Expression patterns of risk genes associated with three evolutionarily relevant syndromes in rhesus macaque and human brains Authors: Kalyani B Karunakaran, Meera Purushottam, N Balakrishnan, Biju Viswanath, Ken-ichi Amemori, Sanjeev Jain Published: 2026 ABSTRACT AbstractDepressive disorder (DD), Alzheimer’s disease (AD), and schizophrenia (SZ) are evolutionarily relevant traits that disrupt neural networks supporting affect and cognition. While genome-wide association studies have identified risk-related genes for these diseases, how the expression of these genes compare across species remains unclear. In this study, we examined the spatial and temporal expression of ~2000 disease-associated genes in human and rhesus macaque brains. Distinct cross-species signatures emerged. DD-linked genes showed broad cortical-subcortical expression in humans but were confined to subcortical regions in macaques. The divergent subset was enriched for neuron differentiation, migration, synaptic signalling, and cognition. SZ-linked genes were expressed across cortical-subcortical-cerebellar structures in humans. AD-linked genes showed postnatal cortical-hippocampal macaque expression, and broader cortical-subcortical human expression. Cross-species spatial comparisons revealed a significant negative correlation for DD genes, suggesting a broader spatial distribution of DD-related gene expression in humans, extending to distributed emotion-cognition networks, compared to affective hubs in macaques. SZ genes exhibited a similar, though non-significant, negative trend, while AD genes showed a weak, non-significant correlation, indicating an absence of systematic expression shifts. Together, evolutionary shifts in gene expression may have shaped emotional and cognitive functions in humans, and susceptibility to psychiatric and neurodegenerative disorders.Subject terms: Depression, Schizophrenia, Genomics, Neuroscience Introduction Neural functions that underlie various psychiatric and neurological syndromes are shaped by complex evolutionary forces. Comparing the molecular signatures of brain regions in humans and non-human primates (NHPs) is important to identify shared and species-specific evolutionary patterns underlying these conditions. In this study, we focused on three syndromes suspected to have different evolutionary trajectories, namely, depressive disorder (DD), Alzheimer’s disease (AD), and schizophrenia (SZ). We examined how the expression of genes linked to the risk of developing these conditions varies across different regions and developmental stages in human and rhesus macaque (Macaca mulatta) brains. Rhesus macaques are apt models for this comparative analysis. They share many neurodevelopmental features with humans, albeit with crucial divergent aspects [1–4]. They are widely studied as NHP models for anxious temperament and depression in ethological and behavioural paradigms using functional magnetic resonance imaging (fMRI), positron emission tomography and microstimulation [5–8]. Multiple studies indicate that examining the spatiotemporal specificities of trait-linked genes in macaque and human brains could help identify the underlying neurocircuitry. In human brains, functional connectivity of regions significantly overlaps with regional gene co-expression patterns [9–14]. Similarly, in macaque brains, expression signatures covary with structural variation, including in regions that diverge from the human brain [15, 16]. The experience of depression, and its pathological presentation, DD, are rooted in affective behaviours conserved across species, particularly primates [17–19]; in the Discussion section, where possible, we maintain this distinction in alignment with the original studies, i.e., between the experience of depression, depression-like behaviours, depressive states, and the severe clinical presentation as DD. In contrast, AD and SZ may primarily affect human-specific cognitive systems [20, 21]. However, the molecular signatures of these syndromes have not been systematically compared across human and macaque brains. Previous comparative transcriptomic studies, such as from PsychENCODE and BICCN, identified human-specific and conserved expression signatures across cortical, subcortical and cerebellar areas [22–24]. Several autism- and SZ-linked genes showing distinct spatial and temporal patterns in humans were identified [22]. Several factors prompted us to select DD, AD and SZ for comparison across the two species. Affective modulation is an ancient and well-established response across many species [19]. Expression of emotion plays a significant role in communication and bonding, which are essential for formation of social units, social cohesion, and species survival [17]. As a broad behavioural repertoire, affective responsivity emerges during infancy and the resulting engrams influence rearing, sexual interactions, pregnancy, and social dominance. The traits that underlie communication of distress, depression and reinforcement of social support may thus undergo positive selection [18]. Shared substrates – from genes to neural circuits – perhaps underlie the experience of affects across species [25, 26]. These are likely to be even more congruent in closely related clades, such as primates [18], where comparable motor, cognitive, and perceptual systems are activated during emotional distress. In contrast, AD as a clinical syndrome is observed in elderly humans. Its clinical analogues have not been identified in NHPs, although AD-related pathologies have been noted in chimpanzees [27]. Longer human lifespan contributing to increased neuronal aging, antagonistic pleiotropy, modulation of immune responses, diet and patterns of recent brain growth may underlie this difference [28–30]. Similarly, the persistence of alleles linked to SZ risk in humans has been attributed to their pleiotropic effects on non-psychiatric phenotypes, such as blood pressure, lung function, etc [31–34]. These may have conferred adaptive advantages in living conditions as humans dispersed across continents. These traits were perhaps subjected to negative selection pressures [35], or showed a complex ‘evolutionary paradox’ [36]. Many SZ- and dementia-related behaviours thus have poor correlates in other primates. A wide range of (rare and common) genetic risks and neural circuits that evolved relatively recently in hominids [21, 37] may underlie these traits. The neural circuits associated with DD, AD and SZ may therefore be differently organized across humans and other primates. The mechanisms underlying these syndromes are also likely regulated across critical life stages (e.g., early prenatal period and adolescence in the dual hit hypothesis of SZ [38]). The expression of genes peak in specific developmental stages, and disease-associated genetic variations may perturb these temporal trajectories, consequently affecting the developmental fates of neural circuitries (e.g., prefrontal cortex in the SZ hypothesis) and signalling systems. Genome-wide association studies (GWASs) identify genes carrying variants contributing to the risk of these disorders. We examined, across human and macaque brains, the spatiotemporal expression patterns of ~2000 genes linked to the risk of developing DD, AD, and SZ (Fig. 1). We hypothesised that the expression patterns of AD and SZ may differ between the two species, while that of DD may be similar, aligning closely with conserved circuitry. Contrary to this expectation, we found that the expression pattern of DD genes diverged sharply and significantly across species, being broadly distributed in humans but restricted in macaques. In contrast, AD and SZ genes showed relatively diffuse patterns in both species, with SZ showing a trend towards divergence