Beyond the Genotype: A Multi-Omic Analysis of APOEe4’s Role in Alzheimer’s Disease

Beyond the Genotype: A Multi-Omic Analysis of APOEe4’s Role in Alzheimer’s Disease | bioRxiv /* */ /* */ <!-- <!-- /*! * 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-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Beyond the Genotype: A Multi-Omic Analysis of APOEe4’s Role in Alzheimer’s Disease View ORCID Profile Yaroslav Markov , Ahana Priyanka , View ORCID Profile Leqi Xu , Weiwei Wang , Kyra Thrush-Evensen , View ORCID Profile John Gonzalez , Daniel Borrus , Jessica Kasamoto , Raghav Sehgal , Grace Zou , Jenel Fraij , Becky C. Carlyle , View ORCID Profile Steve Horvath , David A Bennett , View ORCID Profile Hongyu Zhao , Christopher H. van Dyck , TuKiet T Lam , Morgan E. Levine , Albert T. Higgins-Chen doi: https://doi.org/10.1101/2025.10.16.682426 Yaroslav Markov 1 Program in Computational Biology and Biomedical Informatics, Yale University , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Yaroslav Markov For correspondence: yaroslav.markov{at}yale.edu Ahana Priyanka 2 Department of Computer Science and Engineering, Sri Sivasubramaniya Nadar College of Engineering , Chennai, India Find this author on Google Scholar Find this author on PubMed Search for this author on this site Leqi Xu 3 Department of Biostatistics, Yale School of Public Health , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Leqi Xu Weiwei Wang 4 Department of Molecular Biophysics and Biochemistry, Yale University , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Kyra Thrush-Evensen 1 Program in Computational Biology and Biomedical Informatics, Yale University , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site John Gonzalez 5 Department of Pathology, Yale School of Medicine , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for John Gonzalez Daniel Borrus 6 Department of Psychiatry, Yale School of Medicine , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jessica Kasamoto 1 Program in Computational Biology and Biomedical Informatics, Yale University , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Raghav Sehgal 1 Program in Computational Biology and Biomedical Informatics, Yale University , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Grace Zou 7 Department of Genetics, Yale School of Medicine , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jenel Fraij 1 Program in Computational Biology and Biomedical Informatics, Yale University , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Becky C. Carlyle 8 Department of Physiology, Anatomy & Genetics, University of Oxford , Oxford, UK 9 Kavli Institute for Nanoscience Discovery, University of Oxford , Oxford, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Steve Horvath 10 Department of Human Genetics, University of California , Los Angeles, CA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Steve Horvath David A Bennett 11 Rush Alzheimer’s Disease Center, Rush University Medical Center , Chicago, IL, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Hongyu Zhao 1 Program in Computational Biology and Biomedical Informatics, Yale University , New Haven, CT, USA 3 Department of Biostatistics, Yale School of Public Health , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Hongyu Zhao Christopher H. van Dyck 12 Alzheimer’s Disease Research Unit, Yale School of Medicine , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site TuKiet T Lam 4 Department of Molecular Biophysics and Biochemistry, Yale University , New Haven, CT, USA 13 Keck Mass Spectrometry & Proteomics Resource, Yale School of Medicine , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Morgan E. Levine 5 Department of Pathology, Yale School of Medicine , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Albert T. Higgins-Chen 6 Department of Psychiatry, Yale School of Medicine , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Abstract Full Text Info/History Metrics Supplementary material Preview PDF ABSTRACT Alzheimer’s disease (AD) is characterized by widespread molecular dysregulation, with the APOEe4 allele recognized as its strongest genetic risk factor. However, the mechanisms by which APOEe4 drives distinct molecular changes – whether by exacerbating pathology or triggering compensatory responses – remain incompletely understood. We generated and analyzed proteomic, epigenetic, and genetic data from post-mortem dorsolateral prefrontal cortex samples of a uniquely APOEe4-enriched subset of the Religious Orders Study and Memory and Aging Project (ROSMAP). Specifically, we generated DIA LC-MS proteomic data (n = 302), analyzed previously generated DNA methylation profiles from our group (n = 310), and used published whole-genome sequencing data (n = 254) to compute polygenic risk scores (PRS). In this cohort, 69% (n = 214) were APOEe4 carriers, and 19.6% (n = 42) of them showed no pathological evidence of AD based on NIA-Reagan criteria, enabling identification of APOEe4-related risk and resilience mechanisms. In the absence of AD, APOEe4 carriers exhibited lower levels of 27 proteins, suggesting early synaptic (e.g., VAMP1, SYN3, CASKIN1) and metabolic (e.g., GLUD1, PI4KA) vulnerability. By contrast, APOEe4 carriers with AD displayed marked upregulation of inflammatory and proteostatic proteins (e.g., GNAO1, AHNAK, FGG, HEBP1, APEX1, RAB4A, SLC12A5, LRP1, BAG6) and hypermethylation of cg06329447 in ELAVL4. Network analyses highlighted convergent disruptions in synaptic transmission, metabolism, and proteostasis – key pathways altered in APOEe4-associated AD. Mediation analyses identified GRIPAP1 and GSTK1 as top protein mediators (accounting for ∼26–33% of APOEe4’s effect), with VAMP1, CASKIN1, DPP3, SYN3, and FGG each contributing ∼9–15%. ELAVL4 hypermethylation also mediated ∼12% of the APOEe4 effect, linking epigenetic dysregulation to disease risk. To assess whether the identified proteins reflected broader genetic risk for AD or were specific to APOEe4, we calculated PRS both excluding and including the APOE genomic region. While the non-APOE PRS showed no association with identified molecular markers, the APOE-inclusive PRS was significantly associated with eight AD-related proteins in carriers, indicating they are not explained by polygenic risk outside of APOE. Finally, predictive modeling stratified by APOEe4 status revealed that in non-carriers, PRS most effectively classified AD (AUC = 0.73), whereas in carriers, proteomic and epigenetic markers outperformed PRS (AUC up to 0.74). Together, these findings demonstrate that APOEe4 confers AD risk through early synaptic and metabolic disruptions and later-stage inflammatory and epigenetic changes, laying the groundwork for genotype-tailored biomarker development and therapeutic strategies. Download figure Open in new tab INTRODUCTION Alzheimer’s disease (AD) is an escalating global health challenge, currently affecting over 55 million people worldwide, with projections estimating a rise to 139 million by 2050 due to population aging [ 1 ]. AD imposes substantial socioeconomic burdens, with global costs exceeding $1.3 trillion annually, driven by long-term care expenditures and productivity losses [ 2 ]. Among genetic contributors to late-onset AD, the apolipoprotein E e4 (APOEe4) allele is the most robustly documented risk factor, carried by approximately ∼14% of the global population, with the frequency rising dramatically to ∼40% in patients with AD [ 3 , 4 ]. The allele is associated with a 3-4-fold increased risk (12–15-fold for homozygotes) relative to e3/e3 individuals, as well as with earlier disease onset and more aggressive clinical progression [ 3 , 5 – 8 ]. Its pathogenic influence extends beyond the classical amyloid–tau mechanisms, disrupting lipid metabolism, mitochondrial function, oxidative stress responses, neuroinflammation, synaptic function, and cerebrovascular integrity [ 9 ]; [ 10 – 12 ]. APOEe4 carriers also show earlier reductions of CSF Aβ and elevations of p-tau that precede symptom onset [ 13 – 15 ], suggesting a long preclinical window for molecular intervention. Recent large-scale proteomic work further underscores this point: APOEe4 carriers display a conserved proteomic signature across brain, CSF, and plasma, marked by pro-inflammatory and infection-related immune pathways independent of neurodegenerative diagnosis [ 16 ]. Despite these established associations, a detailed molecular understanding of how APOEe4 drives these changes in the brain remains lacking. Large-scale studies have extensively catalogued protein-level changes associated with AD [ 17 – 23 ]. However, these analyses typically adjust for APOE genotype effects without explicitly investigating genotype-driven biological heterogeneity, obscuring e4-specific changes. Similarly, while epigenetic changes have emerged as critical modulators of gene-environment interactions in neurodegeneration, the impact of APOE genotype on these epigenetic landscapes remains poorly characterized [ 24 – 31 ]. Several CpG sites near genes like ANK1, BIN1, and RHBDF2 consistently associate with AD [ 32 – 35 ], yet few studies examine how APOE genotype might influence these methylation patterns. To address these knowledge gaps, we leverage large, uniquely structured subset of the Religious Orders Study and Memory and Aging Project (ROSMAP), distinguished by an exceptionally high proportion of APOEe4 carriers (69%, n = 214), with 19.6% (n = 42) of them showing no pathological evidence of AD based on NIA-Reagan criteria. This unique cohort composition provides a rare window into the molecular determinants of both AD risk and resilience in the brain among genetically susceptible populations. By generating proteomic and DNA methylation data in this cohort, we provide novel multi-omic evidence supporting a biphasic molecular model of APOEe4-driven pathology, characterized by early-stage synaptic and metabolic vulnerabilities that progress to late-stage inflammatory and proteostatic disruptions in clinically manifest AD. Notably, our analyses identify candidate molecular mediators, highlighting potential genotype-specific therapeutic targets to modify risk specifically in APOEe4 carriers. RESULTS DATASET OVERVIEW Data for this study were derived from post-mortem prefrontal cortex brain samples obtained from participants in the Religious Orders Study and the Memory and Aging Project (ROSMAP) [ 36 ]. All subjects were categorized by both APOE genotype, as described [ 37 ], and Alzheimer’s disease (AD) status. AD pathology was assessed post-mortem using a dichotomized version of the National Institute on Aging–Reagan (NIA-Reagan) criteria [ 38 ], which combines CERAD plaque scores and Braak neurofibrillary tangle staging, as described [ 39 ]. Importantly, these evaluations were performed blinded to clinical data, and classification was based solely on the extent of neuropathology: individuals with intermediate or high levels of pathology were classified as having AD. The cohort was intentionally enriched for APOEe4 carriers, who made up 69.0% (n=214) of the total sample ( Table 1 ). Notably, 19.6% of these carriers (n=42) showed no evidence of AD pathology. Within each APOE genotype group, Non-AD and AD individuals had similar age and sex distributions. This sampling strategy allowed us to specifically examine both risk and resilience mechanisms related to the APOEe4 allele. View this table: View inline View popup Table 1 Molecular data were collected across all combinations of APOE genotype and AD status, with comparable sampling proportions for proteomics, DNA methylation, and polygenic risk scores (PRS) ( Table 1 ). The overlap in data