(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Brain aging patterns among nine neurological disorders: A case-control study [1]

['Chuang Liang', 'Department Of Artificial Intelligence', 'Nanjing University Of Aeronautics', 'Astronautics', 'Nanjing', 'Jiangsu', 'Key Laboratory Of Brain-Machine Intelligence Technology', 'Ministry Of Education', 'Godfrey Pearlson', 'Olin Neuropsychiatry Research Center']

Date: 2026-07 

In this study, structural MRI data from 45,900 healthy controls (HCs) and 2,698 patients with developmental disorders (attention-deficit/hyperactivity disorder [ADHD] and autism spectrum disorder [ASD]), addiction (alcohol use disorder [AUD], tobacco use disorder [TUD], and AUD&TUD-A&TUD), dementia (Alzheimer’s disease [AD], and mild cognitive impairment [MCI]) or other psychiatric disorders (schizophrenia [SZ], bipolar disorder [BP], and major depressive disorder [MDD]), were collected to generate PAD, along with transcriptome data. Then, we calculated the PAD difference between patient and HC as Cohen’s d effect sizes, derived from a linear model that accounted for age, age 2 , sex, and site, and further identified the interpretable brain patterns associated with the PAD difference for each diagnostic group. Finally, enrichment analyses was conducted to identify the biological function of genes relatively over- or underexpressed in association with these patterns. Results showed that while PAD was consistently greater across disorders, different brain disorders showed different degrees of abnormality, the highest effects in dementia (AD: d = 0.97, 95% confidence interval (CI) [0.82,1.13]; p < 0.001 and MCI: d = 0.45, 95% CI [0.34,0.56]; p < 0.001), followed by addiction (A&TUD: d = 0.84, 95% CI [0.44,1.23]; p < 0.001, TUD: d = 0.72, 95% CI [0.49,0.96]; p < 0.001, and AUD d = 0.62, 95% CI [0.39,0.84]; p < 0.001) and psychiatric disorders (SZ: d = 0.53, 95% CI [0.30,0.76]; p < 0.001, BP: d = 0.46, 95% CI [0.22,0.69]; p < 0.001 and MDD: d = 0.28, 95% CI [0.11,0.46]; p < 0.001), but not different from expected in developmental disorders (ASD: d = 0.06, 95% CI [−0.04,0.16]; p = 0.36) and ADHD: d = 0.01, 95% CI [−0.14,0.15]; p = 0.98). Furthermore, higher PAD values in patient groups were linked to specific spatial brain patterns, including the frontotemporal network in psychiatric disorders, default mode network-salience network-putamen-thalamus in addiction and fronto-occipital network in dementia. Prefrontal cortex involvement was common across disorders, and disorder-specific brain patterns associated genes were enriched in different biological processes. A limitation of our study is that psychiatric disorders and addiction have high comorbidity, and these potential confounders were not considered.

Data Availability: The main code used in this study is available at https://github.com/liangchuang11/Brain-age-prediction.git and has been archived with a citable DOI on Zenodo: https://doi.org/10.5281/zenodo.20743298 . The de-identified minimal data required to replicate the findings of this study have been made publicly available without restriction on the GitHub repository: https://github.com/liangchuang11/brain_age_de-identified-minimal_data.git and archived with a DOI on Zenodo: https://doi.org/10.5281/zenodo.20743260 . The multimodal image data of HC, developmental disorders (ADHD and ASD) and dementia (AD and MCI) used in the present study are publicly available from the following consortia: HCP ( https://www.humanconnectome.org/ ), GSP ( https://www.neuroinfo.org/gsp/ ), UKB ( https://www.fmrib.ox.ac.uk/ukbiobank ), ADHD-200 ( https://fcon_1000.projects.nitrc.org/indi/adhd200/ ), ABIDE ( https://fcon_1000.projects.nitrc.org/indi/abide/ ) and ADNI ( https://adni.loni.usc.edu/ ) consortia. Each consortium has its own data access policies and application procedures, which are detailed on their respective websites. The psychiatric disorders (SZ, BP and MDD) and addiction (AUD and TUD) data are protected and are not publicly available due to data privacy and IRB restrictions. Data access requests may be submitted via https://trendscenter.org/contact-us/ or by email to info@trendscenter.org for researchers who meet the criteria for access to data.

(a) The averaged gray matter volume (GMV) from each region of interest (ROI, augmented Schaefer-1016 brain atlas) were extracted from testing (ADHD, ASD, SZ, BP, MDD, AUD, TUD, A&TUD, AD, and MCI) and age-range matched training (healthy participants from HCP, GSP, and UKB) sets. (b) The performance of the extreme gradient boosting (XGBoost) model was verified on training set by 10-fold cross-validation. (c) The XGBoost model was applied to the testing set to generate individual brain age predictions and the PAD difference between specific patient group and HC was compared. (d) Identifying the brain patterns associated with the PAD difference in different diagnostic groups by incorporating interaction terms (Shapley Additive Explanations-SHAP × group) in multiple linear regression models. The age-corrected PAD was the dependent variable, while age, age 2 , sex, site, group, SHAP value, and SHAP × group were the independent variables. (e) Enrichment analyses on the genes that relatively over- or underexpressed in association with PAD difference T-map. ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; SZ, schizophrenia; BP, bipolar disorder; MDD, major depressive disorder; AUD, alcohol use disorder; TUD, tobacco use disorder; A&TUD, AUD and TUD; AD, Alzheimer’s disease; MCI, mild cognitive impairment; HC, healthy control; HCP, Human Connectome Project; GSP, Brain Genomics Superstruct Project; UKB, UK Biobank; PAD, predicted age difference.

