Netherlands Find this author on Google Scholar Find this author on PubMed Search for this author on this site Søren Brunak 12 Department of Health Technology, Technical University of Denmark , Kongens Lyngby, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Søren Brunak Emmanouil T. Dermitzakis 29 Department of Genetic Medicine and Development, University of Geneva Medical School , Geneva, Switzerland Find this author on Google Scholar Find this author on PubMed Search for this author on this site Giuseppe N. Giordano 2 Department of Clinical Science, Genetic and Molecular Epidemiology, Lund University Diabetes Centre , Helsingborg, Sweden Find this author on Google Scholar Find this author on PubMed Search for this author on this site Ramneek Gupta 1 Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen , Copenhagen, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Torben Hansen 1 Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen , Copenhagen, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Leen t Hart 22 Amsterdam Public Health, Health Behaviors & Chronic Diseases , Amsterdam, The Netherlands 30 Department of Cell and Chemical Biology, Leiden University Medical Center , Leiden, The Netherlands 31 Department of Biomedical Data Sciences, Section Molecular Epidemiology, Leiden University Medical Center , Leiden, The Netherlands 32 Department of Epidemiology and Data Science, Amsterdam UMC , Amsterdam, The Netherlands Find this author on Google Scholar Find this author on PubMed Search for this author on this site Andrew T. Hattersley 16 Department of Clinical and Biomedical Sciences, University of Exeter College of Medicine & Health , Exeter, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Andrew T. Hattersley Leanne Hodson 3 Oxford Centre for Diabetes, Endocrinology and Metabolism, University of Oxford , Oxford, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Markku Laakso 17 Internal Medicine, Institute of Clinical Medicine, University of Eastern Finland , Kuopio, Finland Find this author on Google Scholar Find this author on PubMed Search for this author on this site Ruth J.F. Loos 1 Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen , Copenhagen, Denmark 33 The Charles Bronfman Institute for Personalized Medicine, Icahn School of Medicine at Mount Sinai , New York, NY, USA 34 The Mindich Child Health and Development Institute, Icahn School of Medicine at Mount Sinai , New York, New York, USA 35 The Novo Nordisk Foundation Center for Genomic Mechanisms of Disease, Broad Institute of MIT and Harvard , Cambridge, Massachusetts, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jordi Merino 1 Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen , Copenhagen, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Mattias Ohlsson 36 Center for Environmental and Climate Science, Lund University , Lund, Sweden Find this author on Google Scholar Find this author on PubMed Search for this author on this site Oluf Pedersen 37 Center for Clinical Metabolic Research, Herlev and Gentofte University Hospital , Copenhagen, Denmark Find this author on Google Scholar Find this author on PubMed Search for this author on this site Martin Ridderstråle 38 Department of Clinical Science, Lund University , Malmö, Sweden Find this author on Google Scholar Find this author on PubMed Search for this author on this site Hartmut Ruetten 39 Boehringer Ingelheim International GmbH , Ingelheim, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Femke Rutters 21 Amsterdam UMC, Epidemiology and Data Science , Amsterdam, The Netherlands 22 Amsterdam Public Health, Health Behaviors & Chronic Diseases , Amsterdam, The Netherlands Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jochen M. Schwenk 40 Science for Life Laboratory, School of Biotechnology , Solna, Sweden Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jochen M. Schwenk Jeremy Tomlinson 3 Oxford Centre for Diabetes, Endocrinology and Metabolism, University of Oxford , Oxford, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Mark Walker 41 Translational and Clinical Research Institute, Faculty of Medical Sciences, University of Newcastle , Newcastle upon Tyne, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Hanieh Yaghootkar 42 School of Chemistry, College of Health and Science, University of Lincoln , Lincoln, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Fredrik Karpe 3 Oxford Centre for Diabetes, Endocrinology and Metabolism, University of Oxford , Oxford, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Mark I McCarthy 43 Wellcome Trust Centre for Human Genetics, University of Oxford , UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Mark I McCarthy Elizabeth Louise Thomas 44 Research Centre for Optimal Health, School of Life Sciences, University of Westminster , London, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Elizabeth Louise Thomas Jimmy D. Bell 44 Research Centre for Optimal Health, School of Life Sciences, University of Westminster , London, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Andrea Mari 6 Institute of Neuroscience, National Research Council , Padova, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Imre Pavo 45 Eli Lilly Regional Operations Ges.m.b.H. , Vienna, Austria Find this author on Google Scholar Find this author on PubMed Search for this author on this site Ewan R. Pearson 10 Population Health and Genomics, Ninewells Hospital and Medical School, University of Dundee , Dundee, Scotland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ewan R. Pearson Ana Viñuela 10 Population Health and Genomics, Ninewells Hospital and Medical School, University of Dundee , Dundee, Scotland Find this author on Google Scholar Find this author on PubMed Search for this author on this site Paul W. Franks 2 Department of Clinical Science, Genetic and Molecular Epidemiology, Lund University Diabetes Centre , Helsingborg, Sweden 46 Precision Healthcare University Research Institute, Queen Mary University of London , London, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: paul.franks{at}med.lu.se Abstract Full Text Info/History Metrics Data/Code Preview PDF Abstract Objective To delineate organ-specific and systemic drivers of metabolic dysfunction-associated steatotic liver disease (MASLD), we applied integrative causal inference across clinical, imaging, and proteomic domains in individuals with and without type 2 diabetes (T2D). Research Design and Methods We used Bayesian network analyses to quantify causal pathways linking adipose distribution, glycemia, and insulin dynamics with fatty liver using data from the IMI-DIRECT prospective cohort study. Measurements were made of glucose and insulin dynamics (using frequently-sampled metabolic challenge tests), MRI-derived abdominal and liver fat content, serological biomarkers, and Olink plasma proteomics from 331 adults with new-onset T2D and 964 adults free from diabetes at enrolment. The common protocols used in these two cohorts provided the opportunity for replication analyses to be performed. When the direction of the effect could not be determined with high probability through Bayesian networks, complementary two-sample Mendelian randomization (MR) was employed. Results High basal insulin secretion rate (BasalISR) was identified as the primary causal driver of liver fat accumulation in both diabetes and non-diabetes. Excess visceral adipose tissue (VAT) was bidirectionally associated with liver fat, indicating a self-reinforcing metabolic loop. Basal insulin clearance (Clinsb) worsened as a consequence of liver fat accumulation to a greater degree before the onset of T2D. Out of 446 analysed proteins, 34 mapped to these metabolic networks and 27 were identified in the non-diabetes network, 18 in the diabetes network, and 11 were common between the two networks. Key proteins directly associated with liver fat included GUSB, ALDH1A1, LPL, IGFBP1/2, CTSD, HMOX1, FGF21, AGRP, and ACE2. Sex-stratified analyses revealed distinct proteomic drivers: GUSB and LEP were most predictive of liver fat in females and males, respectively. Conclusions Basal insulin hypersecretion is a modifiable, causal driver of MASLD, particularly prior to glycaemic decompensation. Our findings highlight a multifactorial, sex-and disease-stage–specific proteo-metabolic architecture of hepatic steatosis. Proteins such as GUSB, ALDH1A1, LPL, and IGFBPs warrant further investigation as potential biomarkers or therapeutic targets for MASLD prevention and treatment. Introduction About a third of the global adult population is estimated to have metabolic dysfunction-associated steatotic liver disease (MASLD), which is predicted to increase markedly in the coming decades [ 1 ]. In people with type 2 diabetes (T2D), the prevalence of MASLD is about 67% and rising [ 2 ]. Although the epidemiological association between intrahepatic lipid accumulation and T2D is well described, its underlying metabolic and proteomic features remain poorly understood, hindering