availability across modalities is illustrated in Supplementary Figure 1 . Proteomic profiling was performed on 302 of the 310 subjects using Liquid Chromatography–Mass Spectrometry (LC-MS) with Data Independent Acquisition (DIA). This approach identified 4,901 unique proteins, including isoforms and post-translationally modified variants, as quantified by Scaffold DIA software. After grouping by UniProt protein names and excluding proteins with >20% missing data, 1,291 proteins were retained for downstream analysis. DNA methylation data were obtained using the Illumina EPIC array on all 310 subjects. PRS were calculated for a subset of 254 individuals using whole-genome sequencing data and effect size weights from the AD GWAS meta-analysis by [ 40 ], which included 39,918 AD cases and 358,140 controls. BASELINE MOLECULAR CHARACTERISTICS OF ALZHEIMER’S DISEASE We first sought to characterize baseline molecular alterations associated with Alzheimer’s disease (AD) in our APOEe4-enriched cohort. Differential analyses were performed to compare protein abundance and DNA methylation profiles between individuals with and without AD pathology. All models were adjusted for age, sex, neuronal proportion, and post-mortem interval (PMI) to minimize confounding. Proteomics In the proteomics dataset, differential analysis (FDR ≤ 0.05) revealed 32 proteins significantly upregulated and 16 significantly downregulated in AD ( Figure 1 ; Supplementary Table 1 ). Among the upregulated proteins were canonical AD markers, including amyloid precursor protein (APP; logFC = 2.52, p = 5.34 × 10⁻⁹) and tau (MAPT; logFC = 0.23, p = 1.19 × 10⁻³), alongside less commonly reported proteins such as SLC25A1 (logFC = 0.61, p = 7.71 × 10⁻⁴). Downregulated proteins included key regulators of synaptic transmission (GRIPAP1, logFC = –1.36, p = 6.36 × 10⁻⁶), neuronal stress response (HTT, logFC = –0.74, p = 0.0014), and calcium homeostasis (CSDE1, logFC = –0.98, p = 4.12 × 10⁻⁵) [ 41 ]; [ 42 ]; [ 43 ]. Download figure Open in new tab Figure 1 Differential Protein and DNA Methylation Profiles in AD Panel A: Volcano plot of differential protein abundance: log₂ fold change vs. –log₁₀P. Proteins with |log₂FC| ≥ 0.58 and FDR ≤ 0.10 are labeled (red for increased protein level in AD, blue for decreased); points above –log₁₀P = 8 are clipped (red triangles). The dashed horizontal line indicates FDR = 0.10, and the dotted line indicates FDR = 0.05. Panel B: Volcano plot of differential DNA methylation: log₂ fold change (Δβ) vs. –log₁₀P. CpGs meeting FDR ≤ 0.10 are labeled (red for increased methylation in AD, blue for decreased). Points above –log₁₀P = 8 are clipped (red triangles). The dashed horizontal line indicates FDR = 0.10, and the dotted line indicates FDR = 0.05. Panel C: Manhattan plot of genome-wide differential DNA methylation (AD vs. control) in prefrontal cortex, adjusted for age, sex, neuronal proportion, and post-mortem interval. The dashed horizontal line indicates FDR = 0.10, and the dotted line indicates FDR = 0.05. To expand the set for enrichment analysis, we relaxed the FDR cutoff to 0.1, yielding 49 upregulated and 27 downregulated proteins. GO enrichment of the upregulated proteins revealed involvement in immune-related processes, filament organization, blood coagulation, and inflammatory response, emphasizing their likely role in AD pathology. For downregulated proteins, the top driver GO terms included synaptic vesicle recycling via endosome, regulation of modification of synaptic structure, positive regulation of macroautophagy, and vesicle-mediated transport in synapse, highlighting disruptions in synaptic maintenance, plasticity, and intracellular trafficking ( Supplementary Figure 2 ). Collectively, these results align with established models of AD and reflect coordinated dysregulation across immune, synaptic, and structural pathways. DNA Methylation Differential methylation analysis in prefrontal cortex samples comparing AD and non-AD individuals identified six hypermethylated and two hypomethylated CpG sites at FDR < 0.05 (19 and 6, respectively, at FDR < 0.1) ( Figure 1 ; Supplementary Table 2 ). As expected, we observed hypermethylation of ANK1 (cg05066959; ΔBeta = 0.029, p = 4.42 × 10⁻⁸), a well-established epigenetic marker of neurodegeneration [ 44 ] implicated in cytoskeletal dysregulation. Several other loci point to novel candidate targets for future epigenetic studies in AD. These included LY86-AS1 (cg06701353; ΔBeta = 0.007, p = 7.37 × 10⁻⁷), an lncRNA inversely associated with Braak stage [ 45 ], and two CpG sites in ERMAP (cg25285237, cg12410370; ΔBeta = 0.017, p = 1.26 × 10⁻⁶ and ΔBeta = 0.0086, p = 1.30 × 10⁻⁶), a gene involved in macrophage and T-cell modulation during amyloid clearance [ 46 ]. cg06329447 in ELAVL4 (ΔBeta = 0.013, p = 3.60 × 10⁻⁶) was hypermethylated. ELAVL4 is an RNA-binding protein involved in amyloid regulation [ 47 ], and its epigenetic alteration may play a role in AD pathogenesis. Hypomethylated sites included CDHR2 (cg00332146; ΔBeta = –0.020, p = 2.26 × 10⁻⁶), which harbors AD-associated exonic variants [ 48 ]; S1PR4 (cg17518965; ΔBeta = –0.017, p = 3.50 × 10⁻⁶), potentially reflecting increased immune signaling [ 49 ]; and STRN4 (cg04247584; ΔBeta = –0.0054, p = 8.80 × 10⁻⁷), previously implicated in tauopathy mouse models [ 50 ]. Taken together, these differentially methylated sites highlight a combination of known and novel epigenetic changes in AD, including immune, cytoskeletal, and RNA-regulatory processes. IDENTIFYING MOLECULAR MODERATORS OF THE APOE4 X AD RELATIONSHIP To better understand how APOE genotype modifies molecular changes in Alzheimer’s disease (AD), we used interaction models to test whether the relationship between AD status and molecular features differs between APOEe4 carriers and non-carriers. Because our study aimed to uncover molecular factors that may either confer risk or resilience in the context of APOEe4, this analysis was central to our framework. To interpret these interactions, we also categorized individuals into four biologically meaningful genotype–phenotype subgroups: APOEe4−/AD−, APOEe4−/AD+, APOEe4+/AD−, and APOEe4+/AD+ ( Supplementer Table 3 ). This allowed us to better characterize molecular signatures that may underlie susceptibility to, or protection from, AD in APOEe4 carriers. Proteomics Our analysis identified 19 proteins with significant APOE × AD interaction effects at FDR < 0.05 ( Supplementary Table 4 ). To enable broader biological interpretation and downstream network analyses, we also considered proteins at a more lenient threshold (FDR < 0.1), yielding 121 candidates. We examined how their expression changed across the four genotype–phenotype subgroups (APOEe4−/AD−, APOEe4−/AD+, APOEe4+/AD−, and APOEe4+/AD+) to distinguish between shared disease signatures and genotype-specific effects. We found that 94 proteins were significantly decreased in AD among APOEe4 non-carriers ( Figure 2 ). Within this group, 29 proteins also exhibited lower expression in healthy APOEe4 carriers compared to non-carriers, suggesting that APOEe4 may predispose individuals to a lower baseline of protective proteins that are not further altered by AD itself. This may represent a molecular vulnerability that increases susceptibility to disease. A subset of proteins, including GRIPAP1, BSN, EPHA4, and IDH3A, was consistently downregulated in both carriers and non-carriers with AD, suggesting these changes are robust hallmarks of AD irrespective of genotype. Notably, GRIPAP1 was also reduced in healthy carriers relative to non-carriers. Together, these findings point to both APOEe4-independent and APOEe4-specific mechanisms contributing to AD pathogenesis. Download figure Open in new tab Figure 2 Differential Protein Expression Across APOE and AD Status Groups Panel A: Clustered heatmap displaying scaled, covariate-adjusted mean protein levels for each APOE/AD subgroup: APOE4−/AD−, APOE4−/AD+, APOE4+/AD−, and APOE4+/AD+. In the primary heatmap, colors range from light yellow (lower expression) to red (higher expression). The adjacent contrast heatmap shows directional group-level differences (Δ protein levels) for selected pairwise comparisons. Color indicates direction and magnitude of change: darker blue represents a relative decrease, darker red a relative increase, and grey indicates a nonsignificant comparison. Statistical significance is annotated directly within the contrast heatmap: p< 0.001 (***), p≤ 0.01 (**), p≤ 0.05 (*), p< 0.1 (·), and “NS” for nonsignificant (p ≥ 0.1). Panel B: Boxplots of three representative proteins illustrating distinct patterns of differential expression across APOE/AD subgroups. Residual protein levels were adjusted for age, sex, postmortem interval, and neuronal proportion. Each dot represents an individual sample. Statistical comparisons between selected subgroup pairs were performed using Wilcoxon tests. Asterisks indicate significance levels as follows: p < 0.001 (***), p < 0.01 (**), p < 0.05 (*). We also identified several proteins – GNAO1, AHNAK, FGG, HEBP1, APEX1, RAB4A, SLC12A5, LRP1, CRKL, and BAG6 – that were significantly upregulated only in AD among APOEe4 carriers. These changes may reflect either genotype-specific pathological processes or compensatory responses to disease. Based on these patterns, we prioritized 39 proteins with APOEe4-specific expression shifts in AD for further investigation. These proteins likely represent mechanisms of risk amplification or resilience specific to APOEe4 carriers. To determine whether these proteins mediate the effect of APOEe4 on AD risk, we applied mediation analysis. In this context, mediation analysis quantifies how much of APOEe4’s association with AD can be explained by changes in a given protein. Seven proteins showed statistically significant mediation effects (ACME p < 0.05; Supplementary Table 5 ). GRIPAP1 and GSTK1 explained approximately 26–33% of the total APOEe4 effect, suggesting they may play central roles in linking APOEe4 to AD pathology. VAMP1, CASKIN1, DPP3, and SYN3 mediated 9–15% each, with most showing decreased expression in both healthy APOEe4 carriers and AD in non-carriers—consistent with a potential role in early vulnerability. FGG, in contrast, was upregulated in APOEe4 carriers with AD and showed modest mediation (∼10%), indicating it may be part of a different, possibly later-acting disease mechanism. These results highlight potential molecular intermediaries of APOEe4-associated AD risk and offer new targets for further mechanistic and therapeutic investigation. To link these findings to neuropathological hallmarks, we assessed associations between the 39 prioritized proteins and two standard pathological staging systems: CERAD scores (for amyloid plaques) and Braak stage (for tau tangles). Sixteen proteins were significantly associated with CERAD scores (p < 0.05; Supplementary Table 6 ), with directions consistent with their associations with AD and APOEe4 status. Only six proteins – AHNAK, GNAO1, FGG, KIF5C, GNAI2, and LRP1 – were associated with Braak stage, suggesting that the majority of APOEe4-specific proteins may be more strongly tied to amyloid pathology than tau. This supports the notion that APOEe4 primarily modulates the amyloidogenic pathway and highlights several proteins with potential for early biomarker development or targeted intervention. DNA Methylation Given the high dimensionality of this dataset, we first filtered CpG sites associated with AD at an FDR < 0.1. This preprocessing step reduced the number of sites under consideration and improved power for detecting meaningful APOE × AD interaction effects. Among the remaining