In this study, structural magnetic resonance imaging (sMRI) data including HCs (n = 45,900) and common brain disorders (344 ADHDs, 484 ASDs, 152 SZs, 143 BPs, 258 MDDs, 155AUDs, 144 TUDs, 361 ADs, and 657 MCIs) were used to generate the individual brain age predictions in each diagnostic group (including patients and the matched HCs, Fig 1a and 1b ). Our aims including: (1) comparing PAD difference across different diagnostic groups, and among age and sex subgroups ( Fig 1c ); (2) validating PAD difference for consistency across datasets, atlas resolutions, and prediction models in each diagnostic groups; (3) identifying brain patterns associated with the PAD difference in each diagnostic group and evaluating the associations between the PAD and symptoms ( Fig 1d ); 4) identifying the biological function of genes relatively over or under expressed in association with identified brain patterns ( Fig 1e ). By using large imaging samples and incorporating both clinical and genetic analyses, we sought to systematically and comprehensively investigate the deviation from normative brain aging and the underlying neuromolecular mechanisms associated with the PAD difference for 9 common brain disorders. We hypothesized that common brain disorders exhibit distinct degrees of accentuated brain aging, each associated with a specific spatial brain pattern linked to unique gene expression profiles.

Although a number of brain disorders have been investigated using PAD metrics, several issues remain to be addressed. First, while some studies have included relatively large multi-disorder samples, sample size constraints remain a persistent challenge in brain age prediction research. This limits generalizability and repeatability and may be unable to uncover the full spectrum of brain aging variations across different brain disorders. Second, predicted brain age is typically studied as a single whole-brain measure, thereby neglecting specific spatial brain patterns underlying the brain age prediction, leading to limited biological interpretability of the PAD. Third, although genetic differences have been demonstrated to explain a portion of the inter-individual variability in PAD [ 15 ], the genetic mechanisms underlying specific brain patterns associated with PAD difference in common brain diseases remain unclear. To address the above limitations, we combined transcriptome and structural neuroimaging data to uncover the deviation from normative brain aging, examine spatial brain patterns associated with the PAD difference and related gene expression profiles among common brain disorders in large samples. This is an important step towards utilizing PAD as potential biomarkers to assist clinicians in disentangling shared and specific pathophysiological processes of common brain disorders.

Previous studies have demonstrated the phenomena of accentuated brain aging or delayed development in a variety of commonly occurring brain disorders. Brain age prediction based on gray matter in Alzheimer’s disease (AD) revealed that the pathological structural atrophy in AD is associated with accentuated aging, with PAD of +10 years [ 6 ]. Mild cognitive impairment (MCI), a possible precursor to AD, is also associated with increased PAD, which has been shown increase accuracy of predicting conversion of MCI to AD and may serve as a biomarker for early AD risk screening [ 7 ]. The large, worldwide Enhancing Neuro-Imaging Genetics through Meta-analysis (ENIGMA)-schizophrenia (SZ) also reported a significantly higher PAD in SZ, compared to healthy controls (HCs) [ 8 ]. Another large dataset-based structural brain age prediction study reported a moderate increase in PAD in bipolar disorder (BP) compared to HC [ 9 ]. In major depressive disorder (MDD), accentuated brain aging seems to be stage-dependent, occurring at illness onset and disappearing as the illness further advances [ 10 ]. Moreover, higher brain PAD is associated with health-related lifestyle factors, and consistently reported in individuals with alcohol use disorder (AUD) and tobacco use disorder (TUD) [ 11 ]. Attention-deficit/hyperactivity disorder (ADHD) and autism spectrum disorder (ASD) typically emerge in childhood with disorder-related brain changes being highly age-dependent, meaning that patterns of accentuated aging, typical development, or delayed development may switch between different age groups [ 12 – 14 ].

The difference between age predicted from structural or functional neuroimaging data and chronological age, predicted age difference (PAD), can be used as an index to quantify individuals’ deviation from a normative brain aging [ 1 , 2 ]. A positive PAD value indicates that an individual’s brain age is greater than their chronological age, termed accentuated aging, while a negative PAD implies a delay from expected aging. Brain age prediction based on machine learning techniques has been widely used in investigating brain disorders to assess whether these disorders cause deviations in brain aging pattern [ 3 – 5 ]. These observations can provide a starting point for understanding the underlying neuropathological mechanisms of brain disorders.

Post-mortem gene expression data from Allen Institute for Brain Science (AIBS) and PAD difference T-map were used to identify the biological function of genes relatively over- or underexpressed in association with this brain pattern. Specifically, we processed the regional microarray expression data from the Allen Human Brain Atlas (AHBA) [ 38 ] using the abagen toolbox ( https://github.com/netneurolab/abagen ) to map the transcriptomic data onto 1,016 parcellated brain regions from the augmented Schaefer-1016 (details on transcription data processing can be found in Text D in S1 Appendix ) to generate regional gene expression matrix (1,016 ROIs × 15,632 genes) as the predictive variable. PAD difference T-map (1,016 ROIs × 1) of each diagnostic group was used as response variable to identify the gene expression component related to these T-maps by partial least square (PLS) regression. The statistical significance of the variance explained by each component was assessed by permuting the response variables 1,000 times. The significance component with the highest explained variance for PAD difference T-map (PLS T ) was selected for subsequent analysis. The error in estimating PLS T weight for each gene was evaluated by bootstrapping, and the ratio of each gene’s weight to its bootstrap standard error was used to calculate the Z scores. Genes with Z > 3 and Z <−3 (p FDR < 0.05) were defined as PLS T + and PLS T − gene sets, respectively. Metascape [ 39 ] ( https://metascape.org/gp/index.html ) was used to identify gene ontology (GO) terms that significantly enriched in these gene sets. Redundancy elimination and visualization of GO term lists (REViGO) [ 40 ] was used to summarize and visualize gene ontology ( http://revigo.irb.hr ) by removing redundant terms ( Fig 1e ).