the prevention and treatment of MASLD [ 3 ]. Hyperinsulinemia, especially excessive early-phase insulin secretion, is a common prelude to T2D and MASLD [ 4 , 5 ] and the hyper-secretion of insulin often seen in people with T2D is exasperated when MASLD co-occurs [ 6 ]. When adipocyte lipid storage capacity is exceeded, ectopic lipid accumulation occurs, with the liver being a major sink for excess triacylglycerol. An over-abundance of metabolic substrates (glucose and fatty acids) drives liver-specific insulin resistance and corresponding de novo hepatic glucose and lipid production. This degenerative cycle leads to systemic elevation in glucose and lipids – major risk factors for T2D [ 7 – 10 ]. Understanding the plasma protein network and its relationship with glucose and insulin dynamics and adipose distribution may elucidate how these metabolic features interact. Recent proteomic studies have explored specific features of MASLD [ 3 , 11 – 13 ] Nevertheless, a comprehensive system-wide model is lacking, hindering understanding of the causal drivers and consequences of liver fat accumulation and its effects on glycose and insulin homeostasis. Addressing this gap may inform diagnostic, preventive, and therapeutic strategies for MASLD. Here, we applied Bayesian network analysis and Mendelian randomization to the comprehensively phenotyped datasets from the IMI-DIRECT cohorts to determine potential causal networks linking plasma proteins, glucose and insulin dynamics with hepatic lipid accumulation. MATERIALS AND METHODS IMI-DIRECT cohorts and measures To avoid imputation biases, analyses focused on complete case data from IMI-DIRECT, a multicentre prospective cohort study involving European-ancestry adults diagnosed with T2D (n=331) and those without T2D (n=964). The latter group included participants with normal and prediabetic glycemia (ascertained by glycated haemoglobin A1C (HbA1c), fasting glucose, or 2-hour glucose). All participants provided written informed consent at enrolment, and the study protocol was approved by the regional research ethics committees of each clinical study centre [ 14 , 15 ]. Frequently-sampled mixed-meal tolerance tests (MMTT) and oral glucose tolerance tests (fsOGTT) were carried out in the diabetes and non-diabetes cohorts, respectively. Insulin secretion estimates were derived from C-peptide deconvolution using the method described by Van Cauter et al. [ 16 ], expressed as pmol/min per square meter of estimated body surface area. Visceral and subcutaneous fat were assessed using T1-weighted MRI scans, while liver and pancreas fat content were measured using a T2*-based multi-echo technique from MRI scans [ 17 ]. Biospecimens from IMI-DIRECT study centres were manually randomized using a mix-shake-distribute procedure and placed into 96-well plates. Proteins in EDTA plasma samples were quantified at SciLifeLab in Stockholm using the Cardiometabolic, Cardiovascular II, Cardiovascular III, Development, and Metabolism panels provided by Olink Proteomics AB (Uppsala, Sweden), following the guidelines for Proximity Extension Assays (PEA) [ 18 ]. Further details on the study design and core characteristics are provided elsewhere [ 14 , 15 ]. Bayesian network analysis We employed Bayesian networks to create directed acyclic graphs (DAGs) that visually depict potential causal connections and interactions between clinical and protein features, using established methods. In these graphs, each variable is represented as a node, and the arcs connecting them suggest possible cause-and-effect relationships. The network’s skeleton was constructed using the Semi-Interleaved HITON-PC (SI-HITON-PC) algorithm. SI-HITON-PC is a constraint-based algorithm that performs conditional correlation-based independence tests (p-value ≤ 0.01)[ 19 ]. This process selects protein features and determines the presence of arcs between clinical and proteomic nodes. The algorithm identifies the Markov blanket for each node, which is the minimal set of nodes that makes the node conditionally independent of other nodes in the network. Directional arcs were then assigned using a hill-climbing algorithm, which orients the arcs based on causal Bayesian theorem while avoiding cycles in the DAG. These steps were carried out through a 10-fold cross-validation process, aggregating predictive and posterior correlation scores from multiple runs of the model. To create a robust structure that resists network perturbations, we used model averaging, resulting in an averaged network based on the frequency of observed potential arcs across all networks. The strength of each arc was determined by how frequently it appears in the model; the probability of the assigned direction was also calculated. The purpose of these analyses is to determine the probability of true-positive discovery and causal effect directionality. To ensure the networks function appropriately, we inverse-normalized all numeric variables to fit a Gaussian Bayesian network with a multivariate normal distribution. Mendelian randomization analysis Following the construction of the Bayesian networks, we performed bi-directional two-sample Mendelian randomization (MR) analysis [ 20 ] on connections between proteins and clinical variables that were strongly associated but with low directional probability (suggesting bidirectional causal associations). For associations between proteins and clinical variables, we utilized the SMR (Summary-based Mendelian randomization) and HEIDI (Heterogeneity in Dependent Instruments) methodologies [ 21 ]. We applied the top pQTL genetic loci as the instrument cis -pQTLs within a ±1 MB window of the gene encoding the target protein. HEIDI analysis was conducted to assess the heterogeneity of each lead SNP estimate relative to those in linkage disequilibrium (LD), helping distinguish biological pleiotropic effects (p-value_Heidi > 0.01) from those driven solely by LD. For reciprocal associations between clinical variables and proteins, we employed the inverse variance weighted (IVW) method as the primary MR analysis along with MR-Egger. We considered MR findings to be statistically significant if the causal association amongst IVW and MR Egger was directionally concordant and the causal association for IVW passed a false discovery rate (FDR) corrected threshold (p-value 0.05). The MR analysis utilized the latest summary statistics from publicly available databases, detailed in Table S1. Statistical analyses were undertaken using R software version 4.2.1. The Bayesian networks were constructed using the bnlearn package [ 22 ], and the MR analysis was performed using TwoSampleMR [ 23 ]. Bayesian networks were visualized using Cytoscape software version 3.10.3 [ 24 ]. The scripts used can be found at https://github.com/Natalie-Atabaki/BN . Variables from IMI-DIRECT were adjusted for the centre of origin of the samples (known as the ‘centre effect’), age, and sex (when appropriate) in a linear model, including each variable per model. For proteins, in addition to biological adjustments, plate adjustments were made. Residuals were extracted from these models and rank normalized to fulfil the Gaussian Bayesian network assumption. Results Within the IMI-DIRECT cohorts, 331 (with T2D) and 964 (without T2D) participants had the required clinical (n=19) and proteomic (n=468) variables for a complete case analysis. An overview of IMI-DIRECT participant characteristics is shown in Table S1 for the non-diabetes and diabetes cohorts. These clinical variables were initially selected based on their association with hepatic lipid accumulation within the IMI-DIRECT cohorts [ 25 ] or from existing literature [ 7 – 10 , 26 ]; these include glycaemic, beta-cell function, adipose volume, and hepatic lipid content variables. Proteomic feature selection in IMI-DIRECT cohorts Of 446 analysed Olink plasma proteins, 34 were identified as key components in the metabolic network, including CTRC, PON3, KITLG, APOM, CD4, LDLR, MFGE8, LPL, MATN2, FAM3C, FGF21, TGM2, IGFBP1, LEP, AGRP, ANXA4, ACE2, HMOX1, CHL1, MSMB, NRP1, FST, GUSB, GHRL, ANGPTL4, IGFBP2, CDH5, CTSD, SRCRB4D, FCRL5, WFIKKN2, CD300LG, ALDH1A1, and GH1, with 27 present in the non-diabetes network and 18 in the diabetes network, of which 11 were shared across networks. Functional enrichment analysis revealed that these proteins are associated with several cellular components, including protein-lipid macromolecular complexes, LDL particles, and the endoplasmic reticulum lumen. Their molecular functions were enriched in signalling receptor binding, hormone activity, and growth factor interactions. Pathway analysis showed overrepresentation in biological processes related to lipoprotein metabolism, ghrelin processing, and fat-soluble vitamin metabolism, highlighting their potential roles in metabolic regulation across health and disease states ( Figure 1 ) [ 27 – 30 ]. Download figure Open in new tab Figure 1. Proteins identified in Bayesian networks and their functional enrichment