that did not reach significance. Graphical abstract outlining the analyses conducted in the study. Genes harbouring variants associated with Alzheimer’s disease, depressive disorder and schizophrenia were collected from the GWAS catalog. The spatiotemporal expression profiles of these genes were collectively examined in rhesus macaque and human brains to identify clusters of genes with specific temporal and regional signatures and disorder affiliations. Three methods were employed for identifying clusters: (I) hierarchical clustering, (II) t-SNE), and (III) PCA. For functional characterization of the temporal and spatial clusters, their enrichment for GO biological processes was analysed. Results We compiled a list of 438, 1178, and 2293 genes that lie within 50 kb of SNPs (p-value < 10−5) linked to DD, AD and SZ risk, respectively, from the GWAS catalog (Supplementary Data File 1 and Supplementary Methods) [39]. We included variants at both suggestive (p < 10⁻⁵) and genome-wide (p < 10⁻⁸) significance levels because we intended to capture the polygenicity that is intrinsic to these syndromes [40–42], rather than single-gene effects. Direct SNP-level comparisons are complicated by differences in genomic structure between the two species. Therefore, we mapped these SNPs to genes and examined their expression signatures in human and macaque brains using multiple clustering and dimensionality reduction methods (results summarised in Table 1). We used GTEx RNA-sequencing data from various human organ systems to confirm that these genes are predominantly expressed in the nervous system (Supplementary Fig. 1). Table 1Summary of results from clustering and dimensionality reduction analyses.SyndromeSpeciesSpatial Cluster (SC)SC specificitiesTemporal Cluster (TC)TC specificitiesDistribution pattern (t-SNE and PCA)ADMacaque--TC-II↑ Postnatal expressiont-SNE: Distributed with quadrant enrichments; PCA: Broad dispersion with quadrant enrichmentsHumanSC-I↑ Subcortical structures, temporal cortex, frontal, parietal and visual cortices; ↓ Foetal structures (ganglionic eminences, upper rhombic lip)TC-I↑ Late prenatal-adultt-SNE: Distributed, multiple quadrant enrichments; PCA: Broad dispersion with quadrant enrichmentsDDMacaqueSC-I↑ Subcortical structures; ↓ ganglionic eminencesTC-I↑ Across development, postnatal-biasedt-SNE: Confined to one quadrant; PCA: Tight clustering, quadrant enrichmentHuman----t-SNE: Widely distributed; PCA: Broad dispersionSZMacaque----t-SNE: Distributed with quadrant enrichments; PCA: Broad dispersionHumanSC-VI↑ Medial ganglionic eminence, upper rhombic lip, cerebellum, primary motor-sensory cortex; ↑ subcortical structures, and frontal, parietal, visual corticesTC-VI↑ Early-mid prenatalt-SNE: Distributed, multiple quadrant enrichments; PCA: Broad dispersion with quadrant enrichmentsSpatial and temporal clusters (derived from hierarchical clustering) showing FDR-corrected syndrome gene enrichment are indicated. ↑ indicates high expression; ↓ indicates low expression. The distribution patterns derived for each syndrome gene set from t-SNE and PCA have also been described. Summary of results from clustering and dimensionality reduction analyses. Spatial and temporal clusters (derived from hierarchical clustering) showing FDR-corrected syndrome gene enrichment are indicated. ↑ indicates high expression; ↓ indicates low expression. The distribution patterns derived for each syndrome gene set from t-SNE and PCA have also been described. Hierarchical clustering revealed risk gene expression signatures in macaque brains We analyzed the expression of the syndrome-linked genes in rhesus macaque brains using microarray data from NIH Blueprint NHP Atlas [43]. First, we assessed gene expression across 128 macaque brain structures grouped into 13 broad regions. Hierarchical clustering identified four spatial clusters (SCs) (Fig. 2a) with syndrome-specific enrichments (Fig. 2b), of which some survived FDR correction using the Benjamini-Hochberg (BH) method after a hypergeometric test. DD genes showed significant enrichment in SC-I (FDR p = 2.6×10⁻3) (Fig. 2b); 44.1% (68 genes) of the total DD genes for which expression data was available (154 genes) were found in SC-I. To interpret the complex expression patterns seen in Fig. 2a, we calculated the mean expression levels of genes belonging to each spatial cluster across the 13 major areas (Fig. 2c). The cluster showed high expression in ganglionic eminences, and subcortical structures, and less prefrontal cortical involvement compared to the adjacent SC-II. DD-enriched spatial cluster was identified based on differential expression patterns in macaque brain regions. a. The heatmap shows the expression patterns of 1604 disease-associated genes (vertical axis) across 13 broad structures (horizontal axis) in the macaque microarray data (NIH Blueprint Atlas). Relative gene expression of each of the genes across the regions was hierarchically clustered by computing pairwise distances between the data points (log2-transformed probe intensities) using Pearson correlation and identifying closely linked clusters using the average linkage method. The clustered heat map was generated using the Morpheus software. b. The bar graph shows the enrichment of the spatial clusters (Y-axis) for syndrome-related genes, indicated as FDR-corrected statistical significance (shown as -log10(p-value)). c. Relative mean expression values of the spatial clusters across the 13 broad areas shown as z-scores. Second, we examined gene expression across six prenatal and four postnatal stages of macaque brain development. Hierarchical clustering revealed five temporal clusters (TCs) (Fig. 3a). DD genes were strongly enriched in TC-I (FDR p < 2.2 × 10⁻16) (Fig. 3b), a cluster that showed a postnatal bias (Fig. 3c). 43.5% (67 genes) of the total DD genes for which expression data was available (154 genes) were found in TC-I. AD genes were enriched in the postnatally-expressed TC-II (FDR p = 0.017) (Fig. 3b, c); 23.4% (97 genes) of the total AD genes (414 genes) were found in TC-II. We further confirmed that DD genes were strongly biased to TC-I (Supplementary Note 1). Intersections between SCs and TCs aligned with their syndrome associations (Supplementary Fig. 2), e.g., both SC-I and TC-I, detected as DD-enriched clusters, overlapped with each other. DD- and AD-enriched temporal clusters were identified based on differential expression patterns in macaque developmental stages. a. The heatmap shows the expression patterns of 1604 disease-associated genes (vertical axis) across six prenatal and four postnatal stages (horizontal axis) in the macaque brain (Pearson correlation, average linkage method). b. The bar graph shows the enrichment of the temporal clusters (Y-axis) for syndrome-related genes, indicated as FDR-corrected statistical significance (shown as -log10(p-value)). c. Relative mean expression values of the temporal clusters across the ten stages shown as z-scores. Overall, in the macaque brain, strong FDR-corrected enrichment was detected for DD genes in specific spatial and temporal clusters, and FDR-corrected enrichment for AD genes in a specific temporal cluster. See Supplementary Data File 2 for enrichment statistics: odds ratio of enrichment, nominal p, and FDR p. Hierarchical clustering revealed risk gene expression signatures in human brains We analyzed the expression of syndrome-linked genes across 22 human brain structures (Fig. 3a) using RNA-sequencing data from BrainSpan Atlas [44]. Notably, DD genes showed no significant clustering (FDR p > 0.05). Hierarchical clustering identified eight spatial clusters (Fig. 4a). FDR-corrected AD gene enrichment was seen in SC-I (FDR p = 3.96 × 10⁻5) (Fig. 4b). 