where f(x i ) represents the predicted brain age for a single testing sample x i ; E train [f(x)] represents the fixed baseline (average chronological age in training set); M represents the number of input features and is the SHAP value which represents the contribution of feature j to the model prediction of participant i. Moreover, to mitigate potential bias arising from correlations among input features, SHAP values were computed using TreeSHAP with the interventional feature perturbation scheme, which anchors the perturbation process in the empirical data distribution rather than relying on the independence assumption used in conventional perturbation strategies, thereby reducing attribution bias when features are correlated [ 37 ]. Then, a SHAP value matrix with sample size × number of input features (n × M) for testing set was generated. We incorporate interaction terms (SHAP × group) in multiple linear regression models to identify the brain features for PAD difference in each diagnostic group, where the age-corrected PAD was the dependent variable, while age, age 2 , sex, site, group, SHAP value, and SHAP × group were the independent variables. Subsequently, PAD difference T-map was obtained according to statistical t-values of the interaction coefficients. The brain regions with significant interaction coefficients (p FDR < 0.05) were considered as the brain regions associated with the PAD difference ( Fig 1d ).

Age-range matched training HCs were selected from HCP, GSP, and UKB (641 for ADHD, 1,061 for ASD, 2,164 for SZ, 2,177 for BP, 2006 for MDD, 1912 for AUD, 1912 for TUD,1805 for A&TUD, 34,308 for AD and MCI) generating 10 training sets for each diagnostic group. Patients were not included in the training or cross-validation procedures. Details on samples construction for training and testing can be found in Fig B in S3 Appendix . The sample overlap between the corresponding training sets of different diagnostic groups can be found in Table D in S2 Appendix . Schaefer atlas [ 29 ] with 1,000 cortical regions of interest (ROIs) plus 16 subcortical ROIs from Melbourne subcortical brain atlas [ 30 ] were combined to generate 1016-ROIs as the “augmented Schaefer (AS)-1016” [ 31 ]. Averaged GMV based on augmented Schaefer-1016 brain atlas were used as features in brain age prediction ( Fig 1a ). The extreme gradient boosting (XGBoost) algorithm based on gradient boosted decision trees was chosen to predict the brain age, with chronological age of individuals in the healthy training set serving as the prediction target. The parameters were optimized using a 10-fold cross-validation conducted within the healthy training set. The optimal number of training iterations was determined by evaluating the prediction error across 2,000 rounds and early stopping was applied if performance did not improve for 20 consecutive rounds [ 32 ]. The hyperparameters of XGBoost were set up with a maximum depth of 4 and learning rate of 0.05, with other parameters set to default. The performance of predictive model was estimated by correlating the predicted brain age and chronological age, as well as by computing the mean absolute error (MAE) and coefficient of determination (R 2 ), in the healthy training set. After optimization, the trained models were applied to independent patient samples to generate predicted brain age and calculate the PAD (deviation between predicted brain age and chronological age, Fig 1b and 1c ).

Neuroimaging data for the training of brain age prediction were obtained from 3 consortia, including HCP (n = 2,489) [ 16 ], GSP (n = 1,034) [ 17 ], and UKB (n = 39,679) [ 18 ]. ADHD (n = 344) and age-gender-quantity matched HC (n = 344) were obtained from ADHD-200 project [ 19 ]; ASD (n = 484) and matched HC (n = 484) from ABIDE II [ 20 ]; SZ (n = 152) and matched HC (n = 152), BP (n = 143) and matched HC (n = 143) from BSNIP-1 [ 21 ]; MDD (n = 258) and matched HC (n = 258) from Beijing Anding Hospital, West China Hospital of Sichuan, First Affiliated Hospital of Zhejiang and Henan Mental Hospital of Xinxiang [ 22 ]; AUD (n = 155) and TUD (n = 144) from Albuquerque [ 23 ]; AD (n = 361) and matched HC (n = 361), MCI (n = 657) and matched HC (n = 657) from ADNI [ 24 ]. The diagnosis of ADHD, ASD, SZ, BP, and MDD were based on the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) [ 25 ]. Participants with Alcohol Use Disorder Identification Test (AUDIT) > 7/Fagerström Test for Nicotine Dependence (FTND) > 7 were categorized as AUD/TUD [ 26 , 27 ] (A&TUD were from AUD and TUD with AUDIT > 7 and FTND > 7). The diagnosis of MCI and AD was based on the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS/ADRDA) criteria [ 28 ]. More detailed information on the diagnostic criteria and medication status for healthy participants and patients can be found in Text A in S1 Appendix . Demographic information of individuals participated in this study can be found in Table 1 . The scatterplot of ages used in each control subgroup can be found in Fig A in S3 Appendix . Gray matter volume (GMV) from sMRI generated by Statistical Parametric Mapping version 12 (SPM12) was used for brain age prediction. Details on imaging parameters and preprocessing pipeline can be found in Text B in S1 Appendix and Table A in S2 Appendix . In addition, details on the impact of 1.5T scanner acquisitions on predictive performance and PAD difference can be found in Text C in S1 Appendix and Tables B and C in S2 Appendix . The research program was approved by the institution review board at each local site. Written informed consents were obtained after providing a complete description of the research process to each participant.

