Discovering Hidden Links: Harnessing Similarity Network Fusion to Reveal Common Clusters in Healthy Aging, Mild Cognitive Impairment, and Dementia

Discovering Hidden Links: Harnessing Similarity Network Fusion to Reveal Common Clusters in Healthy Aging, Mild Cognitive Impairment, and Dementia | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Discovering Hidden Links: Harnessing Similarity Network Fusion to Reveal Common Clusters in Healthy Aging, Mild Cognitive Impairment, and Dementia View ORCID Profile Taeko Bourque , View ORCID Profile Peter Zhukovsky , View ORCID Profile Cassandra Morrison , View ORCID Profile John A. E. Anderson , the Alzheimer’s Disease Imaging Initiative (ADNI) doi: https://doi.org/10.1101/2025.04.02.25325137 Taeko Bourque 1 Department of Cognitive Science, Carleton University Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Taeko Bourque Peter Zhukovsky 2 Brain Health Imaging Center, Center for Addiction and Mental Health , Toronto, ON Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Peter Zhukovsky Cassandra Morrison 3 Department of Psychology, Carleton University Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Cassandra Morrison John A. E. Anderson 1 Department of Cognitive Science, Carleton University 3 Department of Psychology, Carleton University Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for John A. E. Anderson For correspondence: johnanderson3{at}cunet.carleton.ca Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Cognitively Unimpaired (CU), Mild Cognitive Impairment (MCI), and Alzheimer’s Disease (AD) are diagnostic categories used to categorize degrees of cognitive impairment. Older adults experience brain changes associated with aging and cognitive decline at different ages and progress at varying rates, leading to heterogeneous patterns of cognitive decline. As a result, people within the same diagnostic category often exhibit significant differences in cognitive abilities and brain functions. The present study aimed to investigate how data-driven categories map onto diagnostic categories. Therefore, using data from 515 participants in the Alzheimer’s Disease Neuroimaging Initiative database and a multivariate clustering approach, Similarity Network Fusion, we combined brain imaging features (cortical thickness average, surface area, and volume) with cognitive measures (Alzheimer’s Disease Assessment Scale-Cognitive, Mini-Mental State Exam, Montreal Cognitive Assessment, Clinical Dementia Rating) in older adults who had a clinical diagnosis of CU, MCI, or AD. We identified four data-driven groups spanning a gradient of cognitive and neural severity. Group 1 (87% diagnosed with dementia) showed the most significant impairment, while Group 4 (96% cognitively unimpaired) showed minimal impairment. Groups 2 and 3 captured transitional stages, including an “at-risk” group with early neural and cognitive decline. The current study provides evidence that more nuanced, data-driven approaches can reveal commonalities in individuals’ etiology and underlying neurobiology across traditional diagnoses. These results may lead to more focal therapies and a better understanding of who is at risk for converting to dementia, allowing for earlier detection and treatment of cognitive decline. Introduction Dementia currently affects approximately 55 million people worldwide ( World Health Organization, 2025 ), a number that is expected to double every 20 years ( Prince et al., 2015 ). This fast-growing global health concern can be devastating not only to affected individuals but also to their families, caregivers, and economies. Dementia is used to label impaired daily functioning due to cognitive decline, with Alzheimer’s Disease (AD) accounting for approximately two-thirds of dementia diagnoses ( Jalbert et al., 2008 ; Karantzoulis & Galvin, 2011 ). Despite the advent of new amyloid-clearing drugs, these clinical options remain limited in effectiveness, making treatment options for dementia challenging ( Benninghoff & Perneczky, 2022 ; Buckley & Salpeter, 2015 ; Dyck et al., 2023 ). Identifying those at high risk for dementia in the earliest stages of the disease before extensive brain pathology has occurred would be pivotal in slowing disease progression. Mild Cognitive Impairment (MCI), also known as prodromal (i.e., the beginning of a condition when symptoms first appear) AD, is an early stage of AD and is used to label cognitive decline that is detectable by clinical measures but does not cause impaired functioning in daily tasks ( Baldeiras et al., 2008 ; Dawe et al., 1992 ; Lee, 2023 ; Linn et al., 1995 ; Petersen, 2016 ; Petersen et al., 1999 ). Individuals with MCI may remain stable over time or transition to AD, with an annual conversion rate from MCI to AD of approximately 10 to 18% ( Davatzikos et al., 2011 ; Landau et al., 2010 ; Peterson et al., 2009). Determining who will transition from MCI to AD or dementia remains challenging, particularly because of the heterogeneity of disease progression in neurodegenerative disorders; brain structure and function decline at different ages and rates in other people ( Dunn et al., 2022 ; Maheux et al., 2023 ; Young et al., 2018 ). For example, individuals with differing clinical labels (i.e., CU – cognitively unimpaired, MCI, AD) may have more similar neural and cognitive profiles than individuals within the same diagnostic category. Recent studies highlight this heterogeneity and emphasize the need for new approaches to further the theoretical and empirical understanding of dementia ( Bermejo-Pareja & Del Ser, 2024 ; Dunn et al., 2022 ; Maheux et al., 2023 ; Young et al., 2018 ). For example, Maheux et al. (2023) used disease progression models to predict progressive decline in people with all stages of AD. Disease progression models use computational and statistical methods to analyze patients’ various clinical and biomarker data and learn the variability of disease progression. Maheux et al. used this method on 4600 participants spanning four continents and found that AD Course Map provided a robust and generalizable method for predicting disease progression. Although MCI has been described as a prodromal stage of AD, this topic is controversial because of the heterogeneity in the presentation of both disorders and the unclear progression of MCI to AD ( Bermejo-Pareja & Del Ser, 2024 ; Dawe et al., 1992 ; Lee, 2023 ; Thompson & Hodges, 2002 ). The psychometric boundaries for an MCI diagnosis are evolving and not well standardized ( Lee, 2023 ; Swallow, 2020 ; Vega & Newhouse, 2014 ). Indeed, revised MCI criteria ( Albert et al., 2011 ) may blur the contrast between MCI and mild AD, which some argue comes at the cost of a potential AD diagnosis ( Lee, 2023 ; Morris, 2012 ). Practitioners may also be concerned with a diagnosis’s clinical