31.6% (202 genes) of the total AD genes with available data (639 genes) were found in SC-I. The cluster was associated with adult brain structures, namely, subcortical structures, and temporal, frontal, parietal and visual cortices (Fig. 4c). SZ gene enrichment was seen in SC-VI (FDR p = 0.037) (Fig. 4b); 14.8% (184 genes) of the total SZ genes (1244 genes) were found in SC-VI. Compared to SC-I, SC-VI showed lesser involvement of amygdala, hippocampus, and striatum, and higher involvement of thalamus and cerebellum (Fig. 4c). AD- and SZ-enriched spatial clusters were identified based on differential expression patterns in human brain regions. a. The figure shows the expression patterns of 1953 disease-associated genes (vertical axis) across 22 brain structures (horizontal axis) in the human RNA-sequencing data (BrainSpan Atlas). Relative gene expression of each of the genes across the regions was hierarchically clustered by computing pairwise distances between the data points (log2-transformed probe intensities) using Pearson correlation and identifying closely linked clusters using the average linkage method. The clustered heat map was generated using the Morpheus software. b. The bar graph shows the enrichment of the spatial clusters (Y-axis) for syndrome-related genes, indicated as FDR-corrected statistical significance (shown as -log10(p-value)). c. Relative mean expression values of the spatial clusters across the 22 regions shown as z-scores. We also examined gene expression across 31 temporal points grouped into ten developmental stages (Fig. 5a). Hierarchical clustering identified eight temporal clusters. TC-I was strongly enriched for AD genes (FDR p = 1.16 × 10⁻5) (Fig. 5b) and showed high expression from the late prenatal stage to adulthood (Fig. 5c). 32.5% (208 genes) of the total AD genes (639 genes) were found in TC-I. SZ gene enrichment was seen in TC-VI (FDR p = 9.04 × 10⁻3) (Fig. 5b); 14.9% (185 genes) of the total SZ genes (1244 genes) were found in TC-VI. The cluster was highly expressed from early to late-mid prenatal stages, and maintained high expression in later stages (Fig. 5c). SC and TC intersections aligned with their syndrome affiliations (Supplementary Fig. 3), e.g., both SC-I and TC-I, detected as AD-enriched clusters, and SC-VI and TC-VI, detected as SZ-enriched clusters, overlapped with each other. AD- and SZ-enriched temporal clusters were identified based on differential expression patterns in human developmental stages. a. The figure shows the expression patterns of 1953 disease-associated genes (vertical axis) across four prenatal and six postnatal stages (horizontal axis) in the human brain (Pearson correlation, average linkage method). b. The bar graph shows the enrichment of the temporal clusters (Y-axis) for syndrome-related genes, indicated as FDR-corrected statistical significance (shown as -log10(p-value)). c. Relative mean expression values of the temporal clusters across the ten stages shown as z-scores. Overall, in the human brain, while no cluster was detected for DD genes, multiple clusters with FDR-corrected enrichments were noted for AD and SZ genes. See Supplementary Data File 3 for enrichment statistics. The clustering analyses performed were based on Pearson correlation, which is sensitive to outliers and may not capture non-linear relationships. Hence, we repeated the analyses using Spearman correlation and recapitulated the syndrome gene enrichments. In macaques, we identified four spatial (Supplementary Fig. 4a) and five temporal (Supplementary Fig. 5a) clusters that showed significant overlap with the Pearson-derived clusters (Supplementary Fig. 4b and Supplementary Fig. 5b). One spatial cluster showed significant enrichment for DD genes (Supplementary Fig. 4c), whereas no enrichment was observed for AD or SZ genes. In the temporal analysis, two clusters showed enrichment for DD genes (Supplementary Fig. 5c). In humans, Spearman-based clustering identified spatial (Supplementary Fig. 6a) and temporal (Supplementary Fig. 7a) clusters with significant overlap with Pearson clusters (Supplementary Fig. 6b and Supplementary Fig. 7b). Multiple AD- and SZ-enriched spatial (Supplementary Fig. 6c) and temporal clusters were observed (Supplementary Fig. 7c), while no DD-enriched clusters were detected. In the clustering analyses, we examined the spatial and temporal dimensions separately by retaining both dimensions, while using only the spatial or temporal labels to organize the X-axis and performing hierarchical clustering along the gene (Y) axis. This approach was adopted to identify spatially- and temporally- defined gene clusters for downstream analysis and ease of analysis and interpretation. Nevertheless, we computed the mean expression of DD, AD, and SZ gene sets across 26 regions × 10 developmental stages in humans (Supplementary Fig. 8a-c), and 13 broad regions × 10 developmental stages in macaques (Supplementary Fig. 8d-f). In the human dataset, temporal shifts in regional expression patterns were more evident: AD genes showed a clear postnatal bias (Supplementary Fig. 8b), whereas DD showed a widespread prenatal signature (Supplementary Fig. 8a) and SZ showed a prenatal-perinatal signature (Supplementary Fig. 8c). These differences were less apparent in the macaque data (Supplementary Fig. 8d-f), possibly due to flattening of underlying variation, caused by averaging across spatial dimensions (1878 samples categorized into 13 broad regions), a procedure to which microarray data may be more sensitive. RNA-seq data from humans, however, captures greater dynamic range and may therefore retain sensitivity upon averaging. Together, these suggest that separate spatial and temporal analyses may be more suitable for capturing variation in the macaque dataset. t-SNE and PCA captured distinct expression profiles of DD genes in the two species Compared to AD and SZ, hierarchical clustering of DD genes in macaque brains showed a distinct expression profile (Fig. 2). We speculated that DD genes were perhaps widely distributed in the human dataset, as no significant clustering was seen (Fig. 4). We used t-distributed stochastic neighbour embedding (t-SNE) and principal component analysis (PCA) to examine these expression differences. t-SNE is a tuneable dimensionality reduction method and captures less (e.g., 50), moderately (e.g., 30) and more localized (e.g., 10) expression patterns of syndrome-linked genes depending on the perplexity parameter. In macaques, spatial profiles of DD genes were exclusively confined to and showed strong FDR-corrected enrichment (BH method) in a single quadrant at all levels of perplexity (Fig. 6a, d and Supplementary Fig. 9a, d). All enrichments mentioned henceforth in this section are ones that survived FDR correction using the BH method. In contrast, although AD and SZ (Fig. 6d) genes were enriched in specific quadrants, one of which was shared, they were more widely distributed across the plot than DD genes (Fig. 6b, c). At perplexity 10, AD and SZ genes were enriched in the same two quadrants (Supplementary Fig. 9b, c), but at 50, their enrichments shifted to separate quadrants (Supplementary Fig. 9e, f). Similar patterns were seen with temporal profiles (Fig. 6e-h, Supplementary Note 2 and Supplementary Fig. 10), except that at perplexity 50, AD and SZ genes were enriched in the same quadrant. t-SNE revealed the restricted spatial and temporal expression profiles of DD genes and the distributed expression profiles of AD and SZ genes in the macaque brain. The distribution of a. DD genes, b. AD genes, and c. SZ genes, based on their spatial expression profiles (NIH Blueprint Atlas). d. Enrichment of the quadrants (Y-axis) for syndrome-related genes, indicated as FDR-corrected statistical significance (shown as -log10(p-value)). The