sites, cg06329447 – located in the ELAVL4 gene – stood out due to its increased methylation in AD specifically among APOEe4 carriers (interaction p = 0.03). ELAVL4 is a neuronal RNA-binding protein implicated in amyloid processing and synaptic function [ 47 , 51 ], making it a biologically plausible mediator of APOE-related effects. To assess whether cg06329447 methylation might lie on the causal pathway between APOEe4 and AD, we applied mediation analysis (as previously done for protein markers). This analysis estimated that methylation at cg06329447 accounted for 11.7% of the total APOEe4 effect on AD risk (ACME = 0.0276, 95% CI: 0.00431–0.0600, p = 0.014), suggesting a modest but significant role in mediating genetic risk. We then evaluated the relationship between cg06329447 methylation and neuropathological burden. Methylation at this site was strongly associated with CERAD scores (β = 1.59, p = 1.39×10⁻⁵), and to a lesser extent with Braak scores (β = 1.71, p = 0.000206), again indicating a tighter link to amyloid pathology than to tau. While cg06329447 methylation was not directly correlated with ELAVL4 protein levels, we found a significant APOEe4 × ELAVL4 protein interaction in logistic regression models of AD status (interaction p = 0.015). Specifically, higher ELAVL4 protein levels were protective in APOEe4 non-carriers (log odds = –1.03), but this effect was neutralized in carriers (interaction β = 1.15). ELAVL4 protein also negatively correlated with CERAD scores (β = –0.92, p = 0.0218), but showed no significant association with Braak stage. PROTEIN NETWORK OF APOEe4-RELATED AD MOLECULAR INTERACTIONS To elucidate molecular interactions potentially central to AD pathogenesis, particularly in the context of APOE genotype variation, we constructed a protein interaction network guided by our multi-omic findings. The network was built using known protein–protein interactions from the STRING database [ 52 ] and included proteins that exhibited significant APOE-associated changes in our analyses. We also incorporated ELAVL4, given its association with the differentially methylated CpG site cg06329447 and its established role in AD pathology. To contextualize these findings within the broader AD framework, core AD-related proteins such as MAPT, APOE, and APP were included as central reference nodes. The final network consisted of 43 nodes connected by 36 edges, with an average node degree of 1.67. This reflects tighter clustering than expected by chance (expected number of edges: 15), indicating a significant protein–protein interaction enrichment (p < 1.5×10⁻⁶). We used Markov Cluster Algorithm [ 53 ] to partition the network into functional modules, which we annotated based on known protein functions and Gene Ontology (GO) enrichment terms where applicable ( Figure 3 ). The resulting clusters were as follows: Download figure Open in new tab Figure 3 Multi-Omic-Informed Protein Network Anchored by Key AD-Related Genes A network diagram depicting proteins identified through our analyses and literature-based relevance to Alzheimer’s disease. Nodes represent individual proteins and are color-coded according to manually curated functional clusters, including synaptic signaling, metabolic processes, cytoskeletal organization, and immune-related pathways. Key AD-related proteins (MAPT, APOE, and APP) were manually included as reference nodes to anchor the network within the broader context of AD pathogenesis. ELAVL4 was also incorporated due to its relevance to a differentially methylated CpG site (cg06329447) and its established role in amyloid regulation. Solid edges indicate known protein–protein interactions retrieved from the STRING database, with edge thickness reflecting the strength of interaction evidence. Dashed edges represent interactions that connect different functional clusters. Proteins that did not form any connections within the network are not shown. 1) Cell activation + Chemical synaptic transmission: APP, APOE, BAG6, ELAVL4, FGG, HEBP1, LRP1, MAPT, VGF, SLC12A5, BSN 2) Regulation of adenylate cyclase: GNAO1, GRM2, GNAI2, CASKIN1 3) MET receptor recycling + Lipid metabolic process: PI4KA, CRKL, RAB4A, GRIPAP1 4) TCA cycle: ME3, DLST, GLUD1 5) Protein translation: FARSB, SARS2 6) Vesicle-mediated transport: SYN3, VAMP1 7) Protein folding: HSPA8, DPP3 Fifteen proteins did not connect within the network: NRXN3, OXR1, SLC25A22, HPRT1, AHNAK, ARFGEF3, APEX1, GSTK1, CAP2, LRRC47, SFXN1, VPS26B, DMXL2, SYNPO, KIF5C. Overall, the network was significantly enriched for multiple GO terms related to synaptic and intracellular communication, including Chemical synaptic transmission, Dendrite development, Synapse assembly and modulation, Behavior, Axo-dendritic transport, Regulation of vesicle-mediated transport – as detailed in Supplementary Table 7 . These findings underscore the interconnected nature of synaptic, metabolic, and proteostatic processes modulated by APOE genotype in AD. Notably, HSPA8 – a molecular chaperone involved in protein folding – exhibited high connectivity within the network, linking multiple functional modules through five distinct interactions. INTEGRATION OF POLYGENIC RISK SCORES To further elucidate the genetic underpinnings of the proteomic and epigenetic alterations observed in our study, we calculated Polygenic Risk Scores (PRS) to capture the cumulative impact of common AD-associated variants. We derived two versions of PRS: one excluding the APOE locus (non-APOE PRS) and one encompassing the full genome including the APOE region (APOE PRS). This distinction enabled us to distinguish between genome-wide risk and the APOEe4-driven signal. In logistic regression models assessing neuropathological burden, a one–standard deviation (SD) increase in the non-APOE PRS was associated with a 79% increase in the odds of high AD pathology by NIA-Reagan criteria (OR = 1.79, 95% CI: 1.53–2.09, p = 1.67×10⁻⁴), a 76% increase for Braak stage (OR = 1.76, 95% CI: 1.42–2.17, p = 0.0073), and a 68% increase for CERAD scores (OR = 1.68, 95% CI: 1.44–1.96, p = 7.78×10⁻⁴). In contrast, the APOE-inclusive PRS exhibited substantially stronger associations: each 1-SD increase was linked to a 117% increase in odds of high NIA-Reagan classification (OR = 2.17, 95% CI: 1.57–3.01, p = 6.01×10⁻⁶), an 89% increase for Braak stage (OR = 1.89, 95% CI: 1.26–2.84, p = 0.0032), and a 116% increase for CERAD score (OR = 2.16, 95% CI: 1.56–3.00, p = 8.45×10⁻⁶). These findings underscore the added explanatory power of APOE-linked variants, particularly in relation to amyloid burden as reflected in CERAD scores. To assess the relationship between PRS and molecular features, we tested associations between both types of PRS and the differentially expressed proteins and CpG sites identified in earlier analyses. The non-APOE PRS, calculated using both continuous shrinkage (csPRS; [ 54 ] and summary statistics-based nonparametric (SDPR; [ 55 ] methods, showed no significant associations with either protein or methylation markers after multiple testing correction. This suggests that genome-wide risk outside of APOE may act via pathways not directly captured by proteomic or epigenetic signatures in the prefrontal cortex. In contrast, the APOE-inclusive PRS, calculated using the same two methods, demonstrated significant associations with the expression of eight proteins previously found to be upregulated in AD among APOEe4 carriers. These included: HEBP1 (csPRS: beta=0.18, FDR=0.0046; SDPR: beta=0.18, FDR=0.0057), LRP1 (csPRS: beta=0.17, FDR=0.0174; SDPR: beta=0.15, FDR=0.0282), FGG (csPRS: beta=0.16, FDR=0.0495; SDPR: beta=0.18, FDR=0.0230), AHNAK (csPRS: beta=0.24, FDR=0.0010; SDPR: beta=0.25, FDR=0.0006), GNAO1 (csPRS: beta=0.18, FDR=0.0021; SDPR: beta=0.16, FDR=0.0057), SLC12A5 (csPRS: beta=0.19, FDR=0.0021; SDPR: beta=0.21, FDR=0.0008), BAG6 (csPRS: beta=0.18, FDR=0.0095; SDPR: beta=0.19, FDR=0.0057), and RAB4A (csPRS: beta=0.13, FDR=0.0513; SDPR: beta=0.16, FDR=0.0088). These results offer additional evidence that these proteins are tightly linked to APOEe4-specific mechanisms rather than broader polygeneic risk, and validate their relevance in genetically driven AD pathology. A consolidated summary of the major findings across omic layers is presented in Table 2 , which outlines the direction and significance of each molecular association and its relation to APOEe4-specific pathology. View this table: View inline View popup Table 2 COMPARISON OF APOEe4-RELATED MOLECULAR CHANGES FOR PREDICTIVE MODELING Finally, to evaluate the predictive utility of the molecular features identified in our analyses, we assessed how protein expression, DNA methylation, and polygenic risk scores (PRS) performed in classifying AD status. We focused particularly on whether these modalities capture distinct aspects of AD pathology in APOEe4 carriers versus non-carriers. To summarize proteomic variation, we performed principal component analysis (PCA) separately on each protein cluster defined in our previously constructed interaction network, using only the proteins identified in interaction analyses. The first principal component (PC) from each cluster was used as a representative feature. The 15 proteins that did not cluster were grouped and similarly summarized using PCA. For DNA methylation, we initially examined the predictive power of a single CpG site, cg06329447, which had shown APOEe4-dependent differential methylation. In APOEe4 non-carriers, predictive modeling yielded an area under the curve (AUC) of 0.69 for proteomic PCs, 0.73 for the APOE-inclusive PRS, and 0.62 for cg06329447 methylation. In contrast, in APOEe4 carriers, the AUC values were 0.63 (proteomics), 0.56 (PRS), and 0.66 (cg06329447 methylation), suggesting that non-genetic markers – particularly DNA methylation – have greater predictive value than PRS in carriers. Recognizing that a single CpG site may not fully capture the complexity of APOEe4-related epigenetic dysregulation, we next applied weighted gene co-expression network analysis (WGCNA) [ 56 ] to 67,654 CpG sites nominally associated with AD (p < 0.05). This yielded several co-methylation modules whose eigengenes were tested for their predictive utility. Notably, cg06329447 (in ELAVL4) belonged to the pink module, which was strongly associated with AD (β = 0.26, p = 2×10⁻⁵), CERAD score (β = 0.24, p = 2×10⁻⁴), and Braak stage (β = 0.19, p = 0.03). It was also one of two modules that significantly moderated the APOEe4 × AD interaction (interaction coefficient = 0.65, p = 0.03). All ELAVL4-associated CpGs clustered into either the pink or the midnight blue modules. We then evaluated three sets of predictive models: (1) using the eigengene of the pink module only, (2) using the eigengenes of both the pink and midnight blue modules (representing modules positively associated with AD and that enhance the APOEe4 effect), and (3) using the eigengenes of all four moderating modules, which include the pink and midnight blue modules along with two additional modules – the salmon and light green modules – that were negatively associated with AD and moderated the APOEe4 × AD interaction in the opposite direction (salmon: AD β=–0.42, p=2×10⁻⁴; CERAD β=–0.43, p=2×10⁻⁴; Braak β=–0.30, p=0.05; interaction coefficient=–0.67, p=0.02; light green: AD β=–0.26, p=0.004; CERAD β=–0.25, p=0.007; interaction coefficient=–0.75, p=0.03). Using the eigengene of the pink module alone, the AUC improved to 0.67 in non-carriers and 0.65 in carriers ( Figure 4 ). Adding the eigengene of the midnight blue module did not significantly increase performance (AUC = 0.67 in non-carriers, 0.68 in carriers). However, incorporating all