For the Human Connectome Project (HCP) dataset, ethical approval was granted by the Washington University Institutional Review Board (IRB, 201204036). The Brain Genomics Superstruct Project (GSP) was approved by the Partners HealthCare IRB and the Harvard University Committee on the Use of Human Subjects in Research. The UK Biobank (UKB) study was conducted under a protocol approved by the North West Multi-centre Research Ethics Committee (reference: 16/NW/0274). All data included in the ADHD-200 dataset were obtained under IRB approvals at the respective contributing institutions, including Peking University (PKU), Kennedy Krieger Institute (KKI), NeuroIMAGE, New York University Child Study Center (NYU), Oregon Health and Science University (OHSU), University of Pittsburgh, Washington University in St. Louis, and Brown University. According to the International Neuroimaging Data-Sharing Initiative (INDI) protocol, no additional IRB approval was required for secondary analyses of de-identified publicly available data. The original studies included in the Autism Brain Imaging Data Exchange (ABIDE II) dataset were approved by the IRBs at each participating site (full site list available at https://fcon_1000.projects.nitrc.org/indi/abide/abide_II.html ). In accordance with INDI data usage policies, no additional IRB approval was required for secondary analyses of de-identified publicly available data. The Bipolar-Schizophrenia Network for Intermediate Phenotypes (BSNIP-1) study protocol was approved by the IRBs of Hartford Hospital, the University of Texas Southwestern Medical School, the University of Maryland, the University of Chicago, Wayne State University, and Harvard University. The MDD dataset was approved by the Ethics Committees of Beijing Anding Hospital, West China Hospital of Sichuan University, the First Affiliated Hospital of Zhejiang University, and Henan Mental Hospital of Xinxiang. The AUD and TUD datasets were approved by the University of New Mexico Human Research Review Committee. The Alzheimer’s Disease Neuroimaging Initiative (ADNI) protocol was approved by the Institutional Review Boards of all participating institutions. Detailed information regarding the participating ethics committees and institutional review boards is available at https://adni.loni.usc.edu/about/governance/#core-details . Written informed consents were obtained after providing a complete description of the research process to each participant. For participants younger than 18 years of age, written informed consent was obtained from their parents or legal guardians, with assent obtained from the participants where applicable.

Significant GO terms for biological processes that the PLS T + (a) and PLS T − (b) gene sets in each diagnostic group enriched were plotted in semantic space so that similar terms were represented close to each other. The top 8 significant GO terms were labeled for each analysis with redundant GO terms excluded by REViGO. Markers were scaled based on the log10 q-value (q represents the p-value after FDR correction) for the significance of each GO term. Large blue circles were highly significant, while small red circles were less so. SZ, schizophrenia; BP, bipolar disorder; MDD, major depressive disorder; AUD, alcohol use disorder; TUD, tobacco use disorder; AD, Alzheimer’s disease; MCI, mild cognitive impairment; GO, gene ontology.

The enrichment analyses (details can be found in the “Gene enrichment analyses” section) were conducted to identify the biological function of genes relatively over- or underexpressed in association with PAD difference T-map. The first (the third for MDD) partial least squares (PLS) component explained 8.1%, 5.4%, 5.1%, 7.2%, 7.0%, 6.3%, and 5.3% variance in SZ, BP, MDD, AUD, TUD, AD, and MCI, which were not observed by chance (permutation test, p = 0.006 for SZ, p = 0.014 for BP, p = 0.048 for MDD, p = 0.019 for AUD, p = 0.046 for TUD, p = 0.047 for AD, and p = 0.048 for MCI, Fig K in S3 Appendix ). PLS1 (PLS3 for MDD) scores were positively correlated with the PAD difference T-map in SZ (spatial permutation tests, r = 0.24, p spin = 0.002), BP (r = 0.21, p spin < 0.001), MDD (r = 0.22, p spin < 0.001), AUD (r = 0.21, p spin = 0.002), TUD (r = 0.22, p spin = 0.001), AD (r = 0.23, p spin < 0.001), and MCI (r = 0.22, p spin < 0.001), respectively (Fig L in S3 Appendix ). After correcting for enrichment terms (p FDR < 0.05) and removing redundant terms, the significant Gene Ontology (GO) biological processes for PLS T + and PLS T − gene sets enriched were shown in Fig 7a and 7b , in which no significant GO terms enriched in PLS T + gene set for BP. The top 8 significant GO terms (Tables O and P in S2 Appendix ) were labeled for each diagnostic group (only 3 significant GO terms were identified in PLS T − gene set for MCI). In PLS T + , the cytoplasmic translation, ribonucleoprotein complex biogenesis, and translation were the shared significant GO terms in psychiatric disorders (SZ and MDD); carboxylic acid catabolic process, small molecule biosynthetic process, and small molecule catabolic process were the shared in addiction (AUD and TUD); protein localization to membrane was the shared in dementia (AD and MCI). For PLS T − , the endocytosis, import into cell and membrane organization were the shared in psychiatric disorders (SZ, BP, and MDD); inorganic ion transmembrane transport and synaptic signaling were the shared in addiction (AUD and TUD); deoxyribonucleic acid (DNA) damage response and DNA metabolic process were the shared in dementia (AD and MCI).

Partial correlation between PAD and symptoms in each diagnostic group were calculated by linear models accounting for age, age 2 , sex, and site. Results ( Fig 6 ) showed that PAD was negatively correlated with MMSE in AD (r=−0.17, p = 0.02) and MCI (r=−0.24, p = 8.12e − 06), and with FTND in TUD (r=−0.21, p = 0.01). However, no association (p > 0.05) was observed between PAD and symptom in other brain disorders (including SZ, BP, MDD, and AUD). The high degree of correlations between the whole brain PAD (predicted by the whole brain 1,016 ROIs) and the identified brain PAD (predicted by the identified brain ROIs) in each diagnostic group (r = 0.73 for SZ, r = 0.49 for BP, r = 0.70 for MDD, r = 0.86 for AUD, r = 0.72 for TUD, r = 0.89 for AD, r = 0.42 for MCI) confirmed the effectiveness of the identified regions. The associations between the whole brain PAD and symptoms were largely replicated in the identified brain PAD, with negative association with MMSE in AD (r=−0.19, p = 5.52e − 03) and MCI (r=−0.27, p = 9.13e − 07) and no association in other brain disorders.