distribution of e. DD genes, f. AD genes, g. SZ genes, based on their temporal expression profiles. h. Quadrant-wise enrichment of syndrome-related genes. Distance metric = 1 – Pearson Correlation Coefficient, epsilon = 10 and perplexity = 30. The genes have been color-coded based on their syndrome affiliation. Interestingly, DD genes showed a broadly distributed spatial pattern in humans (Fig. 7a), similar to AD and SZ genes at perplexity 30 (Fig. 7b, c). However, AD and SZ genes showed quadrant-specific enrichments that varied with perplexity (Fig. 7d). At perplexity 30, two distinct, non-overlapping quadrants were each enriched for AD and SZ genes (Fig. 7b, c). At perplexities 10 (Supplementary Fig. 11b, c) and 50 (Supplementary Fig. 11e, f), AD consistently showed enrichment in a single quadrant, while SZ enrichment shifted from two quadrants at perplexity 10 to one at 50. Human temporal data (Fig. 7e-h and Supplementary Fig. 12) showed a comparable pattern, except that at perplexity 50, AD enrichment was confined to a single quadrant, and at perplexity 30, DD genes – despite their overall distributed pattern – showed enrichment within one quadrant. t-SNE revealed the distributed spatial and temporal expression profiles of DD, AD and SZ genes in the human brain. The distribution of a. DD genes, b. AD genes, and c. SZ genes, based on the spatial expression profiles of the three gene sets (BrainSpan Atlas). d. Enrichment of the quadrants (Y-axis) for syndrome-related genes, indicated as FDR-corrected statistical significance (shown as -log10(p-value)). The distribution of e. DD genes, f. AD genes, g. SZ genes, based on their temporal expression profiles. h. Quadrant-wise enrichment of syndrome-related genes. Distance metric = 1 – Pearson Correlation Coefficient, epsilon = 10 and perplexity = 30. The genes have been color-coded based on their syndrome affiliation. We also used PCA, which preserves relative distances, to confirm the contrasting patterns of DD genes. PCA replicated the localised pattern of DD genes seen in macaques with t-SNE: they formed a narrow strip close to the PC1 axis that captures 92% variance (Supplementary Fig. 13a). In contrast, although AD (Supplementary Fig. 13b) and SZ (Supplementary Fig. 13c) genes were distributed along PC1, they showed additional spread along PC2 that captured 2.14% variance. Hence, additional factors – even if weakly influential – likely affect these genes. The tight clustering of DD genes magnified along PC2 and PC3 and they were enriched in two quadrants (Supplementary Fig. 13d). AD and SZ genes retained the dispersed pattern along these two axes, with the former enriched in two quadrants (Supplementary Fig. 13e) and the latter (Supplementary Fig. 13e) in none. In human data, DD, AD, and SZ genes were widely distributed along PC1 (Supplementary Fig. 14a-c) and PC2 (Supplementary Fig. 14d-f). DD genes did not show any quadrant-specific enrichment (Supplementary Fig. 14a, d), while AD and SZ genes each showed enrichment in two distinct quadrants along PC1 and in one quadrant each along PC2. Interestingly, PC1 showed lower dominance (81.5%) in human data compared to macaques (Supplementary Fig. 14a-c). Hence, PC2 and PC3 accounted for more variance (7.1% and 3%) perhaps indicating the influence of additional factors on gene expression in humans. We next examined whether the fewer number of genes with available expression profiles in the macaque dataset (156 vs. 252 in human dataset) was responsible for their restricted expression pattern. We mapped the same subset onto the human PC plot (Fig. 8b) and observed that they were widely dispersed, with the pattern magnifying along PC2 (Fig. 8c, d). See Supplementary Fig. 15 and Supplementary Fig. 16 for corresponding patterns of AD and SZ genes. This subset was enriched (all at FDR p < 0.05) for genes involved in central nervous system differentiation, synapse organisation, neuron migration, stem cell differentiation, trans-synaptic signalling, glutamatergic synaptic transmission, and cognition (Fig. 8e, f). Altogether, we concluded that the restricted macaque pattern is unlikely to be an artifact of gene selection or data coverage, and may instead reflect a species-specific profile. Same set of DD genes showed restricted expression in macaques but broad distribution in humans. The distribution patterns of 156 DD genes (with available expression profiles in the macaque dataset) across PC1 and PC2 in a. macaques and b. humans. Distribution of the same genes across PC2 and PC3 in c. macaques and d. humans. Expression profiles were derived from the NIH Blueprint Atlas (macaque) and BrainSpan Atlas (human) (transformation method: ln(x); scaling method: unit variance scaling; PCA method: singular value decomposition). e. Bar graph showing enrichment of the 156 genes for specific GO biological processes (-log10(p-values)). The dotted black line indicates the significance threshold (-log10(p-value) = 1.3) after BH correction for multiple testing. f. GO hierarchy of enriched biological processes among the 156 DD genes; darker blue indicates higher statistical significance. An independent human microarray dataset of higher spatial resolution (232 brain structures from Allen Human Brain Atlas [45] or AHBA) replicated the dispersed pattern of DD, AD and SZ genes in t-SNE (Supplementary Fig. 17). We also performed hierarchical clustering on these structures, grouped into 13 major areas. This analysis identified eight spatial clusters (AHBA SC-I to SC-VIII) (Supplementary Fig. 18a), of which two showed FDR-corrected enrichment for AD genes (AHBA SC-I and AHBA SC-IV) and one for SZ genes (AHBA SC-VIII) (Supplementary Fig. 18b). We then compiled a list of genes showing higher expression in each of the 232 regions relative to others (as described by Rouillard et al. [46]), and checked their distribution across these three spatial clusters using a hypergeometric test (BH-corrected p-value < 0.05 considered significant). SC-I showed enrichment for a few structures in the cerebellum, limbic system, and the temporal lobe (Supplementary Fig. 19). SC-IV showed a subcortical signature, with enrichment in the diencephalon, midbrain, pons and medulla oblongata (Supplementary Fig. 20). SC-VIII showed enrichment in cortical regions (frontal, parietal, temporal, and occipital), as well as regions in the limbic system and the cerebellum (Supplementary Fig. 21). No cluster was enriched for DD genes, replicating the pattern observed in the BrainSpan (BrSp) dataset. We found significant overlaps between the AHBA and BrSp clusters (Supplementary Fig. 18c), consistent with syndrome enrichments. Specifically, AD-enriched BrSp SC-I overlapped with AD-enriched AHBA SC-IV, and SZ-enriched BrSp SC-VI overlapped with SZ-enriched AHBA SC-VIII. Additionally, both BrSp SC-I and SC-VI jointly overlapped with AHBA-SCII (a cluster with no syndrome enrichments). Lastly, to examine the covariation structure of syndrome-related genes, we performed a co-expression analysis in macaques and humans (Supplementary Note 3). In macaques, two co-expression clusters showed FDR-corrected enrichment for DD genes, whereas no such enrichment was observed for AD or SZ. No co-expression cluster showed FDR-corrected enrichment for any of the three syndromes in the human dataset. Overall, multiple methods consistently showed that DD gene expression was spatially and temporally restricted in macaques but diffuse in humans, while AD and SZ genes were broadly distributed across both species. Gene expression patterns from eQTL-supported and genome-wide significant variants The syndrome genes considered in our study were selected due to their genomic proximity to disease-associated genetic variants. It is challenging to identify the specific