four moderating modules – pink, midnight blue, salmon, and light green – raised the AUC to 0.67 for non-carriers and 0.74 for carriers. This suggests that including DNA methylation modules negatively moderating the APOEe4 × AD relationship enhances predictive performance in carriers. Overall, network-based module integration strengthened the predictive value of DNA methylation for both groups. Download figure Open in new tab Figure 4 Discriminatory Accuracy of Protein-, PRS-, and Methylation-Based Models in APOEe4 Carriers vs Non-Carriers Receiver operating characteristic (ROC) curves for predictive models stratified by APOEe4 carrier status, showing the area under the curve (AUC) for models built using protein markers, polygenic risk scores (PRS), a single CpG site identified in the interaction analysis, and DNA methylation subnetworks derived from WGCNA. Protein predictors were summarized using principal component analysis (PCA) based on their assigned functional clusters from the network analysis. For methylation-based models, we evaluated three sets of predictors: (1) the eigengene of the pink module alone, (2) eigengenes of the pink and midnight blue modules (positively associated with AD and APOEe4 effect), and (3) eigengenes of all four moderating modules, including the salmon and light green modules, which were negatively associated with AD and moderated the APOEe4 × AD interaction in the opposite direction. Collectively, our integrative analyses – including interaction modeling, network construction, mediation testing, and predictive modeling – converge to highlight a distinct molecular signature of APOEe4-associated AD. While many changes are shared across genotypes, specific proteins and CpG methylation patterns show genotype-dependent modulation. This work establishes a comprehensive multi-omic framework for understanding APOEe4-related risk and resilience, and identifies molecular candidates for genotype-tailored diagnostic and therapeutic strategies. DISCUSSION In this study, we leveraged a multi-omic framework encompassing proteomic, epigenetic, and genetic data to characterize the molecular landscape associated with the APOEe4 allele in Alzheimer’s disease (AD). By integrating LC-MS proteomics, DNA methylation profiling, and polygenic risk scores (PRS), we identified a two-phase pattern of APOEe4-associated molecular alterations. In cognitively intact APOEe4 carriers, we observed early reductions in proteins related to synaptic function and metabolism, indicating presymptomatic molecular vulnerability. These findings align with recent positron emission tomography imaging studies showing reduced hippocampal and medial temporal lobe synaptic density in APOEe4 carriers prior to clinical symptoms [ 57 ]; [ 58 ], as well as elevated cerebrospinal fluid levels of the synaptic marker neurogranin in early cognitive impairment [ 59 ], underscoring the clinical relevance of our observations. The identified reductions in metabolic enzyme levels are also consistent with previous reports of decreased cerebral glucose metabolism in cognitively healthy APOEe4 carriers [ 60 – 64 ]. In contrast, APOEe4 carriers with AD exhibited increased inflammatory and proteostatic proteins and distinct epigenetic modifications, suggesting a secondary phase involving APOEe4-specific responses to established pathology. This dual-phase model provides a valuable framework for understanding how APOEe4 contributes to early susceptibility and subsequent progression of AD. INTERPLAY BETWEEN APOE GENOTYPE AND MOLECULAR CHANGES Our findings highlight distinct molecular signatures linked to APOEe4 genotype, emphasizing both early vulnerabilities and disease-specific alterations. Even in the absence of clinical AD, APOEe4 carriers displayed reductions in proteins crucial for synaptic integrity and metabolic homeostasis. Synaptic regulators such as VAMP1, essential for synaptic vesicle fusion and Aβ exocytosis [ 65 ]; [ 66 ], and SYN3, important for dopaminergic neuronal function [ 67 ], were notably reduced, aligning with preclinical models demonstrating synaptic deficits associated with APOEe4 [ 58 ]. Concurrently, decreased levels of metabolic enzymes GLUD1 and PI4KA, key regulators of glutamate-GABA balance and lipid metabolism [ 68 ]; [ 69 ], support prior evidence of metabolic dysfunction in APOEe4 carriers [ 70 ]. Early oxidative imbalances were suggested by reduced antioxidant proteins OXR1 and DPP3 [ 71 – 73 ]. Mediation analyses identified proteins such as GRIPAP1, GSTK1, VAMP1, SYN3, CASKIN1, and DPP3 as significant intermediaries between APOE genotype and AD, collectively accounting for 10–33% of APOEe4’s total effect, reinforcing the importance of synaptic dysfunction and oxidative stress as mechanisms driving early APOEe4-related vulnerability. In the context of clinical AD, APOEe4 carriers demonstrated pronounced upregulation of inflammatory, proteostatic, and stress-response pathways, potentially indicative of compensatory responses. Elevated proteins such as HEBP1 and FGG suggest increased neurovascular and coagulation dysfunction [ 74 ]; [ 75 ], consistent with findings from isogenic iPSC-derived endothelial models showing ApoE4-linked alterations in blood coagulation and barrier integrity pathways [ 76 ]. Stress-response proteins including APEX1 and BAG6 further indicated attempts to mitigate proteostatic and cytoskeletal stress [ 77 ]. The increased expression of LRP1, a receptor modulated by APOEe4 and involved in Aβ clearance, may represent an inadequate compensatory mechanism [ 78 ]. Complementing these proteomic changes, hypermethylation at cg06329447 in ELAVL4, a critical RNA-binding regulator of Aβ production [ 47 , 51 ], emerged as a significant epigenetic alteration in APOEe4 carriers with AD. This hypermethylation mediated approximately 12% of the APOEe4 effect on AD, possibly destabilizing ELAVL4 transcripts and accelerating disease progression. Overall, our data suggest APOEe4 shifts the molecular balance from early synaptic-metabolic deficiencies to late-stage inflammatory and proteostatic stress. A systems-level interaction network analysis, informed by the STRING database, contextualized these molecular changes into distinct functional clusters, highlighting synaptic transmission, adenylate cyclase signaling, metabolic regulation, vesicle-mediated transport, and proteostasis. A prominent cluster linked synaptic transmission and neuroinflammation (APP, APOE, BAG6, ELAVL4, FGG, HEBP1, LRP1, MAPT, VGF, SLC12A5, BSN), underscoring APOEe4’s role in neurovascular dysfunction and synaptic vulnerability [ 79 ]; [ 58 ]. Another cluster centered on disrupted adenylate cyclase signaling (GNAO1, GRM2, GNAI2, CASKIN1), critical for synaptic plasticity and cognitive function [ 80 ]; [ 81 ]; [ 82 ]). Metabolic dysfunction involving mitochondrial enzymes GLUD1, DLST, and ME3 further pointed to bioenergetic deficits linked to APOEe4 [ 68 ]; [ 83 ]. Cellular trafficking disruptions were captured by a module involving MET receptor recycling and lipid metabolism (PI4KA, CRKL, RAB4A, GRIPAP1), aligning with APOEe4-driven lipid dysregulation and impaired receptor trafficking [ 84 ]; [ 85 ]; [ 86 ]; [ 70 ]). Vesicle-mediated transport deficits involving SYN3 and VAMP1 highlighted impaired intracellular trafficking related directly to Alzheimer’s pathology. Finally, proteostasis disruption emerged through proteins involved in protein folding and mitigating oxidative stress (HSPA8, DPP3), emphasizing the intersection between redox imbalance and protein quality control mechanisms ([ 87 ]; [ 72 ]). DPP3 activates neuroprotective NRF2 pathways by competing with KEAP1 binding during oxidative stress [ 88 ]; [ 89 ]), while HSPA8 is a molecular chaperone decreased in AD [ 87 ]. INTEGRATION OF POLYGENIC RISK SCORES AND PREDICTIVE MODELING By deriving two separate Polygenic Risk Scores (PRS) – one excluding the APOE region (non-APOE PRS) and one including it (APOE PRS) – we evaluated how genetic risk factors relate to the molecular signatures identified. The non-APOE PRS showed no significant correlations with our proteomic data, indicating that non-APOE genetic risk might act through pathways not captured by prefrontal cortex proteomics. In contrast, the APOE PRS confirmed significant associations with eight proteins increased in AD among APOEe4 carriers: HEBP1, LRP1, FGG, AHNAK, GNAO1, SLC12A5, BAG6, and RAB4A, further supporting their involvement in APOEe4-driven AD pathology. Predictive modeling using proteomic, epigenetic, and genetic data highlighted distinct molecular signatures in APOEe4 carriers versus non-carriers. In non-carriers, the APOE-inclusive PRS (AUC = 0.73) outperformed non-genetic molecular features, suggesting broader genetic contributions to sporadic AD. In contrast, in APOEe4 carriers, non-genetic markers – particularly WGCNA-integrated epigenetic signatures – demonstrated superior discriminative power compared to the PRS (AUC = 0.56), achieving an AUC of 0.74 when combining multiple methylation modules. Notably, using the eigengene of a single WGCNA module (pink) already improved prediction over individual CpG sites in both groups, with further gains observed in carriers when including additional modules, especially those negatively moderating the APOEe4 × AD interaction (salmon and light green). This suggests that coordinated, network-level methylation changes may better reflect biologically meaningful dysregulation than single-site signals, and that incorporating both risk-and resilience-linked modules enhances model performance in genetically susceptible individuals. Proteomic markers also showed better predictive power in carriers, with an AUC of 0.63. Thus, these findings reinforce the value of omic biomarkers in identifying at-risk individuals and tailoring therapeutic strategies for genetically defined subgroups. CONCLUSION/LIMITATIONS Collectively, our integrated multi-omic analysis delineates a detailed molecular landscape of APOEe4-associated AD pathogenesis, highlighting both the pathogenic and potentially compensatory networks activated by this allele. By merging proteomic, epigenetic, and genetic data, we identified a trajectory in which early synaptic and metabolic deficits predispose APOEe4 carriers to inflammatory, proteomic, and epigenetic dysregulation. These insights lay a promising foundation for the development of genotype-specific biomarkers and personalized therapeutic interventions. Nonetheless, several limitations must be acknowledged. The cross-sectional, post-mortem design of our study limits causal inference, emphasizing the need for longitudinal investigations – ideally incorporating peripheral biomarkers – to capture dynamic molecular changes in living individuals. In addition, the use of bulk tissue may mask cell-type–specific effects, suggesting that single-cell or spatial approaches could further refine our understanding of neuronal versus glial contributions to APOEe4-associated risk and resilience. Functional validation in APOEe4-relevant cellular or animal models remains essential to firmly establish causal mechanisms, while expanding cohort sizes and including participants from diverse genetic backgrounds will enhance the generalizability of our findings. Despite these challenges, our work advances the field by setting the stage for future studies that not only address current limitations but also pave the way toward more effective, personalized strategies for early diagnosis and intervention. METHODS Study Population and Tissue Acquisition Frozen brain tissue samples were obtained from participants in the Religious Orders Study and the Rush Memory and Aging Project (ROSMAP) [ 