and social factors ( Swallow, 2020 ). One approach that may help to overcome the heterogeneity among diagnoses is data-driven clustering. Data-driven clustering methods may reveal subgroups of older adults that exhibit greater similarities in brain or behavior that may be lost with traditional diagnostic categories. For example, promising work by Jacobs et al. (2021) demonstrated using Similarity Network Fusion (SNF; Wang et al., 2014 ) the possibility of studying new data-driven groups of children that transcended traditional diagnostic classification of Autism, Obsessive-Compulsive Disorder (OCD) and Attention-Deficit/Hyperactivity Disorder (ADHD). Through SNF, participants are grouped based on their similarities to other participants within and across different data types, resulting in clusters of individuals with the strongest resemblance. Grouping individuals based on shared neural and cognitive similarities may reveal more nuanced classifications of individuals on the normal to abnormal aging spectrum, which are not captured by the current diagnostic classifications. The objective of this study was to use SNF ( Jacobs et al., 2021 ; Wang et al., 2014 ) to refine traditional diagnostic categories of cognitive impairment in the aging brain by identifying groups of aging individuals that are most similar to each other based on brain and cognitive scores. We hypothesized that SNF would reveal distinct groups of older adults with varying profiles of cognitive decline, in which individuals within each cluster would share similarities in brain imaging and behavioral traits, regardless of their clinical diagnostic status. Furthermore, we anticipate that certain individuals categorized as CU might display preliminary signs of neurocognitive deterioration, potentially increasing their susceptibility to future decline. Methods Data used in preparing this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). The ADNI was launched in 2003 as a public-private partnership led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer’s disease (AD). Participants were selected from ADNI-1, ADNI-GO, ADNI-2, and ADNI-3 cohorts. The study received ethical approval from the review boards of all participating institutions. Written informed consent was obtained from participants or their study partners. Participants Full participant inclusion and exclusion criteria are available at www.adni-info.org . All participants were between the ages of 55 and 90 at baseline, with no evidence of depression. Cognitively healthy older adults exhibited no evidence of memory decline, as measured by the Wechsler Memory Scale, and no evidence of impaired global cognition, as measured by the Mini-Mental Status Examination (MMSE) or Clinical Dementia Rating (CDR). MCI participants scored between 24 and 30 on the MMSE, 0.5 on the CDR, and abnormal scores on the Wechsler Memory Scale. Dementia was defined as participants who had abnormal memory function on the Wechsler Memory Scale, an MMSE score between 20 and 26 and a CDR of 0.5 or 1.0 and a probable AD clinical diagnosis according to the National Institute of Neurological and Communicative Disorders and Stroke and the AD and Related Disorders Association criteria. The ADNI database had 5411 participants across three cohorts (CU, MCI, AD). However, participants were excluded from our analyses if they had missing MRI imaging data or missing cognitive data for either the Alzheimer’s Disease Assessment Scale-Cognitive (ADAS-Cog; Rosen et al., 1984 ), the Mini-Mental State Exam (MMSE; Folstein et al., 1975 ), the Montreal Cognitive Assessment (MoCA; Nasreddine et al., 2005 ), and the Clinical Dementia Rating (CDR; Morris, 1993 ). These criteria resulted in the data from 515 participants (all from either ADNI 1 or ADNI 3) being used for our study. Only the most recent time point for each participant was included. Individuals with higher education levels have lower risks of developing dementia, AD, or vascular dementia; thus, level of education is often used as an indicator of cognitive reserve ( Meng & D’Arcy, 2012 ; Stern et al., 1994 ; Wilson et al., 2019 ). Factors that may increase the risk of developing AD include sex (women are more at risk than men; Gao et al., 1998 ), age (increased age is an important risk factor; Gao et al., 1998 ), and a history of depression ( Ownby et al., 2006 ). The Hachinski Ischemic Score (HIS) identifies the vascular component of cognitive disorders and is commonly used for diagnosis of vascular dementia ( Hachinski et al., 1975 , 2012 ). Therefore, sex, age (years), education (total years), depression (measured as the total raw score of the Geriatric Depression Scale; GDS; Yesavage & Sheikh, 1986 ), and HIS ( Hachinski et al., 1975 ) were used to examine if there were demographic features that differed between the groups obtained from the SNF analysis. Table 1 shows the descriptive statistics of the data-driven groups (see Supplementary Material, Table S1 for descriptive statistics of diagnostic groups). View this table: View inline View popup Download powerpoint Table 1 Descriptive Statistics of Data-Driven Groups Clinical/Cognitive Assessments Four cognitive measures were selected to represent clinical features of neurodegeneration: ADAS-Cog, MMSE, MoCA, and the CDR. Total raw scores for each cognitive measure were used in the SNF analysis. These measures are all used to assess cognitive impairment, including questions that involve orientation to the current time and place, ability to recall and recognize words from lists, naming objects, performing multi-step commands, problem-solving, repeating items forwards and backwards, copying shapes, or general conversation ability. The MoCA is particularly effective at early detection of MCI, whereas the MMSE is easier to administer and more suited to assessing the severity and progression of cognitive impairment when screening for dementia. The ADAS-Cog is used to assess the level of cognitive dysfunction in AD and the CDR is commonly used to assess the severity of cognitive dysfunction in AD or other dementias. MRI Acquisition Parameters For three-quarters of ADNI 1 participants, MRI data were acquired on 1.5T scanners with a typical resolution of 1.25 mm x 1.25 mm x 1.2 mm and a field of view (FOV) of 240 mm x 240 mm x 160 mm. Imaging protocols included T1-weighted MP-RAGE sequences optimized for anatomical detail. For the remaining ADNI 1 participants and all ADNI 3 participants, MRI data were acquired using 3T scanners, providing higher resolution images, typically around 1.0 mm x 1.0 mm x 1.0 mm, with a similar or slightly larger FOV depending on the scanner and specific protocol. The 3T imaging protocols also included T1-weighted MP-RAGE sequences, which offer improved signal-to-noise ratio and contrast, beneficial for detailed structural