genes affected by the variants without additional evidence confirming their impact on gene expression. To address this, we examined whether any of the 438 DD, 1178 AD, and 2293 SZ genes showed altered expression patterns associated with their respective variants. We utilized the PsychENCODE [47] and GTEx [48] databases, which provide summary statistics for expression quantitative trait loci (eQTL) associations of specific SNPs and potential eGenes within 1 MB of the SNPs. eQTL computation in PsychENCODE combines prefrontal cortex RNA-seq data from nine databases, namely, GTEx, BrainSpan, CommonMind, CommonMind-HBCC, BrainGVEX, LIBD, BipSeq, UCLA-ASD, and Yale-ASD. Three tiers of evidence were included in our analysis: genes with (i) > 0.1 FPKM in at least 10 samples, (ii) > 0.1 FPKM in at least 150 samples, and (iii) > 1 FPKM in at least 20% samples. We additionally included eQTL data separately derived from GTEx to capture eQTLs from regions other than the frontal cortex (anterior cingulate cortex, amygdala, hippocampus, caudate nucleus, putamen, nucleus accumbens, hypothalamus, substantia nigra, cerebellum, and spinal cord). We applied a stringent threshold of Bonferroni-adjusted p-value < 0.05 to identify statistically significant SNP-gene associations in specific brain structures. We found that the expression of 61 DD (Supplementary Data File 4), 115 AD (Supplementary Data File 5), and 370 SZ (Supplementary Data File 6) genes was significantly modulated by their corresponding SNPs (final gene list in Supplementary Data File 7). Altogether, 13.9% of DD genes, 9.8% of AD genes, and 16.1% of SZ genes in the proximity of disease-linked variants had supporting transcriptomic evidence. In macaques, t-SNE analysis of the expression profiles of eQTL-prioritised genes recapitulated the tight distribution (Supplementary Fig. 22a, e) and quadrant-specific enrichments of DD genes (Supplementary Fig. 22d, h) in macaques. In contrast, DD genes showed a broader distribution pattern in humans (Supplementary Fig. 23a, e), and no quadrant-specific enrichments (Supplementary Fig. 23d, h). Our approach of including genes mapped to variants at both suggestive and genome-wide significance levels introduces disparities in gene set sizes (see Supplementary Methods). To address this, we restricted the sets to genes mapped to variants that passed the genome-wide significance threshold (p < 10⁻⁸), namely, 310 DD genes (70.8% of total DD genes), 412 AD genes (35%), and 687 SZ genes (30%) (Supplementary Data File 7). With t-SNE, in macaques, we observed tighter expression patterns (Supplementary Fig. 24a, e) and quadrant-specific enrichments of DD genes (Supplementary Fig. 24d, h). In humans, DD genes showed broad expression patterns (Supplementary Fig. 25a, e) and no quadrant-specific enrichments (Supplementary Fig. 25d, h). Functional enrichments linked risk gene expression signatures to biological processes To identify the functional implications of the disease-linked clusters shown in Figs. 2–5, we identified the Gene Ontology (GO) biological processes [49] that showed FDR-corrected enrichment (FDR p < 0.05 after BH correction) in them using hypergeometric tests. In the macaque dataset, DD-associated SC-I showed enrichment for a range of synaptic processes (trans-synaptic signalling, synapse organisation, synaptic vesicle cycle), neurogenic processes (regulation of neurogenesis, dendrite development), and neuroinflammatory processes (leukocyte activation, cytokine production) (Supplementary Fig. 26). No DD-specific cluster was identified in the human dataset; hence, GO processes could not be defined. The AD-associated TC-II (Supplementary Fig. 27) showed enrichment for positive regulation of cell activation. In the human dataset, both AD-enriched SC-I (Supplementary Fig. 28) and TC-I (Supplementary Fig. 29) showed enrichment for a broad range of processes, such as synaptic function and neurotransmission, lipid transport and metabolism, amyloid-beta metabolic process, immune responses, and cellular secretion and membrane remodelling. The SZ-enriched SC-VI (Supplementary Fig. 28) and TC-VI clusters (Supplementary Fig. 29) in the human dataset showed enrichment for several processes, such as neurodevelopment, synaptic function and neurotransmission, intracellular transport and vesicle dynamics, and ion homeostasis and transport. The GO hierarchies of biological processes enriched in macaque and human clusters are shown in Supplementary Figs. 30–32. Some differences in the spread of these hierarchies were noted, e.g., focal vs. broad spreads of macaque TC-II (Supplementary Fig. 30) vs. human TC-I (Supplementary Fig. 32), both enriched for AD genes. However, broader GO spreads in the human clusters may reflect increased gene-function annotation and gene coverage in human data, and not biological complexity (e.g., the DD-enriched macaque SC-I has a broad GO hierarchy, Supplementary Fig. 30). In summary, the identified GO processes underlying the three syndromes in the two species provided insights into shared and species-specific cellular functions. Cross-species correlation analysis revealed conserved and divergent expression profiles In the previous sections, we analysed the macaque and human datasets separately using the same set of methods. In this section, we directly compare the expression of the three syndrome-linked gene sets between the two species. Eleven comparable brain regions were identified from the two datasets: (caudal, medial and lateral) ganglionic eminences, dorsolateral prefrontal, orbital frontal, and cingulate cortices, amygdala, hippocampus, thalamus, primary somatosensory cortex, and primary visual cortex (see Supplementary Data File 8 for details). We used Spearman correlation to calculate the average expression of each gene set in these regions in macaques and humans and examine the strength and direction of association (Fig. 9a-c). We closely examined a measure called ‘difference in ranks’ to identify region-specific deviations contributing to the overall correlation pattern (Fig. 9d-f). We hypothesized three scenarios: (1) a positive correlation would indicate “convergence” in gene set expression ranks across regions within an anatomically/functionally connected neural circuit, suggesting a conserved expression pattern; (2) a negative correlation would indicate “divergence” reflecting a systematic shift in expression across a circuit; (3) a weak correlation would suggest no systematic conservation or shift across regions (region-specific changes between species will still be captured). Expression patterns of DD and SZ genes were negatively correlated in human and macaque brains. Correlation of the expression of a. DD, b. AD and c. SZ genes in eleven comparable macaque and human brain regions. Spearman’s rank correlation was used to examine the relationship between the expression values; R indicates the strength and direction of the association, whereas p indicates the statistical significance of the association. Region-wise difference in ranks between macaques and humans have been shown for d. DD, e. AD and f. SZ genes; the differences have been indicated as data labels next to the bar plots. DD genes showed a moderately strong, statistically significant negative correlation (R = −0.65, p = 0.037) (Fig. 9a) with large negative deviations, i.e., higher-ranked expression in humans than macaques, in orbitofrontal, primary somatosensory, and dorsolateral prefrontal cortices, and caudal ganglionic eminence (Fig. 9d). Hippocampus and amygdala showed negative deviations of smaller magnitude. The correlation was performed using only 11 data points each from the macaque and human datasets, and should therefore be interpreted with caution. Nevertheless, Fig. 9d shows negative