36 ]. All participants enrolled without known dementia and agreed to detailed clinical evaluation and brain donation at death. All studies were approved by an Institutional Review Board of Rush University Medical Center. Each participant signed informed and repository consents and all ROSMAP participants signed an Anatomic Gift Act. For this study, tissue was isolated from Brodmann Area 10 (prefrontal cortex). Sample phenotype data, including clinical diagnoses and neuropathologic assessments, were provided by the Rush University Alzheimer’s Disease Center in accordance with previous publications [ 90 , 91 ]. Alzheimer’s disease (AD) neuropathology was classified based on dichotomized NIA-Reagan diagnostic criteria, which integrate assessments of neurofibrillary tangles and neuritic plaques. Subjects were categorized as having AD pathology if they met the criteria for high or intermediate likelihood and as not having AD if they had low likelihood or no AD pathology. Braak stage was dichotomized for logistic regression analyses into two groups: stages 0-II versus III and above, where stages I-II indicate neurofibrillary tangles predominantly confined to the entorhinal region, stages III-IV indicate limbic region involvement (including the hippocampus), and stages V-VI indicate moderate to severe neocortical involvement. Similarly, CERAD scores were dichotomized, with a diagnosis of AD requiring moderate (probable AD) or frequent neuritic plaques (definite AD) in one or more neocortical regions [ 39 , 91 ]. Proteomic Data Acquisition and Processing Sample Preparation for LC–MS/MS Frozen tissues were lysed with a probe sonicator in solubilization buffer (RIPA buffer with serine/threonine protease and phosphatase inhibitors). Lysate was centrifuged at 14.5 x 103 g for 10 minutes at 4 °C in a tabletop centrifuge to pellet cellular debris, and supernatant was collected and protein precipitated with a standard methanol:chloroform:water method. Protein pellets were washed three times with cold MeOH prior to being resuspended in 80uL solubilization buffer (8M urea in 0.4M ammonium bicarbonate). Cysteine of proteins were reduced with 8 µL of 45 mM dithiothreitol (DTT) and incubated at 37 °C for 30 min, then were alkylated with 8 µL of 100 mM iodoacetamide (IAN) and incubated in the dark at room temperature for 30 min. Protein solution was then diluted with water to bring urea concentration to 2 M. Proteins were then digested with sequencing-grade trypsin (Promega, Madison, WI, USA) at a weight ratio of 1:50 (trypsin/protein). Digestion was allowed to occur at 37 °C overnight and more trypsin was added and incubated at 37°C for additional 4 hours. The digestion was quenched with by adding 0.1% formic acid; and peptide mixture was desalted using C18 spin columns (The Nest Group, Inc., Southborough, MA, USA). Eluted peptides were dried in a rotary evaporator and stored until ready for LC MS/MS data collection. The samples were resuspended in 0.2% trifluoroacetic acid (TFA) and 2% acetonitrile (ACN) in water prior to LC–MS/MS analysis. Mass Spectral Data Collection: Data-Independent Acquisition (DIA) DIA LC–MS/MS was performed using a Waters ACQUITY UPLC M-Class system (Waters Corporation, Milford, MA, USA) coupled to a Q-Exactive HFX (ThermoFisher Scientific, San Jose, CA, USA) mass spectrometer. After injection, the samples were loaded onto a trapping column (nanoEase M/Z Symmetry RP C18 Trap column, 180 µm × 20 mm) at a flow rate of 10 µL/min and separated with a RP C18 column (nanoEase M/Z column Peptide BEH C18, 75 µm × 250 mm). The compositions of mobile phases A and B were 0.1% formic acid in water and 0.1% formic acid in ACN, respectively. The peptides were separated and eluted with a gradient extending from 6% to 25% mobile phase B in 98 min and then to 85% mobile phase B in additional 5 min at a flow rate of 300 nL/min and a column temperature of 37 °C. Column regeneration and up to three blank injections were carried out in between each sample injection to ensure no-carry over. The data were acquired with the mass spectrometer operating in a DIA mode with an isolation window width of 10 Th. The full scan was performed in the range of 400– 1,000 m/z with “Use Quadrupole Isolation” enabled at an Orbitrap resolution of 30,000 at 200 m/z and automatic gain control (AGC) target value of 3 × 106. Fragment ions from each DIA window (e.g., MS2 fragmentation) were generated in the C-trap with higher-energy collision dissociation (HCD) at a normalized collision energy of 28% and detected in the Orbitrap at a resolution of 15,000. DIA spectra were searched against a Homo Sapiens brain proteome fractionated spectral library generated from DDA LC MS/MS spectra (collected from the same Q-Exactive HFX mass spectrometer) using Scaffold DIA software v. 2.1.0 (Proteome Software, Portland, OR, USA). Within Scaffold DIA, raw files were first converted to the mzML format using ProteoWizard v. 3.0.11748. The samples were then aligned by retention time and individually searched with a mass tolerance of 10 ppm for the precursor ions and a fragment mass tolerance of 10 ppm. The data acquisition type was set to “Overlapping margins of 2 Da”, and the maximum missed cleavages was set to 2. Peptides with charge states between 2 and 4 and 6-30 amino acids in length were considered for quantitation, and the resulting peptides were filtered by Percolator v. 3.01 at a threshold FDR of 0.01. Peptide quantification was performed by EncyclopeDIA v. 0.9.2 and five of the highest quality fragment ions were selected for protein quantitation. Proteins containing redundant peptides were grouped to satisfy the principles of parsimony, and proteins were filtered at a threshold of two peptides per protein and with an FDR of 1%. Batch correction and Outlier Removal One control peptide that was missing entirely from one of the batches was excluded from further analyses. The initial set of 4,902 protein identifications was then consolidated into 4,034 groups by summing abundances based on the protein names assigned by Scaffold DIA. Missing values were imputed using a minimum value imputation algorithm, and protein abundance values were subsequently log2-transformed. To correct for technical variation, the RUV4 algorithm [ 92 ] (from the R package ruv) was applied using 13 control peptides as negative controls. Following adjustment, proteins with greater than 20% missing values in the original dataset were removed, leaving 1,305 proteins for further analysis. These protein abundance values were then scaled (0 mean, 1 SD) to facilitate comparisons of effect sizes in downstream statistical testing. Quality control measures included principal component analysis, which identified two subjects as outliers; these subjects were removed from further analyses. The final dataset comprised proteomic data from subjects with the following APOE genotype distribution: APOEe3/3 (n=92), APOEe3/4 (n=193), and APOEe4/4 (n=17). Neuropathological classification based on dichotomized NIA-Reagan criteria resulted in 81 subjects classified as “No AD” and 221 subjects as “AD.” The mean age at death was 89.8 years, with a gender distribution of 200 females and 102 males. DNA Methylation Data Acquisition and Processing DNA methylation (DNAme) profiling was performed using the Illumina EPIC array, as previously described [ 93 ]. To minimize confounding by sex-specific effects, CpG sites located on sex chromosomes, as indicated in the Illumina EPIC manifest, were excluded from further analysis. Furthermore, CpGs in the lowest 10% of variance were removed prior to conducting differential methylation analyses, resulting in a final dataset of 761,813 CpG sites. The APOE genotype distribution for this cohort was as follows: APOEe3/3 (n=96), APOEe3/4 (n=197), and APOEe4/4 (n=17) and generated as reported [ 37 ]. Neuropathological classification based on the dichotomized NIA-Reagan criteria yielded 85 subjects classified as “No AD” and 225 as “AD.” The mean age of the subjects was 89.35 years, with a gender distribution of 205 females and 105 males. Genotyping and Polygenic Risk Score (PRS) Calculation To quantify genetic risk for AD, we derived polygenic risk score (PRS) weights using genome-wide association study (GWAS) summary statistics from Wightman et al. [ 40 ]. The GWAS included 39,918 AD cases and 358,140 controls. PRS weights were applied to imputed genome-wide genotypes from 1,780 ROSMAP participants to generate PRS scores, with genotype data accessed from https://pmc.ncbi.nlm.nih.gov/articles/PMC6080491/ [ 94 ]. Two versions of PRS were constructed: one excluding genetic variants within the APOE genomic region (GRCh37:19:40,000,000-50,000,000; non-APOE PRS) and another including genome-wide genetic variants (APOE PRS). PRS was calculated using two Bayesian-based methods, PRS-CS-auto and SDPR, both of which do not require parameter tuning [ 54 , 55 ]. The number of genetic variants included in PRS estimation varied by method and APOE inclusion criteria: for PRS-CS-auto, 1,048,147 variants were included in the APOE PRS and 1,044,866 in the non-APOE PRS, while for SDPR, the corresponding numbers were 958,326 and 955,300. For these analyses, a subset of 254 subjects was used. In this subset, the APOE genotype distribution was as follows: 79 subjects with APOEe3/3, 164 with APOEe3/4, and 11 with APOEe4/4. Neuropathological classification based on the NIA-Reagan score yielded 67 subjects classified as no AD and 187 as AD, with a mean age of 89.70 years (165 females and 89 males). Estimation of Neuron Proportion We used the DNAme data to estimate neuronal proportions in brain tissue samples via the Cell Epigenotype Specific Model (CETS) [ 101 ] implemented in R. Statistical Analyses Differential methylation and protein abundance analyses were performed using the limma package [ 102 ] (functions lmFit and eBayes), with models adjusted for age, sex, estimated neuronal proportion, and post-mortem interval (PMI). Functional enrichment analysis for differentially abundant proteins was conducted using g:Profiler [ 103 ] with false discovery rate (FDR) correction set at 0.05. Interaction analyses were carried out using logistic regression models of the form: AD ∼ APOE genotype × Protein/DNAme + sex + age + neuronal proportion + PMI implemented with the R stats package. APOE genotype was dichotomized based on the presence of at least one e4 allele. Comparisons of protein and DNAme levels across subject groups stratified by APOE genotype and AD pathology were performed on the covariate-adjusted molecule values (i.e., residuals from linear regression models, e.g., Protein ∼ sex + age + PMI + neuronal proportion) using the Wilcoxon test. Mediation analyses were conducted using the R mediation package [ 104 ]. Network Analysis Protein-protein interaction networks were constructed using the STRING database [ 52 ] to elucidate relationships among selected proteins. The CpG site cg06329447 was mapped to the corresponding gene, ELAVL4, for network inclusion. Clustering of the network was performed using the Markov Cluster Algorithm (MCL) [ 53 ] with an inflation parameter of 2.4 to identify functional modules within the network. Predictive modeling To evaluate the predictive performance of our molecular features for AD status, we constructed logistic regression models with the following form: AD ∼ Predictors + age + sex where “predictors” represent the candidate molecular features (e.g., proteins, CpG module eigengenes, or PRS scores). To facilitate direct comparison across different omic types, we restricted the analysis to the subset of subjects (n=247) with complete data for proteomics, DNAme, and PRS. In