across metabolic pathways. (Top Left) Venn diagram showing the distribution of 34 proteins identified in the Bayesian networks across non-diabetes and type 2 diabetes cohorts; (Top Right) Reactome pathway enrichment highlighting associated biological processes; (Bottom Left) Molecular function enrichment analysis; (Bottom Right) Cellular component enrichment. The background gene set for the hypergeometric test included 18,731proteining-coding genes. Metabolic Bayesian network in IMI-DIRECT Non-Diabetes cohort To construct the metabolic network, we performed structural and parameter learning using clinical and proteomic features as nodes within the non-diabetic cohort (n=964). To evaluate the network’s fit to the data, we compared the Bayesian information criterion (BIC) scores of the learned network with the BIC density of 1000 randomly generated (null) networks (Figure S1). To ensure robustness and parsimony, we restricted the network to arcs with both strength and directional probabilities ≥0.8. Table 1 outlines the conditional density of each node, along with estimates of the parameters of its upstream (parent) variables, indicating the magnitude of their relationships within the network. View this table: View inline View popup Table 1. Conditional regression models for each node in the averaged Bayesian network (non-diabetes cohort, n = 964). Each row lists the dependent variable (Outcome) and its set of parent variables (Exposures), along with coefficient estimates and standard error (SE). All variables were inverse-normal transformed. The constructed network ( Figure 2 ) identified basal insulin secretion rate (BasalISR) as the main direct causal determinant of hepatic lipid content. Insulin clearance, calculated from basal values as (Clinsb), was identified as a downstream effect of liver fat. Download figure Open in new tab Figure 2. Bayesian network of metabolic and proteomic interactions in the IMI-DIRECT non-diabetes cohort (n = 964). The graph displays directed relationships among clinical and proteomic variables. Nodes are color-coded: blue (clinical/metabolic), green (proteins), peach (liver fat as outcome). Solid arrows represent directed associations with high confidence (strength and direction probability ≥ 0.8), while dashed arrows indicate less confident directionality. AGRP: agouti-related peptide; ALDH1A1: aldehyde dehydrogenase 1 family member A1; APOM: apolipoprotein M; BasalISR: basal insulin secretion rate at the beginning of the OGTT; CD4: cluster of differentiation 4; CDH5: cadherin-5; Clins: mean insulin clearance during OGTT; Clinsb: basal insulin clearance-calculated as (mean insulin secretion)/(mean insulin concentration); CTRC: chymotrypsin C; CTSD: cathepsin D; FGF21: fibroblast growth factor 21; Glucagonmin0: fasting glucagon; Glucose: fasting plasma glucose; GlucoseSens: glucose sensitivity; GUSB: β-glucuronidase; HbA1c: glycated haemoglobin A1C; HDL: high-density lipoprotein cholesterol; IGFBP1/2: insulin-like growth factor binding proteins 1 and 2; Insulin: fasting plasma insulin; KITLG: KIT ligand; LDLR: low-density lipoprotein receptor; LEP: leptin; LiverFat: hepatic fat content; LPL: lipoprotein lipase; MFGE8: milk fat globule-epidermal growth factor 8; OGIS: oral glucose insulin sensitivity index according to the method of Mari et al. [ 32 ]; PancFat: pancreas fat; PON3: paraoxonase 3; SAT: subcutaneous adipose tissue; TG: triglycerides; TotGLP1min0: fasting total GLP-1; TwoGlucose/TwoInsulin: 2-hour post-load values from OGTT; VAT: visceral adipose tissue. BasalISR was the parental node for both liver fat and VAT, leading to fat accumulation in the ectopic and visceral abdominal areas. Given VAT’s strong correlation with liver fat, we hypothesised that BasalISR directly and indirectly (via VAT accumulation) drives liver fat accumulation. The derived network revealed a direct inverse effect of basal hyperinsulinemia on Insulin-like Growth Factor Binding Protein-1 (IGFBP-1), which then positively affects IGFBP2. Insulin suppresses IGFBP-1 production in the liver, indicating that individuals with higher basal insulin secretion rates have lower IGFBP1 levels owing to this suppressive feedback loop. The network also suggests a direct causal effect of liver fat on CTSD (Cathepsin D) levels, an enzyme involved in protein degradation. CTSD has also been associated with lipid metabolism, inflammation, and hepatic fibrosis in non-alcoholic fatty liver disease (NAFLD) [ 31 ]. Higher HMOX1 (Heme Oxygenase 1) levels were strongly associated with increased liver fat, which may be owing to its role in metabolic stress response. This could be either a reaction to liver fat-induced oxidative stress or to the products of heme degradation (e.g., iron) that exacerbate lipid peroxidation, contributing to fat storage and inflammation in the liver. Plasma proteins CTSD, IGFBP2, HMOX1, GUSB (β-glucuronidase), LPL (Lipoprotein Lipase), ALDH1A1 (Aldehyde Dehydrogenase 1 Family Member A1) and AGRP (Agouti-related protein) were directly linked to liver fat at different strength and directional probability levels (Table S1 and Figure S2). Metabolic Bayesian network in IMI-DIRECT Diabetes cohort To identify causal networks specific to individuals with T2D, we conducted Bayesian network analyses within the T2D cohort of the IMI-DIRECT study (n = 331) in a similar manner as before (Figure S3, Table 2 , Figure 3 ). In alignment with the non-diabetes network, BasalISR emerged as a putative upstream driver influencing liver fat accumulation with the FGF21 (fibroblast growth factor 21) protein as a downstream target of liver fat. FGF21, predominantly secreted by the liver, is a key modulator of hepatic glucose and lipid metabolism. Elevated circulating FGF21 levels have been reported in patients with NAFLD [ 33 ], possibly reflecting a compensatory response to hepatic lipid accumulation. This response may promote lipolysis, enhance fatty acid oxidation, and improve insulin sensitivity. However, in the context T2D, the effectiveness of this compensatory mechanism may be diminished due to pre-existing metabolic dysregulation. Download figure Open in new tab Figure 3. Bayesian network of metabolic and proteomic interactions in the IMI-DIRECT type 2 diabetes cohort (n = 331). The graph displays directed relationships among clinical and proteomic variables. Nodes are color-coded: blue (clinical/metabolic), green (proteins), peach (liver fat as outcome). Solid arrows represent directed associations with high confidence (strength and direction probability ≥ 0.8), while dashed arrows indicate less confident directionality. APOM: apolipoprotein M; BasalISR: basal insulin secretion rate; CD4: cluster of differentiation 4; Clins: mean insulin clearance during OGTT; Clinsb: basal insulin clearance; CTRC: chymotrypsin C; FGF21: fibroblast growth factor 21; Glucagonmin0: fasting glucagon; Glucose: fasting plasma glucose; GlucoseSens: glucose sensitivity; HDL: high-density lipoprotein cholesterol; HOMA_IR: homeostatic model assessment of insulin resistance; Insulin: fasting plasma insulin; LDLR: low-density lipoprotein receptor; LiverFat: hepatic fat content; LPL: lipoprotein lipase; MATN2: matrilin-2; MFGE8: milk fat globule-EGF factor 8; OGIS: oral glucose insulin sensitivity index according to the method of Mari et al. [ 32 ]; PancFat: pancreas fat; PON3: paraoxonase 3; SAT: subcutaneous adipose tissue; TG: triglycerides; TGM2: transglutaminase 2; TwoGlucose: 2-hour post-load glucose (OGTT); TwoInsulin: 2-hour post-load insulin (OGTT); VAT: visceral adipose tissue. View this table: View inline View popup Download powerpoint Table 2. Conditional regression models for each node in the averaged Bayesian network (type 2 diabetes cohort, n = 331). Each row lists the dependent variable (Outcome) and its set of parent variables (Exposures), along with coefficient estimates and standard error (SE). All variables were inverse-normal transformed. ACE2 (angiotensin-converting enzyme 2) was also directly linked to liver fat, albeit with a weaker association that excluded it from the parsimonious network (Figure S4). As a component of the renin-angiotensin system (RAS), ACE2 mitigates the pro-inflammatory and pro-fibrotic effects of angiotensin II, and its elevation in fatty liver disease may reflect a protective, compensatory adaptation to lipid overload and oxidative stress [ 34 ]. Notably, increased hepatic ACE2 expression has been implicated in facilitating SARS-CoV-2 entry, potentially contributing to liver injury and adverse outcomes in individuals with diabetes [ 35 ]. Additionally, both VAT and HbA1c were strongly associated with liver fat, although no definitive causal direction could be established (shown as dashed lines in Figure 3 ). Detailed information of all arcs is reported in Table S1 and Figure S4. Complementary directional inference with Mendelian randomization For some of the arcs, although the strength of the connection was high, the direction of the connection could not be determined