rank differences in eight of the eleven regions, indicative of negative rank correlations in these areas, suggesting that the observed trend is not driven by a small number of data points. SZ genes also showed a moderate negative correlation (R = −0.59) between species, though this did not reach statistical significance (p = 0.061) (Fig. 9c). The strongest negative deviations were observed in the medial ganglionic eminence, and orbitofrontal, primary somatosensory and dorsolateral prefrontal cortices, with smaller deviations in the caudal ganglionic eminence and hippocampus (Fig. 9f). On the other hand, AD genes showed a weak, statistically non-significant negative correlation (R = −0.12, p = 0.73) between species (Fig. 9b). Lastly, we compared the expression of genes involved in glucose metabolism (GO: 0006006 from the GO database [49]) – a process likely not undergoing a systematic shift (for example, across a specific circuitry) in the two species – and found that they exhibit a weak, statistically non-significant correlation (R = −0.14, p = 0.69) (Supplementary Fig. 33a). Overall, we found clear divergence in the rank-based regional profiles of DD and SZ genes between humans and macaques, suggesting possible changes in gene expression programs. On the other hand, AD genes and glucose metabolism-related genes showed weak negative correlations, suggesting a lack of systematic spatial shifts. Discussion In this study, we examined the spatial and temporal expression profiles of genes associated with three evolutionarily relevant syndromes in macaque and human brains. Three key findings emerged. (I) DD risk genes showed restricted postnatal subcortical expression in macaques but broad cortical and subcortical expression in humans. The divergent subset was enriched for neuronal and behavioural processes. Cross-species spatial comparisons revealed a significant negative correlation for DD genes, suggesting an expansion of DD-related circuitry from conserved affective hubs in macaques to distributed emotion-cognition networks in humans. “Expansion” in the context of our findings refers to broad spatial expression, not an increase in functional scope, integration, or complexity of neural circuits. (II) SZ-linked genes were expressed across cortical, subcortical, and cerebellar structures in humans, showing a negative, though non-significant, correlation across the two species. (III) AD-linked genes showed postnatal cortical-hippocampal expression in macaques and cortical-subcortical expression in humans, with weak and non-significant cross-species correlations, suggesting an absence of systematic expression shifts. These findings are discussed in the following sections. Depression In macaques, DD-risk genes clustered in SC-I and TC-I, implicating subcortical areas and postnatal stages. Subcortical structures such as the ventral tegmental area influence reward processing and decision-making, key components of depression-like behaviours, in NHPs [50, 51]; and the ventral striatum shows abnormal activation in major DD patients [52]. Notably, no specific DD cluster was detected in the human dataset. Instead, hierarchical clustering, t-SNE and PCA showed that DD genes were distributed across various clusters with diverse expression profiles. The broader spatial distribution in humans suggests that DD genes have a more globally distributed expression across networks, a finding supported by the moderately strong and statistically significant negative correlation between human and macaque DD gene expression profiles, particularly in emotion- and cognition-related regions, such as the orbitofrontal cortex, dorsolateral prefrontal cortex, hippocampus and the amygdala. Thus, regions with lower expression in macaques systematically show higher expression in humans, indicating an evolutionary shift in specific neural circuits. It cannot be determined whether this divergence reflects compensatory changes without performing genome-wide disease-agnostic comparisons. Additionally, our current analyses are limited to detecting broad versus narrow spatial expression distributions, and cannot explain functional expansion, redistribution, or regulatory reweighting of neural circuits. Encouragingly, the subset of DD genes with restricted expression in macaques showed a dispersed pattern in humans. The enrichment of this subset for neuron differentiation, migration, synaptic signalling, and importantly, cognition, indicates a possible shift in gene regulation with evolutionary divergence, affecting both cellular processes and behavioural domains, consistent with previous studies [53–55]. Hence, the complexity of cognitive control in humans is likely reflected in the gross and microstructural changes in these regions, e.g., the human prefrontal cortex [1–4]. These factors remain unexamined in our study. In addition, several other factors likely contribute to the observed effects: species-specific differences in the hemispheric symmetry and architecture of columns, unique cell types, species-specific cell-type markers, as well as differences in laminar composition, cell-type composition within specific regions, and the developmental timing of cortical maturation [3, 4]. In macaques, pessimistic valuation, as seen in depressive states, have been correlated with a loss of top-down cognitive control over limbic regions [56]. Our findings suggest that depression-linked risk genes may have expanded their expression footprint in the human brain, reflecting complex cognitive regulation of limbic structures, potentially an adaptation to social and physical environments. Importantly, affective modulation and depression, often resolve spontaneously, or with environmental [57–59] and social manipulation [60–65], implying a flexible neural substrate capable of homeostatic recovery. Together, these results suggest that DD genes in humans have a more distributed role, in line with multi-level changes in the human brain, spanning genes, proteins, lipids, neurons, and connectivity patterns [66]. While dataset differences may contribute, the enrichment of cognition-related genes among those with expanded human expression supports this interpretation. The macaque DD-linked cluster was enriched for synaptic, neurogenic, and neuroinflammatory processes, likely reflecting conservation of core affective pathways that are harder to detect in humans due to complexity arising from recent evolutionary divergence. Divergence in DD genes expression may also arise from species-specific changes in neural progenitor populations, perhaps the caudal ganglionic eminence, which showed a large deviation between species in our analysis. This eminence, distinct from the medial and lateral counterparts [67], populates layer 5 cortical neurons, and various limbic structures. In humans, differences in this cell population may have contributed to the complexity of cortico-limbic regions and their involvement in DD. Lastly, our findings support a model where depression arises in humans from dysregulation of a distributed network regulating affective, perceptual, and conative processes [68]. We detect DD-linked regions from statistically (i) non-significant DD gene enrichments in human spatial clusters, and (ii) significant human-specific deviations in cross-species correlations. A network model of depression involving cortical (ventrolateral, orbital and dorsolateral prefrontal cortices) and subcortical areas (striatum, amygdala and dorsal thalamus) thus emerges, consistent with their hypothesized integrative role in depression-like behaviours. For example, human fMRI studies show decreased prefrontal activation coupled with increased subcortical activation during emotional processing tasks [69]. We do not infer causal relationships among gene