To identify brain features contributed to PAD difference in each diagnostic group (ADHD and ASD were excluded due to no PAD difference, A&TUD were excluded to prevent redundancy), we first calculated the SHAP value for each of the 1,016 features. Then, we estimated a multiple linear regression model with age-corrected PAD as the dependent variable, and age, age 2 , sex, site, group, SHAP value, and the interaction term SHAP × group as independent variables. The SHAP × group interaction captures the influence of each brain feature on PAD difference between patient and HC, and PAD difference T-maps were derived from the t-statistics of the corresponding interaction coefficients. PAD difference T-maps for each diagnostic group were displayed in Fig 5a with false discovery rate (FDR) corrected in Fig 5b , whose reproducibility across different scales of brain atlas was validated (Fig I in S3 Appendix ). Prefrontal cortex (PFC), middle temporal cortex (MTC), and inferior temporal cortex (ITC) were the shared brain regions with significant SHAP × group interaction within psychiatric group. PFC, superior/middle/inferior temporal cortex (STC/MTC/ITC), superior/middle occipital cortex (SOC/MOC), thalamus, and insula were identified in SZ; PFC, MTC, ITC, and lingual gyrus (LG) in BP; and PFC, MTC, ITC, and MOC in MDD. Within addiction group, PFC, precentral cortex (PC), cingulate cortex (CC), insula, putamen, thalamus, and temporal cortex (TC) were the identified shared brain regions. PFC, PC, CC, insula, putamen, thalamus, STC, MTC, ITC, MOC, inferior occipital cortex (IOC), LG, superior parietal cortex (SPC), and fusiform gyrus (FG) were identified in AUD; PFC, PC, CC, insula, putamen, thalamus, STC, MTC, and ITC in TUD. Within dementia group, PFC, MOC, and FG were the shared brain regions. PFC, SOC, MOC, FG, hippocampus, para-hippocampus, MTC, ITC, PC, anterior cingulate cortex (ACC), amygdala, caudate, insula, and thalamus were identified in AD; PFC, MOC, and FG for MCI. The identified brain regions are summarized in Tables L–N in S2 Appendix . Note that PFC was consistently identified for PAD difference across psychiatric disorders, addiction, and dementia. Moreover, the bootstrapping analysis showed that the identified brain features consistently exhibited high selection frequency across 1,000 resampling conditions, indicating their robustness and stability (details on bootstrapping analysis can be found in Text I in S1 Appendix and Fig J in S3 Appendix ).

We validated the robustness of PAD difference sequence pattern among diagnostic categories (dementia > addiction > psychiatric disorders > developmental disorders) when using different training sets (full age range HCP), scales of brain atlas (Schaefer-216/516/816) and prediction models (support vector regression: SVR; back propagation neural network: BPNN; and random forest: RF). Details on the implementation of all these datasets, atlases, and predictive models can be found in Text G in S1 Appendix . Ten-fold cross-validation confirmed the predictive performance of all models (Fig H in S3 Appendix ), which were used to predict the brain age in the diagnostic groups, followed by calculating the PAD difference between patient and HC. The Spearman rank correlation coefficient β was used to evaluate the consistency of the PAD difference sequence pattern among diagnostic categories across datasets, brain atlas resolutions, and prediction models (details can be found in Text H in S1 Appendix ). Results showed that the identified PAD difference sequence pattern among diagnostic group were highly consistent across training sets (β = 0.90, p = 8.80e − 04, Fig 4a and Table K in S2 Appendix ), brain atlas resolutions (β = 1.00/0.98/0.99 and all p < 1.00e − 16 for AS-216/516/816, Fig 4b and Table K in S2 Appendix ) and prediction models (β = 0.83/0.91/0.88 and p = 3.18e − 03/3.07e − 04/1.98e − 03 for SVR/BPNN/RF, Fig 4c and Table K in S2 Appendix ).

The PAD difference between patient and HC in different age range was also investigated by dividing into “young,” “adult,” and “old” subgroups (Fig F in S3 Appendix and Tables G–I in S2 Appendix ). Results ( Fig 3b ) showed that increased PAD relative to HC was most prominently observed in “young” dementia (including AD: d = 1.45, 95% CI [1.16,1.74]; p < 0.001* and MCI: d = 0.59, 95% CI [0.40,0.79]; p < 0.001*) and AUD (d = 0.82, 95% CI [0.41,1.24]; p < 0.001*), and “adult” psychiatric disorders (including SZ: d = 0.72, 95% CI [0.32,1.12]; p < 0.001*, BP: d = 0.48, 95% CI [0.07,0.88]; p = 0.03 and MDD: d = 0.49, 95% CI [0.18,0.80]; p = 2.31e − 03*) and addiction (including TUD: d = 0.82, 95% CI [0.40,1.23]; p < 0.001* and A&TUD: d = 0.99, 95% CI [0.29,1.69]; p = 0.01), with no PAD difference in “young” BP (d = 0.19, 95% CI [−0.22,0.61]; p = 0.39), “old” MDD (d = 0.10, 95% CI [−0.21,0.41]; p = 0.55) and developmental disorders (including ADHD and ASD). We further examined age × group interaction effects on PAD (details can be found in Text F in S1 Appendix ) and identified significant interactions for AD and MCI (Table J in S2 Appendix ), with estimated divergence ages of approximately 57 years for AD and 68 and 80 years for MCI (Fig G in S3 Appendix ). Moreover, PAD difference by sex was also calculated. Results ( Fig 3c ) showed that female PAD was significantly higher than male in SZ (d = 0.64, 95% CI [0.30,0.98]; p < 0.001*) and AUD (d = 0.62, 95% CI [0.26,0.97]; p < 0.001*), while lower than male in AD (d=−0.43, 95% CI [−0.64, −0.22]; p < 0.001*), with no sex difference in other brain disorders.