with high probability (dashed lines in Figures 2 and 3 ). This suggests either a bidirectional association or insufficient data to confidently determine causal direction. To investigate this, we performed two-sample MR analysis for connections from clinical variables to the proteins and SMR-HEIDI to study the association from proteins to the clinical traits. The results are reported in Table 3 . View this table: View inline View popup Download powerpoint Table 3. Mendelian Randomization (MR) Analysis to Determine Directionality of Associations Between Clinical Variables and Proteins. This table presents the results of bidirectional MR and SMR-HEIDI analyses used to infer the direction of causal relationships between clinical traits and proteins for which Bayesian networks indicated strong associations but uncertain directionality ( Figures 2 and 3 ). Highlighted rows represent statistically significant findings after correction, with effect estimates, standard errors, and P-values shown for both forward (clinical → protein) and reverse (protein → clinical) directions. Colour shading indicates the direction with stronger evidence: peach for positive beta estimates, and blue for negative beta estimates. We identified potential positive causal associations from high-density lipoprotein (HDL) to HMOX1, APOM (apolipoprotein M), and KITLG (KIT ligand) proteins. HDL is well-known for its anti-inflammatory and antioxidant properties. The observed association with HMOX1, a stress-response enzyme involved in heme degradation and cytoprotection, may reflect HDL’s role in mitigating oxidative stress through upregulation of HMOX1. APOM, predominantly associated with HDL particles, plays a role in lipid metabolism and endothelial function, so a causal link from HDL aligns with its known role as a carrier protein that influences APOM stability and function [ 36 ]. The association with KITLG, a cytokine involved in haematopoiesis and cell survival, may reflect HDL’s regulatory effects on inflammatory and hematopoietic signalling pathways [ 37 ]. Triglycerides (TG) also appeared to have positive causal effects on LDLR (low-density lipoprotein receptor) and AGRP (agouti-related peptide) protein levels. Elevated TG levels can lead to changes in lipoprotein metabolism, which may upregulate LDLR expression to enhance lipid clearance from circulation [ 38 ]. The link to AGRP, a neuropeptide that promotes appetite and energy intake, may reflect a compensatory mechanism where increased TG levels signal energy abundance, thereby stimulating AGRP expression as part of a homeostatic feedback loop in energy balance. Liver fat showed a positive causal effect on HMOX1 and CTSD. Increased hepatic lipid accumulation can induce oxidative stress and inflammation, which may trigger upregulation of HMOX1 as a protective response [ 39 ]. CTSD, a lysosomal protease involved in protein degradation and lipid metabolism, may be upregulated in response to liver fat to support intracellular lipid turnover and mitigate lipotoxicity [ 31 ]. HbA1c and pancreas fat showed negative causal effects on HMOX1 and CTRC (chymotrypsin C), respectively. Chronic hyperglycaemia, characterised here by elevated HbA1c, may suppress HMOX1 through mechanisms involving oxidative damage and altered transcriptional regulation. The negative association between pancreatic fat and CTRC, a pancreatic enzyme that regulates the activation of digestive enzymes, may be explained by fat infiltration inducing lipotoxicity. This, in turn, can stress and impair the function of acinar cells (the primary source of CTRC) potentially reducing its expression or secretion. MFGE8 (Milk fat globule-EGF factor 8 protein) and TG have bidirectional effects, where TG positively influences MFGE8 and MFGE8 negatively influences TG, suggest a feedback mechanism. Specifically, increased TG levels may upregulate MFGE8 expression to modulate inflammation and phagocytosis of apoptotic cells, which are elevated in metabolic dysfunction [ 40 ]. In turn, elevated MFGE8 may enhance lipid clearance or promote anti-inflammatory effects that reduce circulating TG levels, thus helping to restore metabolic balance. MFGE8 has also been implicated in promoting tissue repair and vascular homeostasis, which may be activated in response to lipid-induced damage. The complementary MR analyses helped assign directionality to ten associations identified through Bayesian network analysis that demonstrated strong associations but where directionality could not be confidently ascertained. Bayesian networks and posterior probabilistic inference in IMI-DIRECT Bayesian networks were constructed to model conditional dependencies and infer potential causal relationships among clinical and proteomic variables in relation to liver fat accumulation. Focusing on liver fat as the outcome of interest, we estimated posterior probabilities by conditioning not only on its direct parent nodes but also individually on each variable in the network. This comprehensive conditioning strategy enabled the identification of both direct and indirect influences on the likelihood of MASLD (defined as liver fat >5%). To facilitate this analysis, all continuous variables were discretized into low, moderate, and high levels using Hartemink’s method [ 41 ] (Table S1). In the non-diabetes network, the baseline probability of MASLD was 34%, which rose to 71% after conditioning on high BasalISR ( Figure 4 ). Proteomic data further revealed distinct modulators of MASLD risk ( Figure 5 ). In this group, the greatest increases (∼30%) in MASLD probability were observed after conditioning on high levels of GUSB and ALDH1A1. GUSB is involved in lysosomal degradation and has been implicated in inflammation and obesity-related metabolic dysfunction [ 3 , 42 ]; its upregulation may signal hepatic stress or immune activation. NAFLD-associated genetic variants have also been reported to correlate with plasma GUSB levels [ 3 ]. ALDH1A1 is a key enzyme in retinoid metabolism and lipid homeostasis and has been linked to hepatic lipid accumulation through its regulation of PPAR signalling [ 43 , 44 ]. Additionally, low levels of IGFBP1 and LPL were associated with increased MASLD risk (∼15%). IGFBP1 is negatively regulated by insulin and is typically suppressed in insulin-resistant states, serving as a surrogate marker of hepatic insulin action [ 45 ]. LPL hydrolyzes circulating triglycerides; reduced activity can impair lipid clearance, promoting ectopic fat deposition in the live [ 46 ]. Download figure Open in new tab Figure 4. Posterior probabilities of MASLD based on clinical predictors across IMI-DIRECT diabetes and non-diabetes cohorts. Radar plots display the posterior probability of MASLD (metabolic dysfunction-associated steatotic liver disease, defined as liver fat >5%) after conditioning on individual clinical variables in (Left) individuals with T2D (type 2 diabetes) and (Right) those without diabetes. Bars represent the conditional probability of MASLD given high (red bars) or low (blue bars) levels of each variable. MASLD: metabolic dysfunction-associated steatotic liver disease; T2D: type 2 diabetes; BasalISR: basal insulin secretion rate; VAT: visceral adipose tissue; SAT: subcutaneous adipose tissue; OGIS: oral glucose insulin sensitivity index; HOMA_IR: homeostatic model assessment of insulin resistance; Clinsb: basal insulin clearance; Clins: dynamic insulin clearance; TG: triglycerides; HDL: high-density lipoprotein cholesterol; HbA1c: glycated haemoglobin A1c; GLP1: glucagon-like peptide 1; OGTT: oral glucose tolerance test. Download figure Open in new tab Figure 5. Posterior probabilities of MASLD based on proteomic predictors in IMI-DIRECT diabetes and non-diabetes cohorts. Radar plots display the conditional probability of MASLD after conditioning on individual proteomic variables in (Left) T2D and (Right) non-diabetes cohorts. Bars indicate high (red) or low (blue) levels of the corresponding protein. MASLD: metabolic dysfunction-associated steatotic liver disease; T2D: type 2 diabetes; GUSB: β-glucuronidase; LEP: leptin; ALDH1A1: aldehyde dehydrogenase 1 family member A1; CTSD: cathepsin D; FGF21: fibroblast growth factor 21; IGFBP1/2: insulin-like growth factor binding proteins 1 and 2; LDLR: low-density lipoprotein receptor; MFGE8: milk fat globule-EGF factor 8; AGRP: agouti-related peptide; KITLG: KIT ligand; ACE2: angiotensin-converting enzyme 2; APOM: apolipoprotein M; CD4: cluster of differentiation 4; PON3: paraoxonase 3; TGM2: transglutaminase 2; MATN2: matrilin-2. In the diabetes network, the unconditioned MASLD probability was 58%, with similar increases (∼20%) observed after conditioning on high fasting insulin, BasalISR, or VAT, and low Clinsb, reinforcing the roles of hyperinsulinemia, insulin resistance, and visceral adiposity in hepatic steatosis. In this network, GUSB again demonstrated the strongest association (∼30%), further supporting its potential role in disease progression. High levels of