analysis. Further details on the MRI acquisition parameters and protocols can be found at ADNI MRI ( https://adni.loni.usc.edu/data-samples/adni-data/neuroimaging/mri/ ). Image Analysis Cortical volume, thickness, and surface area were derived from T1-weighted images using the Desikan-Killiany Atlas in FreeSurfer (6.0). These derivatives were taken from the ADNI database. Imaging metrics for cortical thickness average, surface area (divided by intracranial volume), and cortical volume (divided by intracranial volume) were used as brain-based input features in the SNF analysis. P-tau181 data Tau protein phosphorylated at threonine 181 (P-tau181) measurements were taken from the ADNI database. The Clinical Neurochemistry Laboratory, University of Gothenburg, analyzed Plasma P-tau181 using the Single Molecule array (Simoa) technique ( Karikari, 2020 ). Further details on acquisition parameters and protocols can be found at https://adni.loni.usc.edu/help-faqs/adni-documentation/ . Similarity Network Fusion Similarity Network Fusion ( Wang et al., 2014 ) is a computational method for data integration that leverages common and complementary information in different types of data. Figure 1 illustrates how this method iteratively processes individual similarity networks (per data type), thus pruning or strengthening connections between participants for a final, fused participant similarity network representing all data types. Download figure Open in new tab Figure 1. Overview of the Similarity Network Fusion Approach ( Wang et al. 2014 ) Note. Panel A: Brain imaging (i.e., cortical thickness average, surface area, and cortical volume) and cognitive (i.e., ADAS-Cog, MMSE, MoCA, CDR) data are collected for all participants. Panel B: Euclidean distance matrices are calculated for each data type (i.e., cognitive, cortical thickness average, surface area, cortical volume) and then converted into similarity matrices which capture the pairwise similarities between data points. Panel C: similarity matrices form the basis of participant similarity networks; each node represents an individual and edges encode similarity strength based on each data type. Panel D: each network is iteratively updated with information from other networks. Panel E: a fused participant similarity network is created, retaining shared and modality-specific similarities. All analyses were performed using R Statistical Software (v4.3.1; R Core Team 2023 ; Jacobs et al., 2021 ). The scripts used for analyses have been made available at https://github.com/TaekoBourque/SNF-ADNI . We combined various brain imaging features (i.e., cortical thickness average, surface area, and cortical volume) with cognitive measures (i.e., ADAS-Cog, MMSE, MoCA, CDR) in older adults with a clinical diagnosis of CU, MCI, or AD. These structural imaging and cognitive data types were integrated using the SNF package (v2.3.1; Wang et al., 2021 ). Data Integration and Cluster Determination First, Euclidean distance matrices are calculated for each data type (i.e., cognitive, cortical thickness average, surface area, cortical volume). Following Wang et al. (2014) and Jacobs et al. (2021) , we chose the following hyperparameters: the number of neighbors was set to 18, the normalization hyperparameter to 0.8, and the number of iterations to 10. Subsequently, Euclidean distance matrices, in which lower values indicate greater similarity, are converted into affinity matrices (i.e., similarity matrices) which capture the pairwise similarities between data points. These similarity matrices form the basis of participant similarity networks; each node represents an individual and edges encode similarity strength based on each data type. Finally, a fused participant similarity network is created by iteratively updating each network with information from other networks. At each iteration, individual networks are updated by exchanging structural information with the others, ensuring that shared and modality-specific similarities are retained. To assess the contribution of each feature to the fused similarity structure, we compute Normalized Mutual Information (NMI), which quantifies the degree of overlap between a similarity matrix based on one feature and the final fused matrix (based on all features). NMI ranges from 0 (no mutual information) to 1 (perfect correlation), with higher values indicating a greater influence of that feature on the overall participant similarity structure. A silhouette plot (see Supplementary Material, Figure S1) was used to examine the similarity between participants within a given group and participants in all other groups. Silhouette widths varied; Group 1 (0.87) and Group 3 (0.88) had strong cluster structures, Group 2 (0.52) had a good cluster structure, and Group 4 (0.45) had a slightly weaker structure. Overall, the average silhouette width was 0.55, indicating good cluster structure. Silhouette plots provide a visual representation of object classification, and the silhouette width (−1 to +1) indicates “goodness of fit”; higher silhouette width scores indicate that objects within a cluster are more similar to each other than objects in neighboring clusters. Network Visualization (see Figure 2 , Panel A) was used to illustrate the participant similarity networks in a way that highlights participants’ data-driven group assignment and their clinical diagnosis. Download figure Open in new tab Figure 2. Relative Participant Similarities, Cognitive Scores, and Brain Scores for Data-Driven and Diagnostic Groups Note. CU = Cognitively Unimpaired; MCI = Mild Cognitive Impairment; AD = Alzheimer’s Disease. Panel A: Data-driven groups (left) and those same groups colored by clinical diagnoses (right). Panel B: Boxplots representing the cognitive z-scores for data-driven (left) and diagnostic (right) groups. ADAS = Alzheimer’s Disease Assessment Scale-Cognitive; MMSE = Mini-Mental State Exam; MoCA = Montreal Cognitive Assessment; CDR = Clinical Dementia Rating. Panel C: Mirroring the approach to plotting cognitive scores, z-scores for surface area, cortical volume, and thickness average are plotted in brain space using ggseg (DK atlas). Comparisons Among Data-Driven and Diagnostic Groups We conducted separate one-way ANCOVAs among data-driven groups and diagnostic groups (see Table 2 ) to provide a standardized effect size estimate (eta-squared) of group distinctions for all 210 model features contributing to participant similarity. The p-value was corrected for multiple comparisons using a false discovery rate (pFDR) of 5%. Separate t-tests were conducted among data-driven groups to examine the age and P-tau181 differences (see Figure 4 ) and p-values were corrected for multiple comparisons using a pFDR of 5%. View this table: View inline View popup Table 2. The Top 35 Ranking Model Features (by NMI) Contributing to Participant Similarity and Corresponding Statistical Summaries of Data-Driven and Diagnostic Groups Results SNF Shows Four Groups Spanning a Gradient of Cognitive and Neural Severity Four distinct groups were identified based on participant similarity matrices determined by 210 model features across 1000 iterations of resampling 80% of participants (see Figure 2 ). The four data-driven groups featured an average silhouette width of 0.55, indicating a good cluster structure. These four clusters spanned a gradient from most severely affected cognitively and neurologically (group 1) to relatively unaffected (group 4). The top 10 model features with the highest NMI scores included the surface area of the right lingual, right lateral orbitofrontal, right medial orbitofrontal, right middle temporal, as well as the cortical thickness average of the left rostral anterior cingulate, left and right superior temporal sulcus, left caudal middle frontal, left lateral occipital, and right supramarginal gyrus. We highlight the top 35 contributing to participant similarity ranked by NMI (see Table 2 ; c.f., Jacobs et al., 2021 ). These top features included cortical thickness average and surface area regions; there weren’t any cortical volume features and MMSE was the only cognitive feature. For all 210 SNF input features, see Supplementary Material, Table S2. SNF Reveals an “At-Risk” Group of Cognitively Unimpaired Older Adults Our clustering solution significantly overlaps diagnostic labels for dementia and MCI. Specifically, 87% of individuals in Group 1 were diagnosed with dementia, while the rest were diagnosed with MCI. In Group 2, 60.9% were diagnosed with MCI, 11.7% with dementia, and 24.7% were CU. In the CU groups, we see the most notable differences between diagnostic categories and model results. This group was divided into two: Group 4, which contained 96.2% CU, showed no signs of impairment on neuropsychological tests or brain indices, and Group 3, which contained 78% CU, and 21.2% MCI diagnoses, which exhibited early, statistically significant declines in these metrics compared to Group 4. This finding suggests that participants in Group 3 may be showing signs of incipient neuropathology and cognitive decline. Comparison of Top Contributing Features Between Data-Driven and Diagnostic Groups Separate one-way ANCOVAS showed a significant main effect of group (after accounting for the effects of age, sex, and depression) across the top 35 ranking model features. These top features included 30 measures of cortical thickness average, four measures of surface area, one cognitive measure, and zero measures of cortical volume (see Figure 3 and Table 2 ). Effect sizes were typically smaller, and fewer were significant in diagnostic groups (i.e., CU, MCI, and AD) relative to data-driven groups (i.e., Groups 1 through 4). In comparing data-driven groups across cognitive (see Figure 1 , panel B) and brain (see Figure 2 , panel C) features included in the SNF analysis, Group 1 exhibits the most cognitive impairment, Group 4 exhibits the least, and Groups 2 and 3 exhibit an intermediate level of impairment. Download figure Open in new tab Figure 3. Representation of Top 35 Ranking Model Features (by NMI) in Brain-Space Note. NMI (Normalized Mutual Information) ranges from 0 to 1, with higher values indicating a greater influence of that feature on the overall participant similarity structure. As shown in Table 2 , the top features included cortical thickness average and surface area regions. There weren’t any cortical volume features and MMSE was the only cognitive feature. Age and P-Tau181 Differences Among Data-Driven Groups Separate t-tests showed age and P-tau181 differences among data-driven groups. Group comparisons and relevant statistics can be found in Figure 4 . Briefly, we show that Groups 1 and 2 are older than Groups 3 and 4 and that there are significantly higher P-tau181 levels in Groups 1 and 2 than Group 3. Group 4 may be underpowered to detect that difference. Download figure Open in new tab Figure 4. Age and P-Tau181 Differences Among Data-Driven Groups Note. ns: p > 0.05, *: p <= 0.05, **: p <= 0.01, ***: p <= 0.001, ****: p <= 0.0001. All p-values were corrected for multiple comparisons using a pFDR of 5%. Panel A: age differences across data-driven and diagnostic groups. Groups 3 and 4, which exhibit fewer clinical symptoms, tend to be younger compared to groups more severely affected by neurodegeneration and cognitive impairment. Panel B: P-tau181 levels by group, showing that Group 1 has higher concentrations of P-tau181 than Group 3, while Group 2 also has elevated P-tau181 levels relative to Group 3. No significant differences in P-tau181 levels were observed between Groups 3 and 4, suggesting that the early decline detected by the SNF model may precede an increase in P-tau181 concentrations. Discussion Neurodegenerative diseases like AD and MCI can have highly heterogeneous brain changes and cognitive symptoms. Traditional diagnostic categories (i.e., CU, MCI, AD) oversimplify this complexity, and emphasize categorical differences rather than subtle changes by degree. This study employs Similarity Network Fusion (SNF) to integrate brain imaging and cognitive measures and address the challenge of reconciling diagnostic categories with a person’s underlying etiology and performance on neuropsychological tests. SNF revealed novel, data-driven groups of participants from the ADNI data, grouping individuals who shared more brain and cognitive similarities with each other than their diagnoses would suggest. We predicted that transdiagnostic groups of individuals would emerge based on brain and neuropsychological features, better reflecting their stage on the AD continuum based on brain and cognitive changes together. We also anticipated that some individuals classified as CU would be segregated into a different, more affected group. In line with our prediction that our data fusion approach would reveal different groupings than the diagnostic labels, our similarity network fusion analysis identified four distinct groups of individuals, with each group showing highly similar brain and neuropsychological profiles, varying in degree of neurodegeneration and cognitive function. These findings refine traditional diagnostic categories by revealing more nuanced divisions across the spectrum of cognitive decline. Importantly, the groups spanned a gradient of severity, with surface area, cortical thickness, and MMSE emerging as the most important distinguishing features. Group 1 included the most severely impaired individuals with the majority having an AD diagnosis. Group 2 was mostly participants with MCI and some AD participants. Group 3 was mostly CU, with some MCI; this group had early signs of neuropathology. Finally, Group 