The PAD difference between patient and HC was calculated using linear regression accounting for age, age 2 , sex, and site as covariates. The PAD was higher in all the diagnostic groups than HC ( Fig 3a ), in which strong difference effects was observed in AD (d = 0.97, 95% confidence interval (CI) [0.82,1.13]; p < 0.001*) and A&TUD (d = 0.84, 95% CI [0.44,1.23]; p < 0.001*), moderate effects in TUD (d = 0.72, 95% CI [0.49,0.96]; p < 0.001*), AUD (d = 0.62, 95% CI [0.39,0.84]; p < 0.001*), SZ (d = 0.53, 95% CI [0.30,0.76]; p < 0.001*), BP (d = 0.46, 95% CI [0.22,0.69]; p < 0.001*) and MCI (d = 0.45, 95% CI [0.34,0.56]; p < 0.001*), small effects in MDD (d = 0.28, 95% CI [0.11,0.46]; p < 0.001*), and negligible effects in ASD (d = 0.06, 95% CI [−0.04,0.16]; p = 0.36) and ADHD (d = 0.01, 95% CI [−0.14,0.15]; p = 0.98). In addition, fraction amplitude of low-frequency fluctuation (fALFF) from functional MRI (fMRI) was also used to predict brain age under the same prediction framework (Details on fMRI preprocessing pipeline can be found in Text B in S1 Appendix ). Results showed that no PAD difference was observed in almost all the diagnostic groups (Fig E in S3 Appendix ).

The sample size and age distributions of training and testing sets for each diagnostic group are shown in Fig 2a , with no group differences (p = 0.79–0.99, two-sample t test) in age distribution between patient and HC. For each diagnostic group, an independent XGBoost model was built to predict brain age from averaged GMV extracted from augmented Schaefer-1016 brain atlas (1,000 cortical and 16 subcortical ROIs) through 10-fold cross-validation (details can be found in “Brain age prediction with age correction” section). Results showed excellent predictive performance of the model (without age correction: r = 0.68–0.88, mean absolute error-MAE = 1.65–4.97, coefficient of determination-R 2 = 0.45–0.76; after age correction: r = 0.89–0.93, MAE = 1.15–3.75, R 2 = 0.76–0.84, Fig C in S3 Appendix ), with consistently high performance and low coefficient of variation across folds (without age correction:0.9%–12.1%, after age correction:0.5%–10.5%, Tables E and F in S2 Appendix , details can be found in Text E in S1 Appendix ), indicating the effectiveness of the brain age prediction model and highlighting the necessity of age-bias correction. Then, the group-specific XGBoost models were applied to the corresponding independent 9 diagnostic groups to generate the predicted brain age, which was then correlated with age. Results showed that the correlation between predicted brain age and chronological age, as well as the MAE and R 2 , were consistently better in HC than in patients across all clinical groups (r = 0.75, MAE = 1.56, R 2 = 0.57 for ADHD, and r = 0.77 MAE = 1.37, R 2 = 0.60 for HC; r = 0.82, MAE = 3.26, R 2 = 0.63 for ASD and r = 0.83, MAE = 3.09, R 2 = 0.67 for HC; r = 0.85, MAE = 6.10, R 2 = 0.59 for SZ and r = 0.89, MAE = 4.56, R 2 = 0.75 for HC; r = 0.88, MAE = 5.21, R 2 = 0.72 for BP and r = 0.89, MAE = 4.56, R 2 = 0.75 for HC; r = 0.79, MAE = 5.24, R 2 = 0.58 for MDD and r = 0.83, MAE = 4.64, R 2 = 0.66 for HC; r = 0.75, MAE = 4.52, R 2 = 0.51 for AUD and r = 0.89, MAE = 3.60, R 2 = 0.75 for HC; r = 0.79, MAE = 4.67, R 2 = 0.59 for TUD and r = 0.89, MAE = 4.26, R 2 = 0.69 for HC; r = 0.79, MAE = 4.74, R 2 = 0.52 for A&TUD and r = 0.84, MAE = 3.22, R 2 = 0.70 for HC; r = 0.88, MAE = 3.36, R 2 = 0.71 for AD and r = 0.90, MAE = 2.59, R 2 = 0.77 for HC; r = 0.89, MAE = 2.81, R 2 = 0.70 for MCI and r = 0.91, MAE = 2.59, R 2 = 0.79 for HC, Fig 2b ). We also investigated the impact of sample size on the performance of brain age prediction and observed that the larger training sample size the better performance and the smaller variability (Fig D in S3 Appendix ).

Discussion

By combining large structural neuroimaging datasets, we found strong evidence that PAD was consistently larger in patients than in HC across diagnostic groups. Different brain disorders exhibited different degrees of aging, with the highest effects in dementia, followed by addiction and psychiatric disorders, but not in developmental disorders (dementia > addiction > psychiatric disorders > developmental disorders). These results were highly replicable across different training sets, brain atlas resolutions, and prediction models. The regional brain differences underlying PAD among psychiatric disorders (PFC, MTC, and ITC), addiction (PFC, PC, CC, insula, putamen, thalamus, and TC), and dementia (PFC, MOC, and FG) were associated with specific spatial brain patterns, but also with a shared region in PFC. Furthermore, we also found that the PAD was correlated with symptom in dementia, but not in addiction and psychiatric disorders. Gene enrichment analyses suggested that the gene sets related to brain patterns associated with the PAD difference in common brain disorders were enriched in different biological processes. This study represents an integrative effort to systematically compare PAD differences across 9 common brain disorders, and identify the brain patterns associated with these differences and the biological function of its associated genes, using relatively large datasets within a unified framework. This approach may help to disentangle the pathophysiological complexity of brain disorders from the perspective of brain aging.