FGF21, ALDH1A1, and CTSD (∼15%) also elevated MASLD probability. Across both the non-diabetes and diabetes networks, the most informative combination of markers predicting the highest MASLD risk included elevated BasalISR, VAT, GUSB, and ALDH1A1, alongside low levels of LPL and IGFBP1, highlighting a coordinated disruption of pathways related to insulin resistance, inflammation, and lipid metabolism. Sex-stratified analyses revealed subgroup-specific patterns (Figures S5 and S6). Among non-diabetic males (n = 793), ALDH1A1 and LEP (leptin) showed the strongest effects, while GUSB was most influential in non-diabetic females (n = 171). LEP is central to energy homeostasis and hepatic fat oxidation; its dysregulation may impair hepatic lipid handling, particularly in males [ 47 ]. Elevated LEP levels in fatty liver may indicate hepatic LEP resistance, where LEP fails to stimulate lipid turnover. The greater influence of GUSB in females is consistent with prior findings suggesting sex-specific regulation, potentially mediated by oestrogen and body composition. GUSB activity has been reported to be higher in overweight females (but not males), possibly reflecting hormonal fluctuations and systemic low-grade inflammation during reproductive transitions [ 48 ]. In the T2D cohort (n = 193 males, n = 138 females), LEP remained the most influential proteomic marker in males, while GUSB remained dominant in females. Among clinical variables, BasalISR and VAT were the strongest predictors of MASLD in females, underscoring the pivotal role of insulin secretion and visceral adiposity in hepatic lipid accumulation. In males, BasalISR was most predictive in the non-diabetic state, whereas subcutaneous adipose tissue (SAT) had greater impact in diabetes, highlighting sex-and disease-stage–specific differences in adipose distribution and metabolic regulation. Discussion This analysis focused on determining causal mechanisms underlying liver fat accumulation in adults with and without T2D using comprehensive clinical, imaging, and proteomic phenotyping in the IMI-DIRECT cohorts. We describe complex interactions between insulin dynamics, adipose distribution, and specific circulating proteins in MASLD risk. Several key findings replicated across cohorts, with some effects being specific to sex and disease state. One of the key replicated findings is that BasalISR is a primary causal driver of liver fat accumulation, with a stronger effect in individuals without diabetes, providing evidence that basal hyperinsulinemia is a causal driver of hepatic fat accumulation, rather than simply co-occurring or as a downstream effect. This aligns with prior mechanistic evidence indicating that insulin influences hepatic lipid metabolism via de novo lipogenesis, more through hormonal regulation (insulin levels) than because of substrate availability (glucose levels) [ 49 ]. The close association between excessive insulin secretion and metabolic dyshomeostasis has been shown in several studies, although the causal dynamics between hyperinsulinemia and metabolic dysfunction have hitherto lacked strong evidence [ 5 ]. There is, though, an established causal association of excess adiposity on basal insulin secretion rate, even in the absence of insulin resistance, with diet-restricted weight loss lowering insulin secretion rate without changing insulin sensitivity [ 50 ]. Our results concur with prior research showing that hyperinsulinemia is a modifiable risk factor for metabolic deterioration, even in people with normoglycemia, and that small changes in insulin levels can affect lipid accumulation and adipogenesis independent of glucose homeostasis [ 51 , 52 ]. The directionality of relationships between insulin secretion, insulin resistance, and liver fat is complex [ 53 ], with evidence supporting both insulin resistance driving hyperinsulinemia [ 54 ] and vice versa [ 49 , 55 ]. In our Bayesian networks, an arc from BasalISR to HOMA-IR and from OGIS (insulin sensitivity) to BasalISR was observed, reflecting the complexity of these effects. The Bayesian networks also uncovered a bidirectional relationship between liver fat and visceral fat, suggesting a self-perpetuating cycle of fat redistribution and metabolic stress. In the T2D cohort, while BasalISR remained important, VAT exerted similar causal influence on liver fat, likely reflecting the shift in disease drivers as insulin secretion capacity diminishes. Cross-talk between VAT and the liver, mediated by free fatty acids entering the portal circulation, likely underpins this relationship [ 7 ]. Additionally, a negative association between fasting insulin clearance and liver fat was observed, suggesting hepatic steatosis may impair insulin catabolism, exacerbating systemic hyperinsulinemia. Sex-specific differences in adipose distribution emerged: visceral fat was the strongest predictor of MASLD in females, while subcutaneous fat had a stronger influence in males with diabetes. These findings highlight the need for sex-and disease-stage-specific interventions that consider differences in fat compartment function and metabolic regulation. Proteomic analyses revealed 34 proteins associated with liver fat in the Bayesian networks, with 11 replicating across both diabetes and non-diabetes cohorts. Proteins such as GUSB, ALDH1A1, LPL, IGFBP1/2, CTSD, HMOX1, FGF21, AGRP, and LEP played central roles in modulating MASLD risk. GUSB, a lysosomal enzyme linked to inflammation and cellular stress, showed the strongest association with MASLD in both cohorts and was particularly important in females. ALDH1A1, a key regulator of retinoid metabolism and lipid homeostasis, had strong effects in males, and was consistently associated with MASLD risk across metabolic states. LPL and IGFBP1, inversely associated with MASLD, reflect disturbances in lipid clearance and hepatic insulin sensitivity, respectively. Their depletion likely contributes to ectopic fat accumulation and systemic insulin resistance. CTSD and HMOX1, both enzymes involved in oxidative and lysosomal stress responses, were upregulated in relation to liver fat, suggesting a compensatory reaction to lipotoxicity. FGF21, particularly in T2D, appeared as a stress-induced hepatokine, reinforcing its role as a compensatory factor in advanced metabolic dysfunction. Sex-stratified Bayesian inference further revealed that LEP and ALDH1A1 were more influential in non-diabetic men, while GUSB dominated in females, underscoring sex-specific immune-metabolic mechanisms of liver fat regulation. By estimating posterior probabilities conditioned on single variables and combinations of key markers, we were able to quantify how specific features modify MASLD risk. The greatest increases in MASLD probability were observed following conditioning on combinations of high BasalISR, VAT, GUSB, ALDH1A1and low IGFBP1 and LPL, demonstrating the additive influence of insulin dynamics and proteomic factors. These probabilistic models highlight the clinical utility of combining hormonal and proteomic data to stratify populations by MASLD risk. Where causal effect directionality was unclear in the Bayesian networks, we applied bidirectional Mendelian randomization. This analysis validated ten associations and helped assign directionality: HDL to HMOX1, APOM, KITLG: reflecting HDL’s antioxidant and signalling roles. TG to LDLR, AGRP: suggesting lipid-induced shifts in receptor expression and appetite regulation. Liver fat to HMOX1, CTSD: supporting a model where hepatic lipid accumulation triggers oxidative and lysosomal stress responses. HbA1c and pancreas fat lowering HMOX1 and CTRC: indicating chronic hyperglycaemia and fat infiltration suppress protective or digestive pathways. A feedback loop between TG and MFGE8 was observed, suggesting reciprocal regulation between lipids and anti-inflammatory responses. These findings enhance the interpretability of our network results and help delineate causal vs. consequential biological processes in MASLD. The integration of multiple phenotypic layers (including clinical, imaging, metabolic, and proteomic data) within the causal modelling framework deployed here provides robust inference of directionality and mechanistic insight into complex metabolic interactions. Despite these strengths, we should also highlight several key limitations: The cohort consisted solely of individuals of European ancestry, limiting generalizability. Furthermore, MR assumptions may not hold for all proteins due to pleiotropy or weak instruments, and Bayesian inference depends on the completeness and quality of included variables and assumptions. Although several key findings were replicated, it is important to stress that the differences between the two cohorts (i.e., one is of people without and the other with diabetes at enrolment) means that some or all the findings that did not replicate may simply reflect disease-state-specific effects. Disentangling this is challenging, as no comparable cohorts are available for replication. In