this subset, the APOE genotype distribution was as follows: 76 subjects with APOEe3/3, 160 with APOEe3/4, 11 and 11 with APOEe4/4, with a dichotomized NIA-Reagan classification of 64 “No AD” and 183 “AD” subjects, a mean age of 89.80 years, 161 females, and 86 males. Using the caret package [ 105 ], we performed five-fold cross-validation repeated five times, with out-of-fold predictions for every sample. This approach ensures that every sample contributes to both training and testing in an unbiased manner, and yields stable, smoothed estimates of predictive performance stratified by APOEe4 status. We then extracted false positive and true positive rates and calculated the area under the curve for plotting. WGCNA for DNA Methylation Weighted Gene Co-Expression Network Analysis (WGCNA) [ 56 ] was performed on DNAme data to identify modules of co-methylated CpG sites. CpGs were pre-selected based on their association with AD in linear models adjusted for age, sex, estimated neuronal proportion, and PMI, retaining sites with a nominal unadjusted p-value <0.05; a total of 67,654 CpG sites were included in the analysis. Network construction was conducted with the following parameters in the blockwiseModules function: a soft-thresholding power of 6, a signed network type, a minimum module size of 30, and a maximum block size of 34,000. Module eigengenes (defined as the first principal component of each module) were computed to summarize the methylation profiles of each module. Associations between module eigengenes and AD neuropathology were assessed using linear regression models adjusted for age, sex, neuronal proportion, and PMI. Additionally, moderation analyses were performed using logistic regression models structured as: AD ∼ APOE genotype × Eigengene + covariates to evaluate whether the relationship between DNAme modules and AD differed by APOE genotype. Data Availability All generated data and associated metadata have been made publicly available via Synapse. AUTHOR CONTRIBUTIONS YM, MEL, AHC conceived and designed the analyses. YM processed and analyzed the data, and wrote the initial manuscript. AHC provided supervision. AP assisted with data analysis. LX accessed whole-genome sequencing data and calculated polygenic risk scores (PRS). WW contributed to generating proteomic data and documentation. KTE accessed, cleaned, and harmonized DNA methylation and phenotypic data. JG, DB, JK, RS, GZ, JF contributed intellectually through manuscript editing and reviewing biological findings. BCC advised on proteomics data processing. HZ provided intellectual input on statistical analyses and project direction. SH, CHvD, CG, DAB aided in the design of the analysis. CHvD, DAB also provided original samples and metadata. TL is a subaward co-investigator and generated proteomic data and documentation. All authors reviewed and edited the manuscript. CONFLICT OF INTERESTS AHC has received consulting fees from FOXO Technologies, Inc., and TruDiagnostic for work unrelated to the present manuscript. MEL is a founding PI of Altos Labs. SH works for Altos Labs. MEL and AHC hold patents for epigenetic clocks they developed, unrelated to the present manuscript. All other authors declare no competing interests. BCC receives funding through a sponsored research agreement from GSK. CHvD has received consulting fees from Eisai, Cerevel, BMS, and UCB and grant support for clinical trials from Biogen, Eli Lilly, Eisai, Roche, Genentech, Cerevel, and UCB. ACKNOWLEDGEMENT & FUNDING This study was funded by the NIA (R01AG057912 to Higgins-Chen, Levine). We also thank the Keck MS & Proteomics Resource at Yale School of Medicine for providing the necessary mass spectrometers and the accompany biotechnology tools funded in part by the Yale School of Medicine and by the Office of The Director, National Institutes of Health (S10OD02365101A1, S10OD019967, and S10OD018034). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. TTL had efforts under the R01AG057912 (Levine, PI). ROSMAP is supported by P30AG10161, P30AG72975, R01AG17917. R01AG015819, U01AG072572, and U01AG046152. Funder Information Declared National Institutes of Health, https://ror.org/01cwqze88 , R01AG057912 , S10OD02365101A1 , S10OD019967 , S10OD018034 , P30AG10161 , P30AG72975 National Institutes of Health, https://ror.org/01cwqze88 , R01AG17917 , R01AG015819 , U01AG072572 , U01AG046152 Yale School of Medicine REFERENCES 1. ↵ GBD 2019 Dementia Forecasting Collaborators. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019 . Lancet Public Health. Elsevier BV ; 2022 ; 7 : e105 – 25 . OpenUrl 2. ↵ Wimo A , Seeher K , Cataldi R , Cyhlarova E , Dielemann JL , Frisell O , Guerchet M , Jönsson L , Malaha AK , Nichols E , Pedroza P , Prince M , Knapp M , et al. The worldwide costs of dementia in 2019 . Alzheimers Dement. Wiley ; 2023 ; 19 : 2865 – 73 . OpenUrl 3. ↵ Farrer LA , Cupples LA , Haines JL , Hyman B , Kukull WA , Mayeux R , Myers RH , Pericak-Vance MA , Risch N , van Duijn CM . Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease. A meta-analysis. APOE and Alzheimer Disease Meta Analysis Consortium . JAMA. American Medical Association (AMA) ; 1997 ; 278 : 1349 – 56 . OpenUrl 4. ↵ Guo T , Zhang D , Zeng Y , Huang TY , Xu H , Zhao Y . Molecular and cellular mechanisms underlying the pathogenesis of Alzheimer’s disease . Mol Neurodegener. Springer Science and Business Media LLC ; 2020 ; 15 : 40 . 5. ↵ ene Dose of Apolipoprotein E Type 4 Allele and the Risk of Alzheimer’s Disease in Late Onset Families. 6. Frisoni GB , Altomare D , Thal DR , Ribaldi F , van der Kant R , Ossenkoppele R , Blennow K , Cummings J , van Duijn C , Nilsson PM , Dietrich P-Y , Scheltens P , Dubois B . The probabilistic model of Alzheimer disease: the amyloid hypothesis revised . Nat Rev Neurosci. Springer Science and Business Media LLC ; 2022 ; 23 : 53 – 66 . OpenUrl 7. Neu SC , Pa J , Kukull W , Beekly D , Kuzma A , Gangadharan P , Wang L-S , Romero K , Arneric SP , Redolfi A , Orlandi D , Frisoni GB , Au R , et al. Apolipoprotein E genotype and sex risk factors for Alzheimer disease: A meta-analysis . JAMA Neurol. American Medical Association (AMA ); 2017 ; 74 : 1178 – 89 . OpenUrl 8. ↵ Genin E , Hannequin D , Wallon D , Sleegers K , Hiltunen M , Combarros O , Bullido MJ , Engelborghs S , De Deyn P , Berr C , Pasquier F , Dubois B , Tognoni G , et al. APOE and Alzheimer disease: a major gene with semi-dominant inheritance . Mol Psychiatry. Springer Science and Business Media LLC ; 2011 ; 16 : 903 – 7 . OpenUrl 9. ↵ Fernandez CG , Hamby ME , McReynolds ML , Ray WJ . The role of APOE4 in disrupting the homeostatic functions of astrocytes and microglia in aging and Alzheimer’s disease . Front Aging Neurosci. Frontiers Media SA ; 2019 ; 11 : 14 . 10. ↵ Shi Y , Alzheimer’s Disease Neuroimaging Initiative, Yamada K , Liddelow SA , Smith ST , Zhao L , Luo W , Tsai RM , Spina S , Grinberg LT , Rojas JC , Gallardo G , Wang K , et al. ApoE4 markedly exacerbates tau-mediated neurodegeneration in a mouse model of tauopathy. Nature . Springer Science and Business Media LLC ; 2017 ; 549 : 523 – 7 . OpenUrl 11. Huang Y-WA , Zhou B , Wernig M , Südhof TC . ApoE2, ApoE3, and ApoE4 differentially stimulate APP transcription and Aβ secretion. Cell . Elsevier BV ; 2017 ; 168 : 427 – 41 .e21. OpenUrl 12. ↵ Liu C-C , Liu C-C , Kanekiyo T , Xu H , Bu G . Apolipoprotein E and Alzheimer disease: risk, mechanisms and therapy . Nat Rev Neurol. Springer Science and Business Media LLC ; 2013 ; 9 : 106 – 18 . OpenUrl 13. ↵ Lautner R , Insel PS , Skillbäck T , Olsson B , Landén M , Frisoni GB , Herukka S-K , Hampel H , Wallin A , Minthon L , Hansson O , Blennow K , Mattsson N , et al. Preclinical effects of APOE ε4 on cerebrospinal fluid Aβ42 concentrations . Alzheimers Res Ther . 2017 ; 9 : 87 . 14. Kester MI , Bouwman FH , van Elk EJ , Blankenstein MA , Scheltens P , van der Flier WM . P3-079: CSF biomarker levels , APOE genotype and the effect of aging. Alzheimers Dement. Wiley ; 2008 ; 4 : T541 – T541 . OpenUrl 15. ↵ Shang L , Dong L , Huang X , Wang T , Mao C , Li J , Wang J , Liu C , Gao J . Association of APOE ε4/ε4 with fluid biomarkers in patients from the PUMCH dementia cohort . Front Aging Neurosci. Frontiers Media SA ; 2023 ; 15 : 1119070 . 16. ↵ Shvetcov A , Johnson ECB , Winchester LM , Walker KA , Wilkins HM , Thompson TG , Rothstein JD , Krish V , Imam FB , Global Neurodegeneration Proteomics Consortium (GNPC), Burns JM , Swerdlow RH , Slawson C , et al. APOE ε4 carriers share immune-related proteomic changes across neurodegenerative diseases. Nat Med . Springer Science and Business Media LLC ; 2025 ; 31 : 2590 – 601 . OpenUrl 17. ↵ Johnson ECB , Dammer EB , Duong DM , Ping L , Zhou M , Yin L , Higginbotham LA , Guajardo A , White B , Troncoso JC , Thambisetty M , Montine TJ , Lee EB , et al. Large-scale proteomic analysis of Alzheimer’s disease brain and cerebrospinal fluid reveals early changes in energy metabolism associated with microglia and astrocyte activation . Nat Med. Springer Science and Business Media LLC ; 2020 ; 26 : 769 – 80 . OpenUrl 18. Johnson ECB , Carter EK , Dammer EB , Duong DM , Liu Y , Liu J , Betarbet R , Ping L , Yin L , Beach TG , Peng J , Gaiteri C , Bennett DA , et al. Large-scale deep multi-layer analysis of Alzheimer’s disease brain reveals strong proteomic disease-related changes not observed at the RNA level . Alzheimers Dement. Wiley ; 2021 ; 17 Suppl 3 : e055041 . OpenUrl 19. Seifar F , Fox EJ , Shantaraman A , Liu Y , Dammer EB , Modeste E , Duong DM , Yin L , Trautwig AN , Guo Q , Xu K , Ping L , Reddy JS , et al. Large-scale deep proteomic analysis in Alzheimer’s disease brain regions across race and ethnicity [Internet] . bioRxivorg . 2024 . Available from : doi: 10.1101/2024.04.22.590547 OpenUrl Abstract / FREE Full Text 20. Bai B , Wang X , Li Y , Chen P-C , Yu K , Dey KK , Yarbro JM , Han X , Lutz BM , Rao S , Jiao Y , Sifford JM , Han J , et al. Deep multilayer brain proteomics identifies molecular networks in Alzheimer’s disease progression . Neuron. Elsevier BV ; 2020 ; 106 : 700 . 21. Wang H , Dey KK , Chen P-C , Li Y , Niu M , Cho J-H , Wang X , Bai B , Jiao Y , Chepyala SR , Haroutunian V , Zhang B , Beach TG , et al. Integrated analysis of ultra-deep proteomes in cortex, cerebrospinal fluid and serum reveals a mitochondrial signature in Alzheimer’s disease . Mol Neurodegener. Springer Science and Business Media LLC ; 2020 ; 15 : 43 . 22. Wingo AP , Dammer EB , Breen MS , Logsdon BA , Duong DM , Troncosco JC , Thambisetty M , Beach TG , Serrano GE , Reiman EM , Caselli RJ , Lah JJ , Seyfried NT , et al. Large-scale proteomic analysis of human brain identifies proteins associated with cognitive trajectory in advanced age . Nat Commun. Springer Science and Business Media LLC ; 2019 ; 10 : 1619. 23. ↵ Seyfried NT , Dammer EB , Swarup V , Nandakumar D , Duong DM , Yin L , Deng Q , Nguyen T , Hales CM , Wingo T , Glass J , Gearing M , Thambisetty M , et al. A multi-network approach identifies protein-specific co-expression in asymptomatic and symptomatic Alzheimer’s disease . Cell Syst. Elsevier BV ; 2017 ; 4 : 60 – 72 .e4. OpenUrl 24. ↵ Roth TL , Sweatt JD . Annual Research Review: Epigenetic mechanisms and environmental shaping of the brain during sensitive periods of development . J Child Psychol Psychiatry. Wiley ; 2011 ; 52 : 398 – 408 . OpenUrl 25. Mathews HL , Janusek LW. Epigenetics and psychoneuroimmunology: mechanisms and models. Brain Behav Immun . Elsevier BV ; 2011 ; 25 : 25 – 39 . OpenUrl 26. Dall’Aglio L , Muka T , Cecil CAM , Bramer WM , Verbiest MMPJ , Nano J , Hidalgo AC , Franco OH , Tiemeier H . The role of epigenetic modifications in neurodevelopmental disorders: A systematic review . Neurosci Biobehav Rev . 