expression changes, neural circuits, signalling pathways, and syndromes in this study. Genes, environment, and behaviour may interact in a stochastic and dynamic fashion, constrained by possible variations across these axes. Hence, inferring causality as a linear chain of events may be difficult. Nevertheless, two scenarios can be imagined. One, genetic risk factors and gene expression changes may affect the formation of neural circuits that mediate depression, and modulate their responsiveness to depression-inducing events. Two, depressive behaviour and adverse events may alter gene expression (perhaps through epigenetic mechanisms), and modulate neural circuits and susceptibility to syndromes. Schizophrenia In macaques, SZ-related genes did not achieve FDR significance. In humans, SC-VI show FDR significance. SC-VI showed enrichment in the medial ganglionic eminence and upper rhombic lip. Enrichment in the ganglionic eminences supports the involvement of SZ risk genes in early brain development, consistent with the neurodevelopmental model of SZ [70]. We noted enrichment of SZ genes in the cerebellum and subcortical structures, and multiple cortical areas, including frontal, motor-sensory, parietal, and visual cortices, most of which have been implicated in network-level dysfunctions associated with SZ [71]. In humans, the SZ-enriched TC-VI showed prenatal expression and sustained expression beyond the prenatal period, indicating a tendency for early neurodevelopmental disruptions in SZ. Cross-species spatial comparisons showed a moderately negative correlation, though not statistically significant, suggesting possible divergence in the transcriptomic organization of SZ genes. Human-specific shifts were most pronounced in the medial ganglionic eminence, orbital and dorsolateral prefrontal cortex, and primary somatosensory cortex, with additional, subtler differences in the caudal ganglionic eminence and hippocampus. These regions are functionally linked to cognition, sensorimotor processing, and emotional regulation, all relevant to SZ symptomatology. The human-specific shift in medial ganglionic eminence is particularly notable, as this region gives rise to cortical interneurons and populates cortical layers [72]. This suggests that human-specific changes in SZ gene expression may be nodal to this progenitor population, with the potential to impact cortical development. In line with this, in humans, SZ-linked clusters were enriched for neurodevelopmental, synaptic, regulatory, and cellular processes. These enrichments were not observed in macaque SZ-linked clusters, pointing to a human-specific molecular architecture underpinning SZ risk [73]. The correlation analysis considers only divergences in spatial expression across the two species; the temporal dimension, including differences in lifespan, gestational length, and neurodevelopmental pacing, may also contribute to the observed negative correlation. Alzheimer’s disease In the macaque dataset, AD genes were enriched in the postnatally-expressed TC-II, consistent with the late onset of AD [74]. In the human dataset, AD genes showed high expression in subcortical structures and a wider set of cortical areas – as seen with SC-I – from the late prenatal period to adulthood (TC-I), supporting their role in mature neural circuits. However, the developmental stages sampled in macaques and humans are not directly comparable (Supplementary Methods). The macaque dataset spans from peak neurogenesis (E40) to post-pubertal adulthood (48 months), broadly capturing the transition from juvenile to adult [75], whereas the human dataset ranges from early prenatal stages (8 pcw) to adolescence (13–19 years) and adulthood (21–40 years). Accordingly, we carefully refer to clusters as “prenatal” and “postnatal” relative to the perinatal boundary. Interpretations should therefore be made with respect to this boundary rather than direct comparisons of specific developmental time points. AD clusters in humans were involved in a broad range of synaptic, metabolic, and immune processes, suggesting a complex and multifactorial signature, possibly reflecting species-specific higher vulnerability to AD [76]. Interestingly, AD genes showed a weak, statistically non-significant negative correlation in spatial expression profiles between macaques and humans, similar to genes involved in glucose metabolism, a process that is not likely to have undergone a systematic shift in the two species. For genes involved in core metabolic pathways or neurodegeneration/dementia, these results may reflect neutral changes without major shifts in functional architecture. In contrast, the stronger divergence seen in DD and SZ risk-related genes may reflect human-specific expansion of gene expression in neural circuits involved in cognition and emotion. Limitations of the study Our study has some limitations. First, the technologies used for generating the macaque (microarray) and human (RNA-seq) transcriptomic datasets and their coverage differ. However, several factors indicate that the observed divergence is unlikely to result merely from platform differences: (i) the expression distribution patterns of the syndrome gene sets in the human brain were replicated in the Allen Brain Atlas RNA-seq dataset, (ii) mapping the same subset of 156 DD genes onto human PCs produced broader distributions than in macaques – ruling out coverage as a confound – and this subset showed enrichment for cognition-related processes, adding biological validity, (iii) both datasets were processed using comparable normalization and variance scaling, and (iv) only DD genes showed a systematic shift between species. All these indicate that methodological differences are unlikely to be the primary driver. Additionally, in all analyses, the macaque and human datasets were examined separately and overall patterns were compared, except for Spearman correlations, where expression values were converted to ranks and compared across species. The datasets also differ in spatial resolution: the macaque data include layer-specific dissection and expression information, whereas the human data lack this level of resolution. However, we refrain from analyzing layer-specific patterns and instead base our analyses on 13 broad regions in macaques and 26 in humans. The rank correlation analysis is performed on a subset of eleven comparable regions. Nevertheless, future studies using datasets with directly comparable regions, higher gene coverage and spatial/temporal resolution will help extend these findings. Second, a fuller picture of the evolutionary trajectory of syndrome gene expression would require more transcriptomic data across different vertebrate and invertebrate species. Systematic comparisons across species such as mouse, marmoset, macaque, and human will be an important direction for future work. Nevertheless, within the constraints of the available datasets, we present evidence from the human and macaque datasets using multiple methods with complementary strengths. Third, the findings reported here are limited to the three syndrome gene sets analysed. A more comprehensive analysis would include an expanded set of traits with evolutionary relevance. Fourth, our primary strategy for gene prioritization, proximity-based mapping of genes around a SNP, may have introduced spurious disease-gene associations. To test the robustness of our key findings from transcriptomic patterns, we repeated our analyses with a smaller subset of credible genes supported by eQTL evidence. Fifth, the gene sets used for the three syndromes differ in size and are influenced by variability across the GWASs in which they were identified, warranting cautious interpretation of our findings. However, by using the full set