summary, this study provides a comprehensive systems-level analysis of MASLD pathogenesis. We show that basal insulin hypersecretion is a modifiable, causal driver of liver fat accumulation, particularly in individuals without diabetes. Through probabilistic inference and genetic validation, we demonstrate that insulin dynamics, visceral fat, and circulating proteins such as GUSB and ALDH1A1 interact in sex-and disease-specific ways to influence liver fat risk. These insights can inform precision medicine approaches to MASLD prevention, including targeted metabolic modulation in high-risk individuals and the potential use of plasma proteomics for risk stratification. Future research should examine the longitudinal impact of modifying these pathways and extend analyses to more diverse populations. ETHICS STATEMENTS Approval for the study protocol was obtained from each of the regional research ethics review boards separately (Lund, Sweden: 20130312105459927, Copenhagen, Denmark: H-1-2012-166 and H-1-2012-100, Amsterdam, Netherlands: NL40099.029.12, Newcastle, Dundee and Exeter, UK: 12/NE/0132). All participants provided written informed consent at enrolment. The research conformed to the ethical principles for medical research involving human participants outlined in the Declaration of Helsinki. Data Availability All data produced in the present study are available upon reasonable request to the authors DECLARATION OF INTERESTS AD works for Novo Nordisk Research Centre Oxford. SB has ownerships in Intomics A/S, Hoba Therapeutics Aps, Novo Nordisk A/S, Lundbeck A/S, and managing board memberships in Proscion A/S and Intomics A/S. MR owns stock in Novo Nordisk A/S. MMcC is an employee of Genentech and a holder of Roche stock. Within the past five years, PWF has received consulting honoraria from Eli Lilly Inc., Novo Nordisk Foundation, Novo Nordisk A/S, UBS, Qatar Foundation, and Zoe Ltd. PWF was also an employee of the Novo Nordisk Foundation (2021-2024). PWF has also received investigator-initiated grants (paid to institution) from numerous pharmaceutical companies as part of the Innovative Medicines Initiative of the European Union. Download figure Open in new tab Figure S1. Bayesian Information Criterion (BIC) comparison for non-diabetes network model validation (n=964) Histogram of BIC scores from 1,000 randomly generated null networks (gray bars) with a smoothed density overlay (blue line). The red dashed line shows the BIC of the learned network (BIC =-34854.71), indicating significantly better fit than expected by chance. Download figure Open in new tab Figure S2. Expanded Bayesian network of metabolic and proteomic interactions in the IMI-DIRECT non-diabetes cohort (n = 964). This detailed graph shows all directed relationships (arcs with strength ≥ 0.5) among clinical and proteomic variables. Line thickness reflects the strength of association. Nodes are color-coded as follows: blue for clinical/metabolic variables, green for proteins, and peach for liver fat (outcome). Solid arrows represent directed associations with high confidence (strength and directional probability ≥ 0.8), while dashed arrows indicate less confident directionality. AGRP: agouti-related peptide; ALDH1A1: aldehyde dehydrogenase 1 family member A1; APOM: apolipoprotein M; BasalISR: basal insulin secretion rate at the beginning of the OGTT; CD4: cluster of differentiation 4; CDH5: cadherin-5; Clins: mean insulin clearance during OGTT; Clinsb: basal insulin clearance-calculated as (mean insulin secretion)/(mean insulin concentration); CTRC: chymotrypsin C; CTSD: cathepsin D; FGF21: fibroblast growth factor 21; Glucagonmin0: fasting glucagon; Glucose: fasting plasma glucose; GlucoseSens: glucose sensitivity; GUSB: β-glucuronidase; HbA1c: glycated haemoglobin A1C; HDL: high-density lipoprotein cholesterol; IGFBP1/2: insulin-like growth factor binding proteins 1 and 2; Insulin: fasting plasma insulin; KITLG: KIT ligand; LDLR: low-density lipoprotein receptor; LEP: leptin; LiverFat: hepatic fat content; LPL: lipoprotein lipase; MFGE8: milk fat globule-epidermal growth factor 8; OGIS: oral glucose insulin sensitivity index according to the method of Mari et al. [ 32 ]; PancFat: pancreas fat; PON3: paraoxonase 3; SAT: subcutaneous adipose tissue; TG: triglycerides; TotGLP1min0: fasting total GLP-1; TwoGlucose/TwoInsulin: 2-hour post-load values from OGTT; VAT: visceral adipose tissue. Download figure Open in new tab Figure S3. Bayesian Information Criterion (BIC) comparison for diabetes network model validation (n=331) Histogram of BIC scores from 1,000 randomly generated null networks (gray bars) with a smoothed density overlay (blue line). The red dashed line shows the BIC of the learned network (BIC =-30286.48), indicating significantly better fit than expected by chance. Download figure Open in new tab Figure S4. Expanded Bayesian network of metabolic and proteomic interactions in the IMI-type 2 diabetes cohort (n = 331). This detailed graph shows all directed relationships (arcs with strength ≥ 0.5) among clinical and proteomic variables. Line thickness reflects the strength of association. Nodes are color-coded as follows: blue for clinical/metabolic variables, green for proteins, and peach for liver fat (outcome). Solid arrows represent directed associations with high confidence (strength and directional probability ≥ 0.8), while dashed arrows indicate less confident directionality. ACE2: angiotensin-converting enzyme 2; AGRP: agouti-related peptide; ALDH1A1: aldehyde dehydrogenase 1 family member A1; ANXA4: annexin A4; APOM: apolipoprotein M; BasalISR: basal insulin secretion rate; CD4: cluster of differentiation 4; Clins: mean insulin clearance during OGTT; Clinsb: basal insulin clearance; CTRC: chymotrypsin C; FGF21: fibroblast growth factor 21; FAM3C: family with sequence similarity 3 member C; Glucagonmin0: fasting glucagon; Glucose: fasting plasma glucose; GlucoseSens: glucose sensitivity; HDL: high-density lipoprotein cholesterol; HOMA_IR: homeostatic model assessment of insulin resistance; IGFBP1: insulin-like growth factor binding protein 1; Insulin: fasting plasma insulin; KITLG: KIT ligand; LDLR: low-density lipoprotein receptor; LEP: leptin; LiverFat: hepatic fat content; LPL: lipoprotein lipase; MATN2: matrilin-2; MFGE8: milk fat globule-EGF factor 8; NRP1: neuropilin-1; OGIS: oral glucose insulin sensitivity index according to the method of Mari et al. [ 32 ]; PancFat: pancreas fat; PON3: paraoxonase 3; SAT: subcutaneous adipose tissue; TG: triglycerides; TGM2: transglutaminase 2; TotGLP1min0: fasting total GLP-1; TwoGlucose: 2-hour post-load glucose (OGTT); TwoInsulin: 2-hour post-load insulin (OGTT); VAT: visceral adipose tissue. Download figure Open in new tab Figure S5. Sex-stratified Posterior probabilities of MASLD based on clinical predictors across IMI-DIRECT diabetes and non-diabetes cohorts. Radar plots display MASLD conditional probabilities following conditioning on individual clinical variables, stratified by sex and diabetes status: non-diabetic females, n=171 (top left), non-diabetic males, n=793 (top right), diabetic females, n=139 (bottom left), and diabetic males, n=193 (bottom right). Red bars indicate high levels and blue bars low levels of the respective variable. MASLD: metabolic dysfunction-associated steatotic liver disease; T2D: type 2 diabetes; BasalISR: basal insulin secretion rate; VAT: visceral adipose tissue; SAT: subcutaneous adipose tissue; OGIS: oral glucose insulin sensitivity index; HOMA_IR: homeostatic model assessment of insulin resistance; Clinsb: basal insulin clearance; Clins: dynamic insulin clearance; TG: triglycerides; HDL: high-density lipoprotein cholesterol; HbA1c: glycated haemoglobin A1c; GLP1: glucagon-like peptide 1; OGTT: oral glucose tolerance test. Download figure Open in new tab Figure S6. Sex-stratified Posterior probabilities of MASLD based on proteomic predictors in IMI-DIRECT diabetes and non-diabetes cohorts. Radar plots display MASLD conditional probabilities after conditioning on proteomic variables, stratified by sex and diabetes status: non-diabetic females, n=171 (top left), non-diabetic males, n=793 (top right), diabetic females, n=138 (bottom left), and diabetic males, n=193 (bottom right). Red bars show the conditional probability given high levels, and blue bars indicate conditioning on low levels of each protein. MASLD: metabolic dysfunction-associated steatotic liver disease; T2D: type 2 diabetes; GUSB: β-glucuronidase; LEP: leptin; ALDH1A1: aldehyde dehydrogenase 1 family member A1; CTSD: cathepsin D; FGF21: fibroblast growth factor 21; IGFBP1/2: insulin-like growth factor binding proteins 1 and 2; LDLR: low-density lipoprotein receptor; MFGE8: milk fat globule-EGF factor 8; AGRP: agouti-related peptide; KITLG: KIT ligand; ACE2: angiotensin-converting enzyme 2; APOM: apolipoprotein M; CD4: cluster of differentiation 4; PON3: paraoxonase 3; TGM2: transglutaminase 2; MATN2: matrilin-2. ACKNOWLEDGEMENTS The work leading to this publication has received support from the Innovative Medicines Initiative Joint. Undertaking under grant agreement n°115317 (DIRECT, https://directdiabetes.org/ ), resources of which are composed of financial contribution from the European Union’s Seventh Framework Programme (FP7/2007-2013) and EFPIA companies’ in-kind contribution. We thank all the participants and study centre staff in IMI DIRECT for their contribution to the study. NNA was funded by the Swedish Research Council (VR - Avtals-ID: 2021-06714_3) for her postdoctoral fellowship (2022-2025) and acknowledges support from the Henning och Johan Throne-Holsts Foundation. NNA, RJFL and JM are supported by the Novo Nordisk Foundation grant NNF23SA0084103. R.J.F.L. is supported by grants from the National Institutes of Health (R01DK107786; R01DK110113; R01HG010297; R01HL142302; R01DK075787; R01HL156991; U01HG011723; R01DK123019; R01HL151152; R01HL158884), the Danish National Research Fund (DNRF161) and the Novo Nordisk Foundation (NNF20OC0059313). JM acknowledges awards from EFSD/Novo Nordisk Foundation Future Leaders Award (no. 0094134) and the European Union (HORIZON-EIC-2023-PATHFINDERCHALLENGES-01-101161509). Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or European Innovation Council and SMEs Executive Agency (EISMEA). Neither the European Union nor the granting authority can be held responsible for them. PWF was supported by grants from the Swedish Research Council (#2019-01348), Swedish Foundation for Strategic Research (LUDC-IRC, 15-0067), and the European Commission (ERC-CoG_NASCENT - 681742). References 1. ↵ Younossi ZM , Golabi P , Paik JM , Henry A , Van Dongen C , Henry L . The global epidemiology of nonalcoholic fatty liver disease (NAFLD) and nonalcoholic steatohepatitis (NASH): a systematic review . Hepatology . 2023 ; 77 ( 4 ): 1335 – 47 . OpenUrl CrossRef PubMed 2. ↵ En Li Cho E , Ang CZ , Quek J , Fu CE , Lim LKE , Heng ZEQ , et al. Global prevalence of non-alcoholic fatty liver disease in type 2 diabetes mellitus: an updated systematic review and meta-analysis . Gut . 2023 ; 72 ( 11 ): 2138 – 48 . OpenUrl Abstract / FREE Full Text 3. ↵ Sveinbjornsson G , Ulfarsson MO , Thorolfsdottir RB , Jonsson BA , Einarsson E , Gunnlaugsson G , et al. Multiomics study of nonalcoholic fatty liver disease . Nat Genet . 2022 ; 54 ( 11 ): 1652 – 63 . OpenUrl CrossRef PubMed 4. ↵ Hyogo H , Yamagishi S , Maeda S , Kimura Y , Ishitobi T , Chayama K . Increased insulinogenic index is an independent determinant of nonalcoholic fatty liver disease activity score in patients with normal glucose tolerance . Dig Liver Dis . 2012 ; 44 ( 11 ): 935 – 9 . OpenUrl PubMed 5. ↵ Zhang AMY , Wellberg EA , Kopp JL , Johnson JD. Hyperinsulinemia in Obesity, Inflammation, and Cancer . Diabetes Metab J . 2021 ; 45 ( 4 ): 622 . OpenUrl CrossRef PubMed 6. ↵ Chai SY , Pan XY , Song KX , Huang YY , Li F , Cheng XY , et al. Differential patterns of insulin secretion and sensitivity in patients with type 2 diabetes mellitus and nonalcoholic fatty liver disease versus patients with type 2 diabetes mellitus alone . Lipids Health Dis . 2014 ; 13 : 7 . 7. ↵ Scoditti E , Sabatini S , Carli F , Gastaldelli A . Hepatic glucose metabolism in the steatotic liver . Nat Rev Gastroenterol Hepatol . 2024 . 8. Utzschneider KM , Kahn SE . Review: The role of insulin resistance in nonalcoholic fatty liver disease . J Clin Endocrinol Metab . 2006 ; 91 ( 12 ): 4753 – 61 . OpenUrl CrossRef PubMed Web of Science 9. Liu Z , Zhang Y , Graham S , Wang X , Cai D , Huang M , et al. Causal relationships between NAFLD, T2D and obesity have implications for disease subphenotyping . J Hepatol . 2020 ; 73 ( 2 ): 263 – 76 . OpenUrl CrossRef PubMed 10. ↵ Taylor R . Type 2 diabetes: etiology and reversibility . Diabetes Care . 2013 ; 36 ( 4 ): 1047 – 55 . OpenUrl FREE Full Text 11. ↵ Boel F , Akimov V , Teuchler M , Terkelsen MK , Wernberg CW , Larsen FT , et al. Deep proteome profiling of metabolic dysfunction-associated steatotic liver disease . Commun Med (Lond ). 2025 ; 5 ( 1 ): 56 . OpenUrl PubMed 12. Govaere O , Hasoon M , Alexander L , Cockell S , Tiniakos D , Ekstedt M , et al. A proteo-transcriptomic map of non-alcoholic fatty liver disease signatures . Nat Metab . 2023 ; 5 ( 4 ): 572 – 8 . OpenUrl PubMed 13. ↵ Loesch DP , Garg M , Matelska D , Vitsios D , Jiang X , Ritchie SC , et al. Identification of plasma proteomic markers underlying polygenic risk of type 2 diabetes and related comorbidities . Nat Commun . 2025 ; 16 ( 1 ): 2124 . OpenUrl PubMed 14. ↵ Koivula RW , Heggie A , Barnett A , Cederberg H , Hansen TH , Koopman AD , et al. Discovery of biomarkers for glycaemic deterioration before and after the onset of type 2 diabetes: rationale and design of the epidemiological studies within the IMI DIRECT Consortium . Diabetologia . 2014 ; 57 ( 6 ): 1132 – 42 . OpenUrl CrossRef PubMed 15. ↵ Koivula RW , Forgie IM , Kurbasic A , Vinuela A , Heggie A , Giordano GN , et al. Discovery of biomarkers for glycaemic deterioration before and after the onset of type 2 diabetes: descriptive characteristics of the epidemiological studies within the IMI DIRECT Consortium . Diabetologia . 2019 ; 62 ( 9 ): 1601 – 15 . OpenUrl PubMed 16. ↵ Van Cauter E , Mestrez F , Sturis J , Polonsky KS . Estimation of insulin secretion rates from C-peptide levels. Comparison of individual and standard kinetic parameters for C-peptide clearance . Diabetes . 1992 ; 41 ( 3 ): 368 – 77 . OpenUrl Abstract / FREE Full Text 17. ↵ Thomas EL , Fitzpatrick JA , Malik SJ , Taylor-Robinson SD , Bell JD . Whole body fat: content and distribution . Prog Nucl Magn Reson Spectrosc . 2013 ; 73 : 56 – 80 . OpenUrl CrossRef PubMed 18. ↵ Assarsson E , Lundberg M , Holmquist G , Bjorkesten J , Thorsen SB , Ekman D , et al. Homogenous 96-plex PEA immunoassay exhibiting high sensitivity, specificity, and excellent scalability . PLoS One . 2014 ; 9 ( 4 ): e95192 . OpenUrl CrossRef PubMed 19. ↵ Scutari M , Howell P , Balding DJ , Mackay I . Multiple quantitative trait analysis using bayesian networks . Genetics . 2014 ; 198 ( 1 ): 129 – 37 . OpenUrl Abstract / FREE Full Text 20. ↵ Davey Smith G , Hemani G . Mendelian randomization: genetic anchors for causal inference in epidemiological studies . Hum Mol Genet . 2014 ; 23 ( R1 ): R89 – 98 . OpenUrl CrossRef PubMed Web of Science 21. ↵ Zhu Z , Zhang F , Hu H , Bakshi A , Robinson MR , Powell JE , et al. Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets . Nat Genet . 2016 ; 48 ( 5 ): 481 – 7 . OpenUrl CrossRef PubMed 22. ↵ Scutari MD , J-B . Bayesian Networks With Examples in R . New York : Chapman and Hall/CRC ; 2021 . 23. ↵ Hemani G , Zheng J , Elsworth B , Wade KH , Haberland V , Baird D , et al. The MR-Base platform supports systematic causal inference across the human phenome . Elife . 2018 ; 7 . 24. ↵ Shannon P , Markiel A , Ozier O , Baliga NS , Wang JT , Ramage D , et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks . Genome Res . 2003 ; 13 ( 11 ): 2498 – 504 . OpenUrl Abstract / FREE Full Text 25. ↵ Atabaki-Pasdar N , Ohlsson M , Vinuela A , Frau F , Pomares-Millan H , Haid M , et al. Predicting and elucidating the etiology of fatty liver disease: A machine learning modeling and validation study in the IMI DIRECT cohorts . PLoS Med . 2020 ; 17 ( 6 ): e1003149 . OpenUrl CrossRef PubMed 26. ↵ Leite NC , Villela-Nogueira CA , Cardoso CR , Salles GF . Non-alcoholic fatty liver disease and diabetes: from physiopathological interplay to diagnosis and treatment . World J Gastroenterol . 2014 ; 20 ( 26 ): 8377 – 92 . OpenUrl CrossRef PubMed 27. ↵ Watanabe K , Taskesen E , van Bochoven A , Posthuma D . Functional mapping and annotation of genetic associations with FUMA . Nat Commun . 2017 ; 8 ( 1 ): 1826 . OpenUrl CrossRef PubMed 28. Ashburner M , Ball CA , Blake JA , Botstein D , Butler H , Cherry JM , et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium . Nat Genet . 2000 ; 25 ( 1 ): 25 – 9 . OpenUrl CrossRef PubMed Web of Science 29. Gene Ontology C , Aleksander SA , Balhoff J , Carbon S , Cherry JM , Drabkin HJ , et al. The Gene Ontology knowledgebase in 2023 . Genetics . 2023 ; 224 ( 1 ). 30. ↵ Milacic M , Beavers D , Conley P , Gong C , Gillespie M , Griss J , et al. The Reactome Pathway Knowledgebase 2024 . Nucleic Acids Res . 2024 ; 52 ( D1 ): D672 – D8 . OpenUrl CrossRef PubMed 31. ↵ Ding L , Goossens GH , Oligschlaeger Y , Houben T , Blaak EE , Shiri-Sverdlov R . Plasma cathepsin D activity is negatively associated with hepatic insulin sensitivity in overweight and obese humans . Diabetologia . 2020 ; 63 ( 2 ): 374 – 84 . OpenUrl CrossRef PubMed 32. ↵ Mari A , Pacini G , Murphy E , Ludvik B , Nolan JJ . A model-based method for assessing insulin sensitivity from the oral glucose tolerance test . Diabetes Care . 2001 ; 24 ( 3 ): 539 – 48 . OpenUrl Abstract / FREE Full Text 33. ↵ Tucker B , Li H , Long X , Rye KA , Ong KL . Fibroblast growth factor 21 in non-alcoholic fatty liver disease . Metabolism . 2019 ; 101 : 153994 . 