4 was largely CU, with some MCI. Interestingly, our analysis favored a four-group solution over the three traditional diagnostic categories. This additional group split the CU and MCI groups at the extreme end of the x-coordinates producing Group 3 (see Figure 2 ). This finding suggests that many individuals categorized as CU may carry early signs of MCI, nearly imperceptible by routine cognitive screening. Future studies should investigate whether subjective cognitive decline measures, such as the Cognitive Change Index, might further differentiate these groups. Increasing evidence supports the idea that individuals with subjective cognitive decline, previously labeled ‘the worried well,’ may be in the earliest stages of dementia ( Jessen et al., 2014 ; Mendonça et al., 2016 ; Rabin et al., 2017 ), and thus, Group 3 may be the group that is most likely to progress to cognitive decline. The data-driven groups were sensitive to peripheral markers of AD pathology, providing further evidence for the amyloid/tau/neurodegeneration (AT[N]) framework ( Jack Jr. et al., 2018 ) and showcasing the utility of a data-driven method in stratifying MCI according to the level of P-tau181 burden. Our SNF approach emphasizes a continuum model of cognitive and neuropathological decline rather than a sudden shift that aligns with diagnostic thresholds. Cortical thickness, especially in medial and lateral temporal regions and in precuneus were the most important features for distinguishing groups and was the best predictor of cognitive decline along a gradient consistent with the known progression of AD ( Schwarz et al., 2016 ). Surface area and volume metrics were useful for differentiating the most severely affected (e.g., Group 1) individuals from those unaffected (e.g., Group 4) but were less useful for discerning early changes. Of the cognitive measures, only the MMSE appeared among the top SNF features highlighting its diagnostic utility, despite its limited sensitivity to early decline. Future studies should explore whether other measures, such as the MoCA, provide additional insights into differentiating CU-stable from CU-progressors. Our findings also support prior research suggesting that cortical thinning is a more reliable marker of disease progression than surface area changes ( Bakkour et al., 2013 ; Dickerson et al., 2009 ; Williams et al., 2023 ). Our approach highlights the importance of multi-omic integration for offering insights which transcend single modalities and offer a sensitive approach to re-think traditional diagnostic categories. Specifically, we suggest that conventional diagnostic categories may miss key early variations in disease progression, particularly when traditional neuropsychological screening tools are used as a gatekeeping mechanism, determining whether other complementary diagnostic features may be included. Traditional diagnostic categories, especially those that rely on standard neuropsychological screening tools as a primary diagnostic step, may overlook crucial early signs of disease progression and fail to capture the full spectrum of disease variability, which can be observed through brain imaging metrics. High levels of heterogeneity among participants within diagnostic categories also pose a problem for clinical trials and could cause them to fail to detect subtle but important effects or require vastly more participants to boost power ( Reed et al., 2021 ; Shapiro & Usner, 2012 ). Using SNF-derived groups as a starting point in clinical trials could thus help to improve clinical trial sensitivity and provide more homogeneous subgroups for targeted interventions. Others have used a similar approach to fuse DNA, and molecular measures with cognitive decline in the Religious Orders Study and Rush Memory and Aging Project (ROS/MAP) dataset ( Yang et al., 2023 ) and revealed two subtypes associated with diffuse amyloid plaques. Using these approaches may thus offer more comprehensive views of the aging brain. Our findings mark an important step toward personalized medicine by demonstrating how MRI features, integrated with machine learning, can provide clinicians with tools to detect early cognitive decline. Cognitive and neural decline exist on a continuum, and detecting early signs of deterioration remains a challenge, particularly in patients who do not fit precisely into a single diagnostic category. Surface area and cortical thickness – integral to our model – are essential for capturing these early changes, but without MRI, they are impossible to assess. Limitations and Future Directions Previously, work by Jacobs et al. (2021) , emphasized a lack of overlap between clinical diagnosis and biological and neuropsychological burden ( Jack Jr. et al., 2018 ; Sperling et al., 2011 ) – this finding was not the case for the present results. Our results revealed a similar pattern between data-driven and clinical groups; however, our model emphasized a gradient and appears more sensitive to early change. It is important to note that the sample used is relatively homogeneous – these findings should be validated in other large datasets, such as the Ontario Neurodegenerative Disease Research Initiative (ONDRI; Farhan et al., 2017 ) or the Rush Memory and Aging Project (MAP; Bennett et al., 2012 ). Future work could extend input features to include genetic (i.e., APOE e4 positivity; Yang et al., 2023 ) and protein factors (i.e., abnormal accumulation of hyperphosphorylated tau protein; Yang et al., 2023 ) and explore how these profiles are affected by cognitive reserve or risk factors ( Stern 2002 , 2012 ). Emerging technologies such as ultra-low-field MRI (0.055T) and novel tools like SynthSR offer promising, cost-effective solutions for longitudinally monitoring brain health, even in clinical settings. These advances, coupled with rapid cortical thickness estimation algorithms like FASTsurfer, could allow real-time patient data integration into predictive models like the one outlined in our study, aiding in early diagnosis and tracking cognitive decline progression ( Shen et al., 2021 ). Conclusion Early identification of individuals who present normally on neuropsychology screening instruments, but whose neurological profiles suggest they are declining may allow for more rapid intervention and personalized treatments. The current study adds to recent work highlighting the heterogeneity within diagnostic groups ( Bermejo-Pareja & Del Ser, 2024 ; Dunn et al., 2022 ; Maheux et al., 2023 ; Young et al., 2018 ). We identified new data-driven groups that overlap diagnostic classifications and feature more similar profiles of brain and behavior. Cognitive impairment is a continuum, and the sharp delineations of diagnostic categories (i.e., CU, MCI, AD) do not allow for a more nuanced interpretation. The