The reliability of brain age prediction is a crucial influence on the validity of results regarding deviations from normal aging in common brain disorders. Lack of model accuracy, along with insufficient data quality or quantity, can introduce bias, causing PAD to reflect noise rather than true physiological variation [32]. In our study, the reliability of brain age prediction was aided by the large dataset, that incorporated several prominent neuroimaging datasets for model training, including HCP, GSP, and UKB. By encompassing a wide range of brain disorders, our investigation is one of the most comprehensive efforts in brain age prediction to date. Additionally, by comparing PAD in patients to age-, sex-, and quantity-matched controls, we minimized the influence of noise from model inaccuracies or data quality, thereby allowing us to better attribute the PAD to biologically relevant differences.

The different degrees of accentuated aging in different brain disorders are largely consistent with meta-analyses considering multiple brain disorders which reported the strongest effects in neurodegenerative disorders, especially AD, followed by psychiatric disorders where relatively more pronounced brain aging was measured in SZ than BP and MDD, and without case-control differences in PAD for developmental disorders (including ADHD and ASD) [32,41,42]. Prior neuroimaging research has reported significant accentuated brain aging in AUD and TUD, reflecting age-related GMV atrophy patterns similar to AD [43–45]. We have extended these reports by documenting the replicability of PAD difference across different datasets, brain atlas resolutions and prediction models, providing powerful supporting evidence for both the universality of accentuated brain aging among brain disorders and the specificity across individual disorders. Notably, the observation that the PAD pattern (dementia > addiction > psychiatric disorders > developmental disorders) largely mirrors the mean group ages of these diagnostic categories may reflect the interaction between aging-related brain vulnerability and disease processes. As the brain ages, age-related structural and physiological changes may reduce neural resilience, such that pathological processes superimposed on an aging neural substrate may exert amplified effects on brain structure [46]. In contrast, developmental disorders emerge during childhood or adolescence, a period characterized by heightened neuroplasticity, ongoing myelination, and robust compensatory mechanisms [47,48]. The developing brain possesses greater synaptic exuberance and more efficient repair and reorganization capabilities, which may buffer against the structural impact of disorder-related pathology. Moreover, fALFF-based features failed to detect PAD difference between patients and HC, suggesting that the brain aging patterns may be more sensitive to structural than to functional alterations. In addition, different age stages in common brain disorders exhibit varying degrees of accentuated aging patterns, suggesting that the effects of brain aging may vary across the stages of disorder progression, emphasizing the heterogeneous aging subgroups in common brain disorders. We found sex difference in brain aging measures associated with brain disorders, as previously reported [49], suggesting the critical importance of considering sex in disease investigations. We also observed significant correlations between PAD and symptom in dementia (including AD and MCI), but not in other disorders, suggesting that the progression of dementia may co-occur with and be influenced by accentuated brain aging, while the development of symptoms related to addiction and psychiatric disorders may be more closely associated with the neuropathology inherent to those conditions.

The shared brain regions associated with accentuated aging were frontotemporal cortex in psychiatric disorders; default mode network (DMN), salience network (SAN) and putamen-thalamus in addiction; and fronto-occipital cortex in dementia. GMV reduction in frontotemporal cortex has been widely reported in psychiatric disorders as well as being associated with usual aging [50–52]. Previous studies investigating the interaction of alcohol/tobacco use and aging on brain morphology have reported whole brain GMV decreases with age among both HC and AUD/TUD, but with greater GMV reductions in core brain regions of DMN and SAN, including PFC, PC, CC, insula, and TC for AUD/TUD [44,53–55]. In addition, aging effects in putamen and thalamus have been consistently reported in AUD and TUD [43,56]. Previous studies have reported local changes in the PFC, MOC, and FG in the early stages of dementia, with these abnormalities progressively extending to encompass broader brain regions as the disease advances, ultimately involving the entire brain [57,58], consistent with our results. Our results suggested that frontotemporal network, DMN-SAN-putamen-thalamus and fronto-occipital network were overlapping aging and disease-related brain variations captured by the premature aging hypothesis, which may serve as potential biomarkers for assessing the risk and progression of psychiatric disorders, addiction, and dementia, respectively. More importantly, when combining all brain disorders, aging, and disease-related brain variations in PFC were consistently identified across psychiatric disorders, addiction, and dementia, demonstrating that PFC was involved in both these processes. This dual sensitivity to aging and disease across brain disorders further indicates that common brain disorders might not be completely independent at the level of neuroanatomical phenotype associated with aging. If so, this may provide valuable insights into the underlying neural mechanisms contributing to the substantial epidemiological comorbidity observed among common brain disorders, from the perspective of brain aging.

Enrichment analyses identified the biological functions of relatively over- or underexpressed genes associated with brain patterns linked to PAD difference in common brain disorders. These functions included: translation and transport in psychiatric disorders; catabolic processes, transmembrane transport and synaptic signaling in addiction; and cellular protein localization and DNA damage response and DNA metabolic process in dementia. These findings are partially consistent with the prior literature. A previous study reported that aberrant dendritic spine morphology and function regulated by translation initiating genes were associated with psychiatric disorders, including SZ, BP, and MDD [59,60]. Moreover, psychiatric disorders are linked to membrane and receptor transport, with endocytosis being the primary mechanism governing membrane organization transport, and abnormalities in these processes have been reported [61]. The catabolic process involves the breakdown of substances to release energy utilized by cells and organisms, however, alcohol and tobacco use can significantly disrupt the homeostasis of energy metabolism at both the cellular and whole-body levels, leading to tissue damage [62,63]. Additionally, alcohol and tobacco use can significantly alter ion channel and receptor functions on neuronal membranes, leading to modified neurotransmitter release and receptor expression, particularly affecting synaptic signaling in dopamine and glutamate pathways [64,65]. These alterations not only reinforce addictive behaviors but also induce changes in neural plasticity, thereby solidifying long-term dependence [64]. The abnormal aggregation and mis-localization of amyloid β and Tau proteins have been reported in dementia, which can lead to neurotoxicity and synaptic dysfunction, thereby affecting neuronal health and cognitive function [66]. Spontaneous DNA damage arise from both endogenous sources (such as DNA metabolism) and exposure to exogenous damage-inducing agents, and the vast majority of these lesions are efficiently removed by conserved DNA repair systems that specifically recognize such damage [67]. However, many studies have reported the accumulation of unrepaired DNA damage during natural aging but also occurs at high levels during neurodegeneration due to defects in DNA damage response, ultimately leading to dementia [68]. Hence, our results suggested that the different biological processes identified in psychiatric disorders, addiction and dementia can explain the accentuated aging phenomenon driven by brain structural abnormalities in corresponding diseases, which may be the shared molecular genetic mechanisms between brain age gaps and brain disorders.