34. ↵ Soldo J , Heni M , Konigsrainer A , Haring HU , Birkenfeld AL , Peter A . Increased Hepatic ACE2 Expression in NAFL and Diabetes-A Risk for COVID-19 Patients? Diabetes Care . 2020 ; 43 ( 10 ): e134 – e6 . OpenUrl FREE Full Text 35. ↵ Cano L , Desquilles L , Ghukasyan G , Angenard G , Landreau C , Corlu A , et al. SARS-CoV-2 receptor ACE2 is upregulated by fatty acids in human MASH . JHEP Rep . 2024 ; 6 ( 1 ): 100936 . OpenUrl PubMed 36. ↵ Borup A , Christensen PM , Nielsen LB , Christoffersen C . Apolipoprotein M in lipid metabolism and cardiometabolic diseases . Curr Opin Lipidol . 2015 ; 26 ( 1 ): 48 – 55 . OpenUrl PubMed 37. ↵ Sethumadhavan A , Mani M . Kit activates interleukin-4 receptor and effector signal transducer and activator of transcription 6 independent of its cognate ligand in mouse mast cells . Immunology . 2020 ; 159 ( 4 ): 441 – 9 . OpenUrl 38. ↵ Palmisano BT , Yu S , Neuman JC , Zhu L , Luu T , Stafford JM . Low-density lipoprotein receptor is required for cholesteryl ester transfer protein to regulate triglyceride metabolism in both male and female mice . Physiol Rep . 2021 ; 9 ( 4 ): e14732 . OpenUrl PubMed 39. ↵ Malaguarnera L , Madeddu R , Palio E , Arena N , Malaguarnera M . Heme oxygenase-1 levels and oxidative stress-related parameters in non-alcoholic fatty liver disease patients . J Hepatol . 2005 ; 42 ( 4 ): 585 – 91 . OpenUrl CrossRef PubMed Web of Science 40. ↵ Rout M , Malone-Perez MW , Park G , Lerner M , Kimble Frazer J , Apple B , et al. Contribution of circulating Mfge8 to human T2DM and cardiovascular disease . Gene . 2024 ; 927 : 148712 . 41. ↵ Hartemink AJ. Principled computational methods for the validation discovery of genetic regulatory networks : Massachusetts Institute of Technology ; 2001 . 42. ↵ Basinska A , Florianczyk B . Beta-glucuronidase in physiology and disease . Ann Univ Mariae Curie Sklodowska Med . 2003 ; 58 ( 2 ): 386 – 9 . OpenUrl PubMed 43. ↵ Kiefer FW , Vernochet C , O’Brien P , Spoerl S , Brown JD , Nallamshetty S , et al. Retinaldehyde dehydrogenase 1 regulates a thermogenic program in white adipose tissue . Nat Med . 2012 ; 18 ( 6 ): 918 – 25 . OpenUrl CrossRef PubMed 44. ↵ Paik J , Kim A , Fogassy K , Snyder JM , Brabb T , Dill-McFarland KA , et al. Weight loss and metabolic effects of an ALDH1A1-specific inhibitor, FSI-TN42, in a diet induced mouse model of obesity . Int J Obes (Lond) . 2025 ; 49 ( 3 ): 507 – 15 . OpenUrl PubMed 45. ↵ Stanley TL , Fourman LT , Zheng I , McClure CM , Feldpausch MN , Torriani M , et al. Relationship of IGF-1 and IGF-Binding Proteins to Disease Severity and Glycemia in Nonalcoholic Fatty Liver Disease . J Clin Endocrinol Metab . 2021 ; 106 ( 2 ): e520 – e33 . OpenUrl CrossRef PubMed 46. ↵ Shang R , Rodrigues B . Lipoprotein lipase as a target for obesity/diabetes related cardiovascular disease . J Pharm Pharm Sci . 2024 ; 27 : 13199 . 47. ↵ Polyzos SA , Kountouras J , Mantzoros CS . Leptin in nonalcoholic fatty liver disease: a narrative review . Metabolism . 2015 ; 64 ( 1 ): 60 – 78 . OpenUrl PubMed 48. ↵ Funk JL , Wertheim BC , Frye JB , Blew RM , Nicholas JS , Chen Z , et al. Association of ss-glucuronidase activity with menopausal status, ethnicity, adiposity, and inflammation in women . Menopause . 2023 ; 30 ( 2 ): 186 – 92 . OpenUrl PubMed 49. ↵ Johnson JD . On the causal relationships between hyperinsulinaemia, insulin resistance, obesity and dysglycaemia in type 2 diabetes . Diabetologia . 2021 ; 64 ( 10 ): 2138 – 46 . OpenUrl CrossRef PubMed 50. ↵ van Vliet S , Koh HE , Patterson BW , Yoshino M , LaForest R , Gropler RJ , et al. Obesity Is Associated With Increased Basal and Postprandial beta-Cell Insulin Secretion Even in the Absence of Insulin Resistance . Diabetes . 2020 ; 69 ( 10 ): 2112 – 9 . OpenUrl CrossRef PubMed 51. ↵ Dankner R , Chetrit A , Shanik MH , Raz I , Roth J . Basal state hyperinsulinemia in healthy normoglycemic adults heralds dysglycemia after more than two decades of follow up . Diabetes Metab Res Rev . 2012 ; 28 ( 7 ): 618 – 24 . OpenUrl CrossRef PubMed 52. ↵ Templeman NM , Skovso S , Page MM , Lim GE , Johnson JD . A causal role for hyperinsulinemia in obesity . J Endocrinol . 2017 ; 232 ( 3 ): R173 – R83 . OpenUrl Abstract / FREE Full Text 53. ↵ Czech MP . Insulin action and resistance in obesity and type 2 diabetes . Nat Med . 2017 ; 23 ( 7 ): 804 – 14 . OpenUrl CrossRef PubMed 54. ↵ Boucher J , Kleinridders A , Kahn CR . Insulin receptor signaling in normal and insulin-resistant states . Cold Spring Harb Perspect Biol . 2014 ; 6 ( 1 ). 55. ↵ Janssen J . Hyperinsulinemia and Its Pivotal Role in Aging, Obesity, Type 2 Diabetes, Cardiovascular Disease and Cancer . Int J Mol Sci . 2021 ; 22 ( 15 ). View the discussion thread. Back to top Previous Next Posted June 02, 2025. Download PDF Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following A Biological-Systems-Based Analyses Using Proteomic and Metabolic Network Inference Reveals Mechanistic Insights into Hepatic Lipid Accumulation: An IMI-DIRECT Study Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share A Biological-Systems-Based Analyses Using Proteomic and Metabolic Network Inference Reveals Mechanistic Insights into Hepatic Lipid Accumulation: An IMI-DIRECT Study Natalie N. Atabaki , Daniel E. Coral , Hugo Pomares-Millan , Kieran Smith , Harry H. Behjat , Robert W. Koivula , Andrea Tura , Hamish Miller , Katherine Pinnick , Leandro Agudelo , Kristine H. Allin , Andrew A. Brown , Elizaveta Chabanova , Piotr J Chmura , Ulrik P. Jacobsen , Adem Y. Dawed , Petra J.M. Elders , Juan J. Fernandez-Tajes , Ian M. Forgie , Mark Haid , Tue H. Hansen , Elizaveta L. Hansen , Angus G. Jones , Tarja Kokkola , Sebastian Kalamajski , Anubha Mahajan , Timothy J. McDonald , Donna McEvoy , Mirthe Muilwijk , Konstantinos D. Tsirigos , Jagadish Vangipurapu , Sabine van Oort , Henrik Vestergaard , Jerzy Adamski , Joline W. Beulens , Søren Brunak , Emmanouil T. Dermitzakis , Giuseppe N. Giordano , Ramneek Gupta , Torben Hansen , Leen t Hart , Andrew T. Hattersley , Leanne Hodson , Markku Laakso , Ruth J.F. Loos , Jordi Merino , Mattias Ohlsson , Oluf Pedersen , Martin Ridderstråle , Hartmut Ruetten , Femke Rutters , Jochen M. Schwenk , Jeremy Tomlinson , Mark Walker , Hanieh Yaghootkar , Fredrik Karpe , Mark I McCarthy , Elizabeth Louise Thomas , Jimmy D. Bell , Andrea Mari , Imre Pavo , Ewan R. Pearson , Ana Viñuela , Paul W. Franks medRxiv 2025.06.02.25328773; doi: https://doi.org/10.1101/2025.06.02.25328773 Share This Article: Copy Citation Tools A Biological-Systems-Based Analyses Using Proteomic and Metabolic Network Inference Reveals Mechanistic Insights into Hepatic Lipid Accumulation: An IMI-DIRECT Study Natalie N. Atabaki , Daniel E. Coral , Hugo Pomares-Millan , Kieran Smith , Harry H. Behjat , Robert W. Koivula , Andrea Tura , Hamish Miller , Katherine Pinnick , Leandro Agudelo , Kristine H. Allin , Andrew A. Brown , Elizaveta Chabanova , Piotr J Chmura , Ulrik P. Jacobsen , Adem Y. Dawed , Petra J.M. Elders , Juan J. Fernandez-Tajes , Ian M. Forgie , Mark Haid , Tue H. Hansen , Elizaveta L. Hansen , Angus G. Jones , Tarja Kokkola , Sebastian Kalamajski , Anubha Mahajan , Timothy J. McDonald , Donna McEvoy , Mirthe Muilwijk , Konstantinos D. Tsirigos , Jagadish Vangipurapu , Sabine van Oort , Henrik Vestergaard , Jerzy Adamski , Joline W. Beulens , Søren Brunak , Emmanouil T. Dermitzakis , Giuseppe N. Giordano , Ramneek Gupta , Torben Hansen , Leen t Hart , Andrew T. Hattersley , Leanne Hodson , Markku Laakso , Ruth J.F. Loos , Jordi Merino , Mattias Ohlsson , Oluf Pedersen , Martin Ridderstråle , Hartmut Ruetten , Femke Rutters , Jochen M. Schwenk , Jeremy Tomlinson , Mark Walker , Hanieh Yaghootkar , Fredrik Karpe , Mark I McCarthy , Elizabeth Louise Thomas , Jimmy D. Bell , Andrea Mari , Imre Pavo , Ewan R. Pearson , Ana Viñuela , Paul W. Franks medRxiv 2025.06.02.25328773; doi: https://doi.org/10.1101/2025.06.02.25328773 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Endocrinology (including Diabetes Mellitus and Metabolic Disease) Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4435) Dentistry and Oral Medicine (444) Dermatology (382) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1509) Epidemiology (15227) Forensic Medicine (30) Gastroenterology (1124) Genetic and Genomic Medicine (6597) Geriatric Medicine (668) Health Economics (997) Health Informatics (4534) Health Policy (1368) Health Systems and Quality Improvement (1613) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15916) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (146) Nephrology (667) Neurology (6599) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1144) Occupational and Environmental Health (957) Oncology (3332) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (691) Primary Care Research (711) Psychiatry and Clinical Psychology (5447) Public and Global Health (9230) Radiology and Imaging (2198) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a003db60f87f593a',t:'MTc3OTUzNzE3Mg=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();