current study provides evidence that more nuanced, data-driven approaches can allow for better representation of a participant’s brain and behavior profile that could allow for earlier detection and treatment. Data Availability Data used in the preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database ( http://www.loni.ucla.edu/ADNI ). The scripts used for analyses have been made available at https://github.com/TaekoBourque/SNF-ADNI . http://www.loni.ucla.edu/ADNI https://github.com/TaekoBourque/SNF-ADNI We have no known conflict of interest to disclose. This project was supported by an NSERC Discovery Grant (DGECR-2022-00309) and a Canada Research Chair (Tier II, CRC-2020-00174) to JAEA Acknowledgments Data collection and sharing for the Alzheimer’s Disease Neuroimaging Initiative (ADNI) is funded by the National Institute on Aging (National Institutes of Health Grant U19 AG024904). The grantee organization is the Northern California Institute for Research and Education. In the past, ADNI has also received funding from the National Institute of Biomedical Imaging and Bioengineering, the Canadian Institutes of Health Research, and private sector contributions through the Foundation for the National Institutes of Health (FNIH) including generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. Metabolomics Consortium (ADMC) will include the following language: Data collection and sharing for this project was funded by the Alzheimer’s Disease Metabolomics Consortium (National Institute on Aging R01AG046171, RF1AG051550 and 3U01AG024904-09S4). Footnotes Data used in the preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database ( http://www.loni.ucla.edu/ADNI ). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in the analysis or writing of this report. ADNI investigators include (complete listing available at http://www.loni.ucla.edu/ADNI/Collaboration/ADNI Authorship list.pdf). References ↵ Albert , M. S. , DeKosky , S. T. , Dickson , D. , Dubois , B. , Feldman , H. H. , Fox , N. C. , … & Phelps , C. H. ( 2011 ). The diagnosis of mild cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease . Alzheimer’sdementiatia , 7 ( 3 ), 270 – 279 . doi: 10.1016/j.jalz.2011.03.008 OpenUrl CrossRef PubMed Web of Science ↵ Bakkour , A. , Morris , J. C. , Wolk , D. A. , & Dickerson , B. C . ( 2013 ). The effects of aging and Alzheimer’s disease on cerebral cortical anatomy: specificity and differential relationships with cognition . Neuroimage , 76 , 332 – 344 . doi: 10.1016/j.neuroimage.2013.02.059 OpenUrl CrossRef PubMed ↵ Baldeiras , I. , Santana , I. , Proença , M. T. , Garrucho , M. H. , Pascoal , R. , Rodrigues , A. , … & Oliveira , C. R. ( 2008 ). Peripheral oxidative damage in mild cognitive impairment and mild Alzheimer’s disease . Journal of Alzheimer’s Disease , 15 ( 1 ), 117 – 128 . doi: 10.3233/JAD-2008-15110 OpenUrl CrossRef ↵ Bennett , D.A , Schneider , J.A. , Buchman , A.S. , Barnes , L.L. , Boyle , P.A. , & Wilson , R.S . ( 2012 ). Overview and findings from the rush Memory and Aging Project . Current Alzheimer Research , 9 ( 6 ), 646 – 663 . doi: 10.2174/156720512801322663 OpenUrl CrossRef PubMed Web of Science ↵ Benninghoff , J. , & Perneczky , R . ( 2022 ). Anti-dementia medications and anti-alzheimer’s disease drugs: side effects, contraindications, and interactions . In NeuroPsychopharmacotherapy (pp. 1 – 10 ). Cham: Springer International Publishing. doi: 10.1007/978-3-319-56015-1_195-1 OpenUrl CrossRef ↵ Bermejo-Pareja , F. , & Del Ser , T . ( 2024 ). Controversial past, splendid present, unpredictable future: a brief review of Alzheimer disease history . Journal of Clinical Medicine , 13 ( 2 ), 536 . doi: 10.3390/jcm13020536 OpenUrl CrossRef PubMed ↵ Buckley , J. S. , & Salpeter , S. R . ( 2015 ). A risk-benefit assessment of dementia medications: systematic review of the evidence . Drugs & aging , 32 , 453 – 467 . doi: 10.1007/s40266-015-0266-9 OpenUrl CrossRef PubMed ↵ Davatzikos , C. , Bhatt , P. , Shaw , L. M. , Batmanghelich , K. N. , & Trojanowski , J. Q . ( 2011 ). Prediction of MCI to AD conversion, via MRI, CSF biomarkers, and pattern classification . Neurobiology of aging , 32 ( 12 ), 2322 – e19 . doi: 10.1016/j.neurobiolaging.2010.05.023 OpenUrl CrossRef ↵ Dawe , B. , Procter , A. , & Philpot , M . ( 1992 ). Concepts of mild memory impairment in the elderly and their relationship dementiatia – a review . International Journal of Geriatric Psychiatry , 7 ( 7 ), 473 – 479 . doi: 10.1002/gps.930070704 OpenUrl CrossRef ↵ Dickerson , B. C. , Feczko , E. , Augustinack , J. C. , Pacheco , J. , Morris , J. C. , Fischl , B. , & Buckner , R. L . ( 2009 ). Differential effects of aging and Alzheimer’s disease on medial temporal lobe cortical thickness and surface area . Neurobiology of aging , 30 ( 3 ), 432 – 440 . doi: 10.1016/j.neurobiolaging.2007.07.022 OpenUrl CrossRef PubMed Web of Science ↵ Dunn , R. , Clayton , E. , Wolverson , E. , & Hilton , A . ( 2022 ). Conceptualising comorbidity and multimorbiditydementiantia: a scoping review and syndemic framework . Journal of Multimorbidity and Comorbidity , 12 , doi: 10.1177/2633556522112843 OpenUrl CrossRef ↵ Farhan , S. M. , Bartha , R. , Black , S. E. , Corbett , D. , Finger , E. , Freedman , M. , … & ONDRI Investigators . ( 2017 ). The Ontario neurodegenerative disease research initiative (ONDRI) . Canadian Journal of Neurological Sciences , 44 ( 2 ), 196 – 202 . doi: 10.1017/cjn.2016.415 OpenUrl CrossRef ↵ Folstein , M. F. , Folstein , S. E. , & McHugh , P. R . ( 1975 ). “Mini-mental state”: a practical method for grading the cognitive state of patients for the clinician . Journal of psychiatric research , 12 ( 3 ), 189 – 198 . OpenUrl CrossRef PubMed Web of Science ↵ Gao , S. , Hendrie , H. C. , Hall , K. S. , & Hui , S . ( 1998 ). The relationships between age, sex, and the incidencedementiantia and Alzheimer disease: a meta-analysis . Archives of general psychiatry , 55 ( 9 ), 809 – 815 . doi: 10.1001/archpsyc.55.9.809 OpenUrl CrossRef PubMed Web of Science ↵ Hachinski VC , Iliff LD , Zilhka E , et al. ( 1975 ). Cerebral blood flowdementiantia . Archives of Neurology , 32 ( 9 ), 632 – 637 . doi: 10.1001/archneur.1975.00490510088009 OpenUrl CrossRef PubMed Web of Science ↵ Hachinski , V. , Oveisgharan , S. , Romney , A. K. , & Shankle , W. R . ( 2012 ). Optimizing the Hachinski ischemic scale . Archives of neurology , 69 ( 2 ), 169 – 175 . doi: 10.1001/archneurol.2011.1698 OpenUrl CrossRef PubMed ↵ Jack Jr , C. R. , Bennett , D. A. , Blennow , K. , Carrillo , M. C. , Dunn , B. , Haeberlein , S. B. , … & Silverberg , N. ( 2018 ). NIA-AA research framework: toward a biological definition of Alzheimer’s disease . Alzheimer’s & Dementia , 14 ( 4 ), 535 – 562 . doi: 10.1016/j.jalz.2018.02.018 OpenUrl CrossRef PubMed ↵ Jacobs , G. R. , Voineskos , A. N. , Hawco , C. , Stefanik , L. , Forde , N. J. , Dickie , E. W. , Lai , M.