The main finding in our study, that neurodegenerative disorders have the largest PAD is consistent with the post-mortem literature documenting widespread neuronal loss, gliosis, and cortical thinning in dementing illness [69]. Addictions such as AUD and TUD involve toxic agents that result in neuronal injury and brain shrinkage [70], hence their PAD are similarly large. Post-mortem studies of SZ, BP, and MDD do not document universal evidence of neuronal loss/gliosis and their moderately increased PAD values maybe reflect more subtle neuropil retractions limited to fronto-temporal regions [50]. Finally, neuroimaging studies have shown that the differences in cortical structure vary with developmental stage in ADHD and ASD [71,72], which may be a potential reason for minimally changed PAD observed in this study.

Several limitations should be considered. One is that all the datasets included in this study were collected from multiple sites and brain age prediction models were trained on pooled multi-site data without explicit control for site effects at the prediction stage. However, the deconfounding approach, leave-one-site-out (LOSO) validation and site difference analysis of PAD demonstrate that site is not a major confounding factor for brain age prediction in this study, and that the observed PAD differences are unlikely to be driven by site-related acquisition effects (details on deconfounding, LOSO validation, and site difference analysis can be found in Text J in S1 Appendix, Figs M–O in S3 Appendix, and Tables Q–U in S2 Appendix). In addition, site and other confounding factors, including age, age2, and sex, were included in all statistical models, testing for group differences and clinical associations to further minimize confounding effects. Second, we processed a substantial number of cross-sectional MRI to evaluate the dynamic PAD difference across various age ranges within each diagnostic group. We acknowledge that using cross-sectional data to analyze a dynamic process is not optimal. However, previous studies demonstrated that cross-sectional and longitudinal approaches produce similar age-related patterns in both normal aging and brain disorders [73]. Third, psychiatric disorders, such as SZ, BP, and MDD have very large co-morbidities with AUD and TUD [74]. We did not account for these potential confounding effects which are well known to affect PAD. Fourth, the age-stratified analyses relied on data-driven subgroups, which may introduce arbitrary age boundaries, reduce statistical power, and complicate biological interpretation. Therefore, these analyses should be interpreted as exploratory. Fifth, although the use of age-range matched HC and disorder-specific training sets may reduce extrapolation error, it may also potentially limit model generalizability and hindering the comparability of PAD estimates across diagnostic groups. However, our site-wise performance analyses indicated excellent cross-site generalizability (Table S in S2 Appendix), and diagnostic group differences were evaluated using standardized effect sizes (Cohen’s d) rather than direct comparisons of PAD values. On the other hand, aligning the age range between training and testing sets may, to some extent, bring their distributions closer, which could reduce the strictness of out-of-distribution evaluation and therefore only provides limited assurance that PAD estimates remain unbiased under more heterogeneous clinical samples and acquisition conditions. Although the site difference analysis of PAD provides supporting evidence for a certain degree of out-of-distribution generalization capability, these results should still be interpreted with caution, as it does not fully capture more extreme distribution shifts that may exist in real-world clinical settings. Sixth, although TreeSHAP with interventional perturbation was used to mitigate bias introduced by multicollinearity, and bootstrapping analysis confirmed the robustness of the identified brain regions, considering the inherent spatial dependence among input features, the identified brain regions associated with PAD difference may be exploratory findings for the broad spatial patterns and should be interpreted with caution. Moreover, although demographic-matched data sets and additional covariate adjustments were used to reduce potential confounding effects, other demographic and biological factors, such as ethnicity, non-psychiatric chronic diseases, lifestyle behaviors, and environmental exposures, were not consistently available across all datasets and therefore could not be explicitly modeled in the present study. Consequently, residual confounding effects cannot be fully excluded, and future studies with more comprehensive phenotypic harmonization are needed. Finally, the gene expression analyses were based on post-mortem transcriptomic data from a limited number of donors (n = 6), as transcriptomic data for the brain disorder cohorts included in this study were not available, which may limit the generalizability of these findings.

Collectively, the present study utilized large datasets to evaluate PAD difference across multiple brain disorders compared to HC, and identified brain patterns associated with the PAD difference, as well as the biological functions of associated genes. The PAD was increased in common brain disorders, with the greatest effects in dementia, followed by addiction and psychiatric disorders, and no measurable effect in developmental disorders. Each brain disorder showed a specific brain pattern associated with the PAD difference, but with shared involvement of the PFC, whose associated genes were enriched in different biological processes. Our results suggested that the risk factors for different brain disorders may converge on neurological mechanisms related to brain aging processes, and the PFC may underlie the important neural processes with respect to the universality of accentuated brain aging. The biological processes of genes related to identified brain patterns in psychiatric disorders, addiction, and dementia may underlie the molecular substrates of accentuated aging phenomenon driven by brain structural abnormalities, which can help us to better understand brain alterations related to aging at a molecular level. From a clinical perspective, genetically-informed and spatially specific PAD patterns could be tested in future studies for their usefulness as biomarkers to guide critical clinical decision-making such as transitions between MCI and AD; differentiation of SZ and BP during the first episode of psychosis and persistence of ADHD into adulthood.

[END]
---
[1] Url: https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1004860

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/