2018 ; 94 : 17 – 30 . OpenUrl CrossRef PubMed 27. Kubota T , Takae H , Miyake K . Epigenetic mechanisms and therapeutic perspectives for neurodevelopmental disorders . Pharmaceuticals (Basel). MDPI AG ; 2012 ; 5 : 369 – 83 . OpenUrl 28. Article Navigation Journal Article Epigenetic alterations induced by environmental stress associated with metabolic and neurodevelopmental disorders. 29. Kubota T , Miyake K , Hariya N , Mochizuki K. Understanding the epigenetics of neurodevelopmental disorders and DOHaD . J Dev Orig Health Dis. Cambridge University Press (CUP) ; 2015 ; 6 : 96 – 104 . OpenUrl 30. Reichard J , Zimmer-Bensch G . The epigenome in neurodevelopmental disorders . Front Neurosci. Frontiers Media SA ; 2021 ; 15 : 776809 . 31. ↵ Lardenoije R , Iatrou A , Kenis G , Kompotis K , Steinbusch HWM , Mastroeni D , Coleman P , Lemere CA , Hof PR , van den Hove DLA , Rutten BPF. The epigenetics of aging and neurodegeneration . Prog Neurobiol. Elsevier BV ; 2015 ; 131 : 21 – 64 . OpenUrl 32. ↵ De Jager PL , Srivastava G , Lunnon K , Burgess J , Schalkwyk LC , Yu L , Eaton ML , Keenan BT , Ernst J , McCabe C , Tang A , Raj T , Replogle J , et al. Alzheimer’s disease: early alterations in brain DNA methylation at ANK1, BIN1, RHBDF2 and other loci. Nat Neurosci . Springer Science and Business Media LLC ; 2014 ; 17 : 1156 – 63 . OpenUrl 33. Lunnon K , Smith R , Hannon E , De Jager PL , Srivastava G , Volta M , Troakes C , Al-Sarraj S , Burrage J , Macdonald R , Condliffe D , Harries LW , Katsel P , et al. Methylomic profiling implicates cortical deregulation of ANK1 in Alzheimer’s disease . Nat Neurosci. Springer Science and Business Media LLC ; 2014 ; 17 : 1164 – 70 . OpenUrl 34. Gasparoni G , Bultmann S , Lutsik P , Kraus TFJ , Sordon S , Vlcek J , Dietinger V , Steinmaurer M , Haider M , Mulholland CB , Arzberger T , Roeber S , Riemenschneider M , et al. DNA methylation analysis on purified neurons and glia dissects age and Alzheimer’s disease-specific changes in the human cortex . Epigenetics Chromatin [Internet]. Springer Science and Business Media LLC ; 2018 ; 11 . Available from : doi: 10.1186/s13072-018-0211-3 OpenUrl CrossRef PubMed 35. ↵ Alves VC , Carro E , Figueiro-Silva J . Unveiling DNA methylation in Alzheimer’s disease: a review of array-based human brain studies . Neural Regen Res. Medknow ; 2024 ; 19 : 2365 – 76 . OpenUrl 36. ↵ Bennett DA , Buchman AS , Boyle PA , Barnes LL , Wilson RS , Schneider JA. Religious Orders Study and Rush Memory and Aging Project . J Alzheimers Dis. SAGE Publications ; 2018 ; 64 : S161 – 89 . OpenUrl 37. ↵ Yu L , Lutz MW , Farfel JM , Wilson RS , Burns DK , Saunders AM , De Jager PL , Barnes LL , Schneider JA , Bennett DA . Neuropathologic features of TOMM40’523 variant on late-life cognitive decline . Alzheimers Dement. Wiley ; 2017 ; 13 : 1380 – 8 . OpenUrl 38. ↵ Consensus recommendations for the postmortem diagnosis of Alzheimer’s disease. The National Institute on Aging, and Reagan Institute Working Group on Diagnostic Criteria for the Neuropathological Assessment of Alzheimer’s Disease . Neurobiol Aging . 1997 ; 18 : S1 – 2 . OpenUrl CrossRef PubMed Web of Science 39. ↵ Bennett DA , Schneider JA , Bienias JL , Evans DA , Wilson RS . Mild cognitive impairment is related to Alzheimer disease pathology and cerebral infarctions . Neurology. Ovid Technologies (Wolters Kluwer Health ); 2005 ; 64 : 834 – 41 . OpenUrl 40. ↵ Wightman DP , Jansen IE , Savage JE , Shadrin AA , Bahrami S , Holland D , Rongve A , Børte S , Winsvold BS , Drange OK , Martinsen AE , Skogholt AH , Willer C , et al. A genome-wide association study with 1,126,563 individuals identifies new risk loci for Alzheimer’s disease . Nat Genet. Springer Science and Business Media LLC ; 2021 ; 53 : 1276 – 82 . OpenUrl 41. ↵ Davis EJ , Solsberg CW , White CC , Miñones-Moyano E , Sirota M , Chibnik L , Bennett DA , De Jager PL , Yokoyama JS , Dubal DB . Sex-specific association of the X chromosome with cognitive change and tau pathology in aging and Alzheimer disease . JAMA Neurol. American Medical Association (AMA ); 2021 ; 78 : 1249 – 54 . OpenUrl 42. ↵ Lindquist JA , Mertens PR . Cold shock proteins: from cellular mechanisms to pathophysiology and disease . Cell Commun Signal. Springer Science and Business Media LLC ; 2018 ; 16 : 63 . 43. ↵ Zhang H , Li Q , Graham RK , Slow E , Hayden MR , Bezprozvanny I . Full length mutant huntingtin is required for altered Ca2+ signaling and apoptosis of striatal neurons in the YAC mouse model of Huntington’s disease . Neurobiol Dis. Elsevier BV ; 2008 ; 31 : 80 – 8 . OpenUrl 44. ↵ Smith AR , Smith RG , Burrage J , Troakes C , Al-Sarraj S , Kalaria RN , Sloan C , Robinson AC , Mill J , Lunnon K . A cross-brain regions study of ANK1 DNA methylation in different neurodegenerative diseases . Neurobiol Aging. Elsevier BV ; 2019 ; 74 : 70 – 6 . OpenUrl 45. ↵ Cao M , Li H , Zhao J , Cui J , Hu G . Identification of age-and gender-associated long noncoding RNAs in the human brain with Alzheimer’s disease . Neurobiol Aging. Elsevier BV ; 2019 ; 81 : 116 – 26 . OpenUrl 46. ↵ Liu H , Zhao J , Lin Y , Su M , Lai L . Administration of anti-ERMAP antibody ameliorates Alzheimer’s disease in mice . J Neuroinflammation. Springer Science and Business Media LLC ; 2021 ; 18 : 268 . 47. ↵ Kang M-J , Abdelmohsen K , Hutchison ER , Mitchell SJ , Grammatikakis I , Guo R , Noh JH , Martindale JL , Yang X , Lee EK , Faghihi MA , Wahlestedt C , Troncoso JC , et al. HuD regulates coding and noncoding RNA to induce APP→Aβ processing . Cell Rep. Elsevier BV ; 2014 ; 7 : 1401 – 9 . OpenUrl 48. ↵ D’Aoust LN , Cummings AC , Laux R , Fuzzell D , Caywood L , Reinhart-Mercer L , Scott WK , Pericak-Vance MA , Haines JL . Examination of candidate exonic variants for association to Alzheimer disease in the Amish . PLoS One. Public Library of Science (PLoS ); 2015 ; 10 : e0118043 . OpenUrl 49. ↵ Schuster C , Huard A , Sirait-Fischer E , Dillmann C , Brüne B , Weigert A . S1PR4-dependent CCL2 production promotes macrophage recruitment in a murine psoriasis model . Eur J Immunol. Wiley ; 2020 ; 50 : 839 – 45 . OpenUrl 50. ↵ Maphis NM , Jiang S , Binder J , Wright C , Gopalan B , Lamb BT , Bhaskar K . Whole genome expression analysis in a mouse model of tauopathy identifies MECP2 as a possible regulator of tau pathology . Front Mol Neurosci. Frontiers Media SA ; 2017 ; 10 : 69 . 51. ↵ Bronicki LM , Jasmin BJ . Emerging complexity of the HuD/ELAVl4 gene; implications for neuronal development, function, and dysfunction . RNA. Cold Spring Harbor Laboratory ; 2013 ; 19 : 1019 – 37 . OpenUrl 52. ↵ Szklarczyk D , Kirsch R , Koutrouli M , Nastou K , Mehryary F , Hachilif R , Gable AL , Fang T , Doncheva NT , Pyysalo S , Bork P , Jensen LJ , von Mering C. The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res . Oxford University Press (OUP) ; 2023 ; 51 : D638 – 46 . OpenUrl 53. ↵ Enright AJ , Van Dongen S , Ouzounis CA. An efficient algorithm for large-scale detection of protein families. Nucleic Acids Res . Oxford University Press (OUP) ; 2002 ; 30 : 1575 – 84 . OpenUrl 54. ↵ Ge T , Chen C-Y , Ni Y , Feng Y-CA , Smoller JW . Polygenic prediction via Bayesian regression and continuous shrinkage priors . Nat Commun. Springer Science and Business Media LLC ; 2019 ; 10 : 1776. 55. ↵ Zhou G , Zhao H . A fast and robust Bayesian nonparametric method for prediction of complex traits using summary statistics . PLoS Genet. Public Library of Science (PLoS ); 2021 ; 17 : e1009697 . OpenUrl 56. ↵ Langfelder P , Horvath S . WGCNA: an R package for weighted correlation network analysis . BMC Bioinformatics. Springer Nature ; 2008 ; 9 : 559 . 57. ↵ Snellman A , Tuisku J , Koivumäki M , Wahlroos S , Aarnio R , Rajander J , Karrasch M , Ekblad LL , Rinne JO . SV2A PET shows hippocampal synaptic loss in cognitively unimpaired APOE ε4/ε4 homozygotes . Alzheimers Dement . 2024 ; 20 : 8802 – 13 . OpenUrl CrossRef PubMed 58. ↵ He K , Li B , Wang J , Wang Y , You Z , Chen X , Chen H , Li J , Huang Q , Guo Q , Huang YH , Guan Y , Chen K , et al. APOE ε4 is associated with decreased synaptic density in cognitively impaired participants . Alzheimers Dement. Wiley ; 2024 ; 20 : 3157 – 66 . OpenUrl 59. ↵ Fan Y , Gao Y , Therriault J , Luo J , Ba M , Zhang H , Alzheimer’s Disease Neuroimaging Initiative. The effects of CSF neurogranin and APOE ε4 on cognition and neuropathology in mild cognitive impairment and Alzheimer’s disease . Front Aging Neurosci . 2021 ; 13 : 667899 . OpenUrl PubMed 60. ↵ Reiman EM , Caselli RJ , Chen K , Alexander GE , Bandy D , Frost J . Declining brain activity in cognitively normal apolipoprotein E ɛ4 heterozygotes: A foundation for using positron emission tomography to efficiently test treatments to prevent Alzheimer’s disease. Proc Natl Acad Sci U S A . Proceedings of the National Academy of Sciences ; 2001 ; 98 : 3334 – 9 . OpenUrl Abstract / FREE Full Text 61. Reiman EM , Caselli RJ , Yun LS , Chen K , Bandy D , Minoshima S , Thibodeau SN , Osborne D . Preclinical evidence of Alzheimer’s disease in persons homozygous for the epsilon 4 allele for apolipoprotein E . N Engl J Med. Massachusetts Medical Society ; 1996 ; 334 : 752 – 8 . OpenUrl 62. Reiman EM , Chen K , Alexander GE , Caselli RJ , Bandy D , Osborne D , Saunders AM , Hardy J . Correlations between apolipoprotein E epsilon4 gene dose and brain-imaging measurements of regional hypometabolism. Proc Natl Acad Sci U S A . Proceedings of the National Academy of Sciences ; 2005 ; 102 : 8299 – 302 . OpenUrl Abstract / FREE Full Text 63. Small GW , Ercoli LM , Silverman DH , Huang SC , Komo S , Bookheimer SY , Lavretsky H , Miller K , Siddarth P , Rasgon NL , Mazziotta JC , Saxena S , Wu HM , et al. Cerebral metabolic and cognitive decline in persons at genetic risk for Alzheimer’s disease. Proc Natl Acad Sci U S A . Proceedings of the National Academy of Sciences ; 2000 ; 97 : 6037 – 42 . OpenUrl Abstract / FREE Full Text 64. ↵ Small GW , Mazziotta JC , Collins MT , Baxter LR , Phelps ME , Mandelkern MA , Kaplan A , La Rue A , Adamson CF , Chang L . Apolipoprotein E type 4 allele and cerebral glucose metabolism in relatives at risk for familial Alzheimer disease . JAMA. American Medical Association (AMA ); 1995 ; 273 : 942 – 7 . OpenUrl 65. ↵ Liu Y , Sugiura Y , Lin W . The role of synaptobrevin1/VAMP1 in Ca2+-triggered neurotransmitter release at the mouse neuromuscular junction . J Physiol. Wiley ; 2011 ; 589 : 1603 – 18 . OpenUrl 66. ↵ Sevlever D , Zou F , Ma L , Carrasquillo S , Crump MG , Culley OJ , Hunter TA , Bisceglio GD , Younkin L , Allen M , Carrasquillo MM , Sando SB , Aasly JO , et al. Genetically-controlled Vesicle-Associated Membrane Protein 1 expression may contribute to Alzheimer’s pathophysiology and susceptibility . Mol Neurodegener. Springer Science and Business Media LLC ; 2015 ; 10 : 18 . 67. ↵ Zaltieri M , Grigoletto J , Longhena F , Navarria L , Favero G , Castrezzati S , Colivicchi MA , Della Corte L , Rezzani R , Pizzi M , Benfenati F , Spillantini MG , Missale C , et al. α-synuclein and synapsin III cooperatively regulate synaptic function in dopamine neurons . J Cell Sci. The Company of Biologists ; 2015 ; 128 : 2231 – 43 . OpenUrl 68. ↵ Fan L-Y , Yang J , Li M-L , Liu R-Y , Kong Y , Duan S-Y , Guo G-Y , Yang J-H , Xu Y-M . Single-nucleus transcriptional profiling uncovers the reprogrammed metabolism of astrocytes in Alzheimer’s disease . Front Mol Neurosci . 2023 ; 16 : 1136398 . 