-C. , Szatmari , P. , Schachar , R. , Crosbie , J. , Arnold , P. D. , Goldenberg , A. , Erdman , L. , & Ameis , S. H . ( 2021 ). Integration of brain and behavior measures for identification of data-driven groups cutting across children with ASD, ADHD, or OCD . Neuropsychopharmacology: Official Publication of the American College of Neuropsychopharmacology , 46 ( 3 ), 643 – 653 . doi: 10.1038/s41386-020-00902-6 OpenUrl CrossRef PubMed ↵ Jalbert , J. J. , Daiello , L. A. , & Lapane , K. L . ( 2008 ). Dementia of the Alzheimer type . Epidemiologic reviews , 30 ( 1 ), 15 – 34 . doi: 10.1093/epirev/mxn008 OpenUrl CrossRef PubMed Web of Science ↵ Jessen , F. , Amariglio , R. E. , Van Boxtel , M. , Breteler , M. , Ceccaldi , M. , Chételat , G. , … & Subjective Cognitive Decline Initiative . ( 2014 ). A conceptual framework for research on subjective cognitive decline in preclinical Alzheimer’s disease . Alzheimer’s & Dementia , 10 ( 6 ), 844 – 852 . doi: 10.1016/j.jalz.2014.01.001 OpenUrl CrossRef PubMed ↵ Karantzoulis , S. , & Galvin , J. E . ( 2011 ). Distinguishing Alzheimer’s disease from other major formsdementiantia . Expert review of neurotherapeutics , 11 ( 11 ), 1579 – 1591 . doi: 10.1586/ern.11.155 OpenUrl CrossRef PubMed Web of Science ↵ Karikari , T. , Pascoal , T. , Ashton , N. , Janelidze , S. , Benedet , A. , & Rodriguez , J . ( 2020 ). Plasma phospho-tau181 as a biomarker for Alzheimer’s disease: development and validation of a prediction model using data from four prospective cohorts . Lancet Neurol , 19 ( 5 ), 422 – 433 . OpenUrl CrossRef PubMed ↵ Landau , S. M. , Harvey , D. , Madison , C. M. , Reiman , E. M. , Foster , N. L. , Aisen , P. S. , … & Jagust , W. J . ( 2010 ). Comparing predictors of conversion and decline in mild cognitive impairment . Neurology , 75 ( 3 ), 230 – 238 . doi: 10.1212/WNL.0b013e3181e8e8b8 OpenUrl CrossRef PubMed ↵ Lee , J . ( 2023 ). Mild cognitive impairment in relation to Alzheimer’s disease: An investigation of principles, classifications, ethics, and problems . Neuroethics , 16 ( 2 ), 16 . doi: 10.1007/s12152-023-09522-5 OpenUrl CrossRef ↵ Linn , R. T. , Wolf , P. A. , Bachman , D. L. , Knoefel , J. E. , Cobb , J. L. , Belanger , A. J. , … & D’Agostino , R. B. ( 1995 ). The ‘preclinical phase’ of probable Alzheimer’s disease: a 13-year prospective study of the Framingham cohort . Archives of neurology , 52 ( 5 ), 485 – 490 . doi: 10.1001/archneur.1995.00540290075020 OpenUrl CrossRef PubMed Web of Science ↵ Maheux , E. , Koval , I. , Ortholand , J. , Birkenbihl , C. , Archetti , D. , Bouteloup , V. , … & Durrleman , S. ( 2023 ). Forecasting individual progression trajectories in Alzheimer’s disease . Nature Communications , 14 ( 1 ), 761 . doi: 10.1038/s41467-022-35712-5 OpenUrl CrossRef PubMed ↵ Mendonça , M. D. , Alves , L. , & Bugalho , P . ( 2016 ). From subjective cognitive complaintsdementiantia: who is at risk?: a systematic review . American Journal of Alzheimer’s Disease & Other Dementias® , 31 ( 2 ), 105 – 114 . doi: 10.1177/1533317515592331 OpenUrl CrossRef PubMed ↵ Meng , X. , & D’arcy , C . ( 2012 ). Education dementiantia in the context of the cognitive reserve hypothesis: a systematic review with meta-analyses and qualitative analyses . PloS one , 7 ( 6 ), e38268 . doi: 10.1371/journal.pone.0038268 OpenUrl CrossRef PubMed ↵ Morris , J. C . ( 1993 ). The Clinical Dementia Rating (CDR) current version and scoring rules . Neurology , 43 ( 11 ), 2412 – 2412 . OpenUrl FREE Full Text ↵ Morris , J. C . ( 2012 ). Revised criteria for mild cognitive impairment may compromise the diagnosis of Alzheimer disease dementia . Archives of neurology , 69 ( 6 ), 700 – 708 . doi: 10.1001/archneurol.2011.3152 OpenUrl CrossRef PubMed ↵ Nasreddine , Z. S. , Phillips , N. A. , Bédirian , V. , Charbonneau , S. , Whitehead , V. , Collin , I. , … & Chertkow , H. ( 2005 ). The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment . Journal of the American Geriatrics Society , 53 ( 4 ), 695 – 699 . doi: 10.1111/j.1532-5415.2005.53221.x OpenUrl CrossRef PubMed Web of Science ↵ Ownby , R. L. , Crocco , E. , Acevedo , A. , John , V. , & Loewenstein , D . ( 2006 ). Depression and risk for Alzheimer disease: systematic review, meta-analysis, and metaregression analysis . Archives of general psychiatry , 63 ( 5 ), 530 – 538 . doi: 10.1001/archpsyc.63.5.530 OpenUrl CrossRef PubMed Web of Science ↵ Petersen , R. C . ( 2016 ). Mild cognitive impairment . CONTINUUM: lifelong Learning in Neurology , 22 ( 2 ), 404 – 418 . doi: 10.1212/CON.0000000000000313 OpenUrl CrossRef PubMed ↵ Petersen , R. C. , Smith , G. E. , Waring , S. C. , Ivnik , R. J. , Tangalos , E. G. , & Kokmen , E . ( 1999 ). Mild cognitive impairment: clinical characterization and outcome . Archives of neurology , 56 ( 3 ), 303 – 308 . doi: 10.1001/archneur.56.3.303 OpenUrl CrossRef PubMed Web of Science Petersen , R. C. , Roberts , R. O. , Knopman , D. S. , Boeve , B. F. , Geda , Y. E. , Ivnik , R. J. , … & Jack , C. R. ( 2009 ). Mild cognitive impairment: ten years later . Archives of neurology , 66 ( 12 ), 1447 – 1455 . doi: 10.1001/archneurol.2009.266 OpenUrl CrossRef PubMed Web of Science ↵ Prince , M. , Wimo , A. , Guerchet , M. , Ali , G. C. , Wu , Y. T. , & Prina , M. ( 2015 ). World Alzheimer Report 2015: The global impacdementiaentia: An analysis of prevalence, incidence, cost and trends . ( Doctoral dissertation, Alzheimer’s Disease International ). ↵ R Core Team ( 2023 ). R: A Language and Environment for Statistical Computing . R Foundation for Statistical Computing, Vienna, Austria . https://www.R-project.org/ ↵ Rabin , L. A. , Smart , C. M. , & Amariglio , R. E . ( 2017 ). Subjective cognitive decline in preclinical Alzheimer’s disease . Annual Review of Clinical Psychology , 13 ( 1 ), 369 – 396 . doi: 10.1146/annurev-clinpsy-032816-045136 OpenUrl CrossRef PubMed ↵ Reed , G. W. , Parikh , P. , & Nissen , S . ( 2021 ). Importance of Internal Variability in Clinical Trials of Cardiovascular Disease . Canadian Journal of Cardiology , 37 ( 9 ), 1404 – 1414 . doi: 10.1016/j.cjca.2021.06.020 OpenUrl CrossRef PubMed ↵ Rosen WG , Mohs RC , Davis KL . ( 1984 ). A new rating scale for Alzheimer’s disease . American Journal of Psychiatry . 