69. ↵ Bojjireddy N , Botyanszki J , Hammond G , Creech D , Peterson R , Kemp DC , Snead M , Brown R , Morrison A , Wilson S , Harrison S , Moore C , Balla T . Pharmacological and genetic targeting of the PI4KA enzyme reveals its important role in maintaining plasma membrane phosphatidylinositol 4-phosphate and phosphatidylinositol 4,5-bisphosphate levels . J Biol Chem. Elsevier BV ; 2014 ; 289 : 6120 – 32 . OpenUrl 70. ↵ Yang LG , March ZM , Stephenson RA , Narayan PS . Apolipoprotein E in lipid metabolism and neurodegenerative disease . Trends Endocrinol Metab . 2023 ; 34 : 430 – 45 . OpenUrl CrossRef PubMed 71. ↵ Oliver PL , Finelli MJ , Edwards B , Bitoun E , Butts DL , Becker EBE , Cheeseman MT , Davies B , Davies KE . Oxr1 is essential for protection against oxidative stress-induced neurodegeneration . PLoS Genet. Public Library of Science (PLoS ); 2011 ; 7 : e1002338 . OpenUrl 72. ↵ Menale C , Robinson LJ , Palagano E , Rigoni R , Erreni M , Almarza AJ , Strina D , Mantero S , Lizier M , Forlino A , Besio R , Monari M , Vezzoni P , et al. Absence of dipeptidyl peptidase 3 increases oxidative stress and causes bone loss . J Bone Miner Res. Oxford University Press (OUP) ; 2019 ; 34 : 2133 – 48 . OpenUrl 73. ↵ Volkert MR , Crowley DJ . Preventing neurodegeneration by controlling oxidative stress: The role of OXR1 . Front Neurosci. Frontiers Media SA ; 2020 ; 14 : 611904 . 74. ↵ Yagensky O , Kohansal-Nodehi M , Gunaseelan S , Rabe T , Zafar S , Zerr I , Härtig W , Urlaub H , Chua JJ . Increased expression of heme-binding protein 1 early in Alzheimer’s disease is linked to neurotoxicity. Elife [Internet]. eLife Sciences Publications , Ltd ; 2019 ; 8 . Available from : doi: 10.7554/eLife.47498 OpenUrl CrossRef PubMed 75. ↵ Bian Z , Yamashita T , Shi X , Feng T , Yu H , Hu X , Hu X , Bian Y , Sun H , Tadokoro K , Takemoto M , Omote Y , Morihara R , et al. Accelerated accumulation of fibrinogen peptide chains with Aβ deposition in Alzheimer’s disease (AD) mice and human AD brains . Brain Res. Elsevier BV ; 2021 ; 1767: 147569 . 76. ↵ Rieker C , Migliavacca E , Vaucher A , Baud G , Marquis J , Charpagne A , Hegde N , Guignard L , McLachlan M , Pooler AM . Apolipoprotein E4 expression causes gain of toxic function in isogenic human induced pluripotent stem cell-derived endothelial cells . Arterioscler Thromb Vasc Biol. Ovid Technologies (Wolters Kluwer Health ); 2019 ; 39 : e195 – 207 . OpenUrl 77. ↵ Miyauchi M , Matsumura R , Kawahara H . BAG6 supports stress fiber formation by preventing the ubiquitin-mediated degradation of RhoA . Mol Biol Cell. American Society for Cell Biology (ASCB ); 2023 ; 34 : ar34. 78. ↵ Tachibana M , Holm M-L , Liu C-C , Shinohara M , Aikawa T , Oue H , Yamazaki Y , Martens YA , Murray ME , Sullivan PM , Weyer K , Glerup S , Dickson DW , et al. APOE4-mediated amyloid-β pathology depends on its neuronal receptor LRP1 . J Clin Invest. American Society for Clinical Investigation ; 2019 ; 129 : 1272 – 7 . OpenUrl 79. ↵ Quinn JP , Kandigian SE , Trombetta BA , Arnold SE , Carlyle BC. VGF as a biomarker and therapeutic target in neurodegenerative and psychiatric diseases . Brain Commun. Oxford University Press (OUP) ; 2021 ; 3 : fcab261. 80. ↵ Sweet RA , MacDonald ML , Kirkwood CM , Ding Y , Schempf T , Jones-Laughner J , Kofler J , Ikonomovic MD , Lopez OL , Garver ME , Fitz NF , Koldamova R , Yates NA . Apolipoprotein E*4 (APOE*4) genotype is associated with altered levels of glutamate signaling proteins and synaptic coexpression networks in the prefrontal cortex in mild to moderate Alzheimer disease . Mol Cell Proteomics. Elsevier BV ; 2016 ; 15 : 2252 – 62 . OpenUrl 81. ↵ Qin W , Li F-Y , Liu W-Y , Li Y , Cao S-M , Wei Y-P , Li Y , Wang Q , Wang Q-G , Jia J-P . The genetic landscape of early-onset Alzheimer’s disease in China . Alzheimers Dement. Wiley ; 2025 ; 21 : e14486 . OpenUrl 82. ↵ Ohkubo N , Mitsuda N , Tamatani M , Yamaguchi A , Lee YD , Ogihara T , Vitek MP , Tohyama M . Apolipoprotein E4 stimulates cAMP response element-binding protein transcriptional activity through the extracellular signal-regulated kinase pathway . J Biol Chem. Elsevier BV ; 2001 ; 276 : 3046 – 53 . OpenUrl 83. ↵ Sheu K-FR , Brown AM , Haroutunian V , Kristal BS , Thaler H , Lesser M , Kalaria RN , Relkin NR , Mohs RC , Lilius L , Lannfelt L , Blass JP . Modulation by DLST of the genetic risk of Alzheimer’s disease in a very elderly population . Ann Neurol. Wiley ; 1999 ; 45 : 48 – 53 . OpenUrl 84. ↵ Minogue S , Waugh MG , De Matteis MA , Stephens DJ , Berditchevski F , Hsuan JJ . Phosphatidylinositol 4-kinase is required for endosomal trafficking and degradation of the EGF receptor . J Cell Sci. The Company of Biologists ; 2006 ; 119 : 571 – 81 . OpenUrl 85. ↵ Jain N , Ganesh S . Emerging nexus between RAB GTPases, autophagy and neurodegeneration . Autophagy. Informa UK Limited ; 2016 ; 12 : 900 – 4 . OpenUrl 86. ↵ Asiamah EA , Feng B , Guo R , Yaxing X , Du X , Liu X , Zhang J , Cui H , Ma J . The contributions of the endolysosomal compartment and autophagy to APOE ɛ4 allele-mediated increase in Alzheimer’s disease risk . J Alzheimers Dis. SAGE Publications ; 2024 ; 97 : 1007 – 31 . OpenUrl 87. ↵ Dong Y , Li T , Ma Z , Zhou C , Wang X , Li J . HSPA1A, HSPA2, and HSPA8 are potential molecular biomarkers for prognosis among HSP70 family in Alzheimer’s disease. Dis Markers . Hindawi Limited ; 2022 ; 2022: 9480398. 88. ↵ Lu K , Alcivar AL , Ma J , Foo TK , Zywea S , Mahdi A , Huo Y , Kensler TW , Gatza ML , Xia B . NRF2 induction supporting breast cancer cell survival is enabled by oxidative stress– induced DPP3–KEAP1 interaction . Cancer Res. American Association for Cancer Research (AACR ); 2017 ; 77 : 2881 – 92 . OpenUrl 89. ↵ Kerr F , Sofola-Adesakin O , Ivanov DK , Gatliff J , Gomez Perez-Nievas B , Bertrand HC , Martinez P , Callard R , Snoeren I , Cochemé HM , Adcott J , Khericha M , Castillo-Quan JI , et al. Direct Keap1-Nrf2 disruption as a potential therapeutic target for Alzheimer’s disease . PLoS Genet . 2017 ; 13 : e1006593 . OpenUrl CrossRef PubMed 90. ↵ Bennett DA , Wilson RS , Schneider JA , Evans DA , Beckett LA , Aggarwal NT , Barnes LL , Fox JH , Bach J . Natural history of mild cognitive impairment in older persons . Neurology. Ovid Technologies (Wolters Kluwer Health ); 2002 ; 59 : 198 – 205 . OpenUrl 91. ↵ Bennett DA , Schneider JA , Arvanitakis Z , Kelly JF , Aggarwal NT , Shah RC , Wilson RS . Neuropathology of older persons without cognitive impairment from two community-based studies . Neurology. Ovid Technologies (Wolters Kluwer Health ); 2006 ; 66 : 1837 – 44 . OpenUrl 92. ↵ Gagnon-Bartsch JA , Jacob L , Speed TP. Removing unwanted variation from high dimensional data with negative controls . Berkeley : Tech Reports from Dep Stat Univ California. Citeseer ; 2013 ;: 1 – 112 . 93. ↵ Thrush KL , Bennett DA , Gaiteri C , Horvath S , van Dyck CH , Higgins-Chen AT , Levine ME . Aging the brain: multi-region methylation principal component based clock in the context of Alzheimer’s disease. Aging (Albany NY). Impact Journals , LLC ; 2022 ; 14 : 5641 – 68 . OpenUrl 94. ↵ De Jager PL , Ma Y , McCabe C , Xu J , Vardarajan BN , Felsky D , Klein H-U , White CC , Peters MA , Lodgson B , Nejad P , Tang A , Mangravite LM , et al. A multi-omic atlas of the human frontal cortex for aging and Alzheimer’s disease research . Sci Data. Springer Science and Business Media LLC ; 2018 ; 5 : 180142 . 95. Li H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM [Internet] . arXiv [q-bio.GN] . 2013 . Available from: http://arxiv.org/abs/1303.3997 96. Sherry ST , Ward MH , Kholodov M , Baker J , Phan L , Smigielski EM , Sirotkin K. dbSNP: the NCBI database of genetic variation . Nucleic Acids Res. Oxford University Press (OUP) ; 2001 ; 29: 308 – 11 . 97. Benjamin D , Sato T , Cibulskis K , Getz G , Stewart C , Lichtenstein L . Calling Somatic SNVs and Indels with Mutect2 [Internet] . bioRxiv. bioRxiv ; 2019 . Available from : doi: 10.1101/861054 OpenUrl Abstract / FREE Full Text 98. Cingolani P , Platts A , Wang LL , Coon M , Nguyen T , Wang L , Land SJ , Lu X , Ruden DM . A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3. Fly (Austin) . Informa UK Limited ; 2012 ; 6 : 80 – 92 . OpenUrl 99. Cingolani P , Patel VM , Coon M , Nguyen T , Land SJ , Ruden DM , Lu X . Using Drosophila melanogaster as a model for genotoxic chemical mutational studies with a new program , SnpSift. Front Genet. Frontiers Media SA ; 2012 ; 3 : 35 . 100. Liberzon A , Birger C , Thorvaldsdóttir H , Ghandi M , Mesirov JP , Tamayo P . The Molecular Signatures Database (MSigDB) hallmark gene set collection . Cell Syst. Elsevier BV ; 2015 ; 1 : 417 – 25 . OpenUrl 101. ↵ Guintivano J , Aryee MJ , Kaminsky ZA . A cell epigenotype specific model for the correction of brain cellular heterogeneity bias and its application to age, brain region and major depression . Epigenetics. Informa UK Limited ; 2013 ; 8 : 290 – 302 . OpenUrl 102. ↵ Ritchie ME , Phipson B , Wu D , Hu Y , Law CW , Shi W , Smyth GK. limma powers differential expression analyses for RNA-sequencing and microarray studies . Nucleic Acids Res. Oxford University Press (OUP ); 2015 ; 43 : e47 . OpenUrl 103. ↵ Kolberg L , Raudvere U , Kuzmin I , Adler P , Vilo J , Peterson H. g:Profiler-interoperable web service for functional enrichment analysis and gene identifier mapping (2023 update). Nucleic Acids Res . Oxford University Press (OUP) ; 2023 ; 51 : W207 – 12 . OpenUrl 104. ↵ Tingley D , Yamamoto T , Hirose K , Keele L , Imai K . Mediation:RPackage for causal mediation analysis . J Stat Softw. Foundation for Open Access Statistic ; 2014 ; 59 : 1 – 38 . OpenUrl 105. ↵ Kuhn M . Building Predictive Models inRUsing thecaretPackage . J Stat Softw. Foundation for Open Access Statistic ; 2008 ; 28 : 1 – 26 . OpenUrl View the discussion thread. Back to top Previous Next Posted October 16, 2025. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. 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 Beyond the Genotype: A Multi-Omic Analysis of APOEe4’s Role in Alzheimer’s Disease Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv 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 Beyond the Genotype: A Multi-Omic Analysis of APOEe4’s Role in Alzheimer’s Disease Yaroslav Markov , Ahana Priyanka , Leqi Xu , Weiwei Wang , Kyra Thrush-Evensen , John Gonzalez , Daniel Borrus , Jessica Kasamoto , Raghav Sehgal , Grace Zou , Jenel Fraij , Becky C. Carlyle , Steve Horvath , David A Bennett , Hongyu Zhao , Christopher H. van Dyck , TuKiet T Lam , Morgan E. Levine , Albert T. Higgins-Chen bioRxiv 2025.10.16.682426; doi: https://doi.org/10.1101/2025.10.16.682426 Share This Article: Copy Citation Tools Beyond the Genotype: A Multi-Omic Analysis of APOEe4’s Role in Alzheimer’s Disease Yaroslav Markov , Ahana Priyanka , Leqi Xu , Weiwei Wang , Kyra Thrush-Evensen , John Gonzalez , Daniel Borrus , Jessica Kasamoto , Raghav Sehgal , Grace Zou , Jenel Fraij , Becky C. Carlyle , Steve Horvath , David A Bennett , Hongyu Zhao , Christopher H. van Dyck , TuKiet T Lam , Morgan E. Levine , Albert T. Higgins-Chen bioRxiv 2025.10.16.682426; doi: https://doi.org/10.1101/2025.10.16.682426 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 Pathology Subject Areas All Articles Animal Behavior and Cognition (7629) Biochemistry (17660) Bioengineering (13881) Bioinformatics (41911) Biophysics (21436) Cancer Biology (18578) Cell Biology (25482) Clinical Trials (138) Developmental Biology (13371) Ecology (19887) Epidemiology (2067) Evolutionary Biology (24302) Genetics (15599) Genomics (22483) Immunology (17728) Microbiology (40364) Molecular Biology (17163) Neuroscience (88537) Paleontology (666) Pathology (2830) Pharmacology and Toxicology (4821) Physiology (7637) Plant Biology (15129) Scientific Communication and Education (2045) Synthetic Biology (4290) Systems Biology (9817) Zoology (2269)