141 , 1356 – 1364 . OpenUrl CrossRef PubMed Web of Science ↵ Schwarz , C. G. , Gunter , J. L. , Wiste , H. J. , Przybelski , S. A. , Weigand , S. D. , Ward , C. P. , … & Alzheimer’s Disease Neuroimaging Initiative . ( 2016 ). A large-scale comparison of cortical thickness and volume methods for measuring Alzheimer’s disease severity . NeuroImage: Clinical , 11 , 802 – 812 . doi: 10.1016/j.nicl.2016.05.017 OpenUrl CrossRef PubMed ↵ Shapiro , A. , & Usner , D. ( 2012 , July ). Clinical trials from the statistician’s side: Reducing variability . Retina Today. https://retinatoday.com/articles/2012-july-aug/clinical-trials-from-the-statisticians-side-reducing-variability (Accessed 17 March, 2025 ) ↵ Shen , X. N. , Huang , Y. Y. , Chen , S. D. , Guo , Y. , Tan , L. , Dong , Q. , & Yu , J. T . ( 2021 ). Plasma phosphorylated-tau181 as a predictive biomarker for Alzheimer’s amyloid, tau and FDG PET status . Translational psychiatry , 11 ( 1 ), 585 . doi: 10.1038/s41398-021-01709-9 OpenUrl CrossRef PubMed ↵ Sperling , R. A. , Aisen , P. S. , Beckett , L. A. , Bennett , D. A. , Craft , S. , Fagan , A. M. , … & Phelps , C. H. ( 2011 ). Toward defining the preclinical stages of Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease . Alzheimer’s & Dementia , 7 ( 3 ), 280 – 292 . doi: 10.1016/j.jalz.2011.03.003 OpenUrl CrossRef PubMed Web of Science ↵ Stern , Y . ( 2002 ). What is cognitive reserve? Theory and research application of the reserve concept . Journal of the international neuropsychological society , 8 ( 3 ), 448 – 460 . doi: 10.1017/S1355617702813248 OpenUrl CrossRef PubMed Web of Science ↵ Stern , Y . ( 2012 ). Cognitive reserve in ageing and Alzheimer’s disease . The Lancet Neurology , 11 ( 11 ), 1006 – 1012 . DOI: 10.1016/S1474-4422(12)70191-6 OpenUrl CrossRef PubMed Web of Science ↵ Stern , Y. , Gurland , B. , Tatemichi , T. K. , Tang , M. X. , Wilder , D. , & Mayeux , R . ( 1994 ). Influence of education and occupation on the incidence of Alzheimer’s disease . Jama , 271 ( 13 ), 1004 – 1010 . doi: 10.1001/jama.1994.03510370056032 OpenUrl CrossRef PubMed Web of Science ↵ Swallow , J . ( 2020 ). Constructing classification boundaries in the memory clinic: negotiating risk and uncertainty in constituting mild cognitive impairment . Sociology of Health & Illness , 42 , 99 – 113 . doi: 10.1111/1467-9566.13016 OpenUrl CrossRef PubMed ↵ Thompson , S. A. , & Hodges , J. R . ( 2002 ). Mild cognitive impairment: a clinically useful but currently ill-defined concept? . Neurocase , 8 ( 6 ), 405 – 410 . doi: 10.1076/neur.8.5.405.16182 OpenUrl CrossRef PubMed Web of Science ↵ Van Dyck , C. H. , Swanson , C. J. , Aisen , P. , Bateman , R. J. , Chen , C. , Gee , M. , … & Iwatsubo , T. ( 2023 ). Lecanemab in early Alzheimer’s disease . New England Journal of Medicine , 388 ( 1 ), 9 – 21 . doi: 10.1056/nejmoa2212948 OpenUrl CrossRef ↵ Vega , J. N. , & Newhouse , P. A . ( 2014 ). Mild cognitive impairment: diagnosis, longitudinal course, and emerging treatments . Current psychiatry reports , 16 , 1 – 11 . doi: 10.1007/s11920-014-0490-8 OpenUrl CrossRef ↵ Wang B. , Mezlini A. M. , Demir F. , Fiume M. , Tu Z. , Brudno M. , Haibe-Kains B. , Goldenberg A. ( 2021 ). SNFtool: Similarity Network Fusion . R package version 2.3.1 , https://CRAN.R-project.org/package=SNFtool . ↵ Wang , B. , Mezlini , A. M. , Demir , F. , Fiume , M. , Tu , Z. , Brudno , M. , Haibe-Kains , B. , & Goldenberg , A . ( 2014 ). Similarity network fusion for aggregating data types on a genomic scale . Nature Methods , 11 ( 3 ), 333 – 337 . doi: 10.1038/nmeth.2810 OpenUrl CrossRef PubMed ↵ Williams , M. E. , Elman , J. A. , Bell , T. R. , Dale , A. M. , Eyler , L. T. , Fennema-Notestine , C. , … & Kremen , W. S. ( 2023 ). Higher cortical thickness/volume in Alzheimer’s-related regions: Protective factor or risk factor? . Neurobiology of aging , 129 , 185 – 194 . doi: 10.1016/j.neurobiolaging.2023.05.004 OpenUrl CrossRef PubMed ↵ Wilson , R. S. , Yu , L. , Lamar , M. , Schneider , J. A. , Boyle , P. A. , & Bennett , D. A . ( 2019 ). Education and cognitive reserve in old age . Neurology , 92 ( 10 ), e1041 – e1050 . doi: 10.1212/WNL.0000000000007036 OpenUrl Abstract / FREE Full Text ↵ World Health Organization . ( 2025 ). Dementia . World Health Organization . https://www.who.int/health-topics/dementia#tab=tab_2 (Accessed 18 March, 2025 ) ↵ Yang , M. , Matan-Lithwick , S. , Wang , Y. , De Jager , P. L. , Bennett , D. A. , & Felsky , D. ( 2023 ). Multi-omic integration via similarity network fusion to detect molecular subtypes of ageing . Brain Communications , 5 ( 2 ) , doi: 10.1093/braincomms/fcad110 OpenUrl CrossRef ↵ Yesavage , J. A. , & Sheikh , J. I . ( 1986 ). 9/Geriatric depression scale (GDS) recent evidence and development of a shorter version . Clinical gerontologist , 5 ( 1-2 ), 165 – 173 . doi: 10.1300/J018v05n01_09 OpenUrl CrossRef PubMed ↵ Young , A. L. , Marinescu , R. V. , Oxtoby , N. P. , Bocchetta , M. , Yong , K. , Firth , N. C. , … & Alexander , D. C. ( 2018 ). Uncovering the heterogeneity and temporal complexity of neurodegenerative diseases with Subtype and Stage Inference . Nature communications , 9 ( 1 ), 4273 . doi: 10.1038/s41467-018-05892-0 OpenUrl CrossRef PubMed View the discussion thread. Back to top Previous Next Posted April 04, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Discovering Hidden Links: Harnessing Similarity Network Fusion to Reveal Common Clusters in Healthy Aging, Mild Cognitive Impairment, and Dementia Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Discovering Hidden Links: Harnessing Similarity Network Fusion to Reveal Common Clusters in Healthy Aging, Mild Cognitive Impairment, and Dementia Taeko Bourque , Peter Zhukovsky , Cassandra Morrison , John A. E. Anderson , the Alzheimer’s Disease Imaging Initiative (ADNI) medRxiv 2025.04.02.25325137; doi: https://doi.org/10.1101/2025.04.02.25325137 Share This Article: Copy Citation Tools Discovering Hidden Links: Harnessing Similarity Network Fusion to Reveal Common Clusters in Healthy Aging, Mild Cognitive Impairment, and Dementia Taeko Bourque , Peter Zhukovsky , Cassandra Morrison , John A. E. Anderson , the Alzheimer’s Disease Imaging Initiative (ADNI) medRxiv 2025.04.02.25325137; doi: https://doi.org/10.1101/2025.04.02.25325137 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 Geriatric Medicine Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4435) Dentistry and Oral Medicine (444) Dermatology (382) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1509) Epidemiology (15229) Forensic Medicine (30) Gastroenterology (1124) Genetic and Genomic Medicine (6600) Geriatric Medicine (668) Health Economics (997) Health Informatics (4538) Health Policy (1368) Health Systems and Quality Improvement (1613) Hematology (541) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15916) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (146) Nephrology (667) Neurology (6599) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1144) Occupational and Environmental Health (957) Oncology (3333) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (691) Primary Care Research (711) Psychiatry and Clinical Psychology (5447) Public and Global Health (9232) Radiology and Imaging (2198) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a00e366499b24193',t:'MTc3OTY0NTc1OA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();