of genes located within 50 kb of SNPs crossing the nominal significance threshold in the GWAS catalog, we ensure that the results remain credible within the scope of the genetic data that is currently available. We also repeat our key analyses with a smaller subset of genes mapping to variants passing genome-wide significance. Lastly, the Global Burden of Disease (GBD) report by the Institute for Health Metrics and Evaluation (IHME) shows significant sex differences in disease prevalence and age of onset across DD, AD and SZ. Based on data from 1990–2021, DD and AD are more prevalent in females than males [77, 78]. Males have a higher lifetime risk of SZ, with onset likely occurring between 18–25 years in men and 25–35 years in women [79]. These sex differences in disease manifestation may be reflected in the expression patterns identified in our study. Studies have reported sex-biased gene expression across several macaque and human brain regions, hormone-mediated regulation of sex-biased genes, sex differences in neuroimmune function, and the implications of these processes for human disease [80, 81]. The datasets used here are, in principle, suitable for sex-stratified analyses (human: 19 females and 23 males; macaque: 2 females and 2 males). However, this line of enquiry is beyond the current scope of the study, and will be an important direction for future work. In summary, our findings reveal convergent and divergent spatiotemporal patterns underlying three evolutionarily relevant syndromes across macaque and human brains. They suggest that evolutionary shifts in gene expression, such as in DD circuitry, helped shape human emotion and cognition, while contributing to disease susceptibility. Methods Compilation of syndrome-associated risk genes Genes harbouring variants associated with the three syndromes were compiled from the GWAS catalog [39], namely, 438, 1178, and 2293 genes associated with DD, AD and SZ (Supplementary Data File 1), respectively. The genes were found within 50 kb upstream or downstream of the syndrome-associated SNPs. To capture the full spectrum of the known disease genetic architecture, we included variants at ‘suggestive’ levels of significance (p-value < 1E-05), and those reaching genome-wide significance (p-value < 1E-08); see Supplementary Methods for a discussion on the implications of this approach. The rhesus macaque orthologs (taxon ID: 9544) of the human genes were retrieved from the Homologene database (https://www.ncbi.nlm.nih.gov/datasets/gene/). Hierarchical clustering, t-SNE and PCA of expression profiles We analyzed microarray data from macaque brains available in the NIH Blueprint Atlas [43] and RNA-sequencing data from human brains in BrainSpan Atlas [44] to identify gene clusters with distinct spatial and temporal expression profiles (see Supplementary Methods for full data description). The macaque dataset included 1878 samples (categorised for ease of analysis into 13 broad regions) across ten developmental stages, and the human dataset included 524 samples (from 22 broad regions) across 31 temporal points. Spatial and temporal analyses were performed by structuring the data to emphasize one axis while retaining resolution along the other. For spatial analysis, gene expression values were organized across brain regions (top X-axis in Figs. 2a and 3a), and gene clustering along the Y-axis was performed with respect to the regional axis. Developmental stage-specific information was retained without averaging, but was not used as a basis for clustering. For temporal analysis, expression values were organized across developmental stages (top X-axis in Figs. 2c and 3c), while retaining region-specific information. Gene clustering was performed on the Y-axis with respect to the development axis to identify time-dependent patterns. Average probe intensity or RPKM values were log2-transformed, and hierarchical clustering was performed using Morpheus software (Supplementary Methods); RPKM is Reads Per Kilobase Million [82]. Pairwise distances were calculated using Pearson correlation, and clusters were identified using the average linkage method. The enrichment of the syndrome-linked genes for those expressed in specific tissues across various organ systems was computed using RNA-sequencing data from GTEx (Supplementary Methods) [83]. After hierarchical clustering of the Blueprint and BrainSpan spatial expression profiles of the syndrome genes, we asked whether a substantial number of genes within a particular spatial cluster exhibited a significant bias toward high expression in any of the structures in the dataset. We compiled a list of genes highly expressed in any of the structures relative to others, as described by Rouillard et al. [46] in the form of a gene matrix transposed (GMT) file. This GMT file was uploaded as a custom database to WebGestalt [84] for performing an over-representation analysis with the syndrome genes. Statistical significance of the overlap between the list of genes belonging to a spatial cluster and the genes highly expressed in a specific structure was computed using the hypergeometric distribution (Supplementary Methods). The log2-transformed probe intensity (or RPKM) values were assembled into a data matrix containing syndrome-associated genes (rows) and samples (columns), and used as an input for PCA performed with the web-based tool Clustvis (Supplementary Methods) [85]. The data matrix was pre-processed to include rows and columns with less than 70% missing values. The log2probe intensity (or RPKM) values in the matrix were centered using the unit variance scaling method. Singular value decomposition (SVD) with imputation was used to extract PCs. t-SNE plots were constructed using the Morpheus software [82] based on the spatial and temporal expression profiles of syndrome-associated genes across samples, using the following parameters: distance metric = 1 – Pearson Correlation Coefficient, epsilon = 10, perplexity = 30 (varied to 10 and 50 to detect local and global patterns), and maximum number of iterations = 1000. The perplexity parameter defines the effective number of nearest neighbors of each data point, and highlights the local and global aspects of the data structure. For independent replication of the human spatial results, we used microarray data comprising 3702 samples from Allen Brain Atlas [45]. Normalized microarray data was available for 29,130 genes from the brains of six healthy donors, each dissected into 363–946 samples – yielding a total of 3702 samples – from 13 areas in the brain. Functional enrichment analyses The enrichment of the clusters for the genes associated with specific biological processes (Gene Ontology [49]) was computed using WebGestalt [84]. WebGestalt first calculates the distribution of genes belonging to a particular functional category in the input list. It then compares it with the background distribution of genes in that category among all the genes annotated with any functional category in the selected database. Statistical significance of enrichment was assessed using the hypergeometric test and corrected for multiple hypotheses using the BH method. Annotations with a BH-corrected p-value < 0.05 were considered significant. Supplementary information Supplementary Information Supplementary Data Files Acknowledgements The authors thank Dr Ole Andreassen (University of Oslo) for helpful suggestions on the work. Author contributions SJ conceived the project. KBK collected the datasets, designed the methodology and performed the analyses. SJ, KA, BV, NB, and MP supervised the analyses. KBK and SJ wrote the manuscript. All authors edited and approved the manuscript. Funding This work was supported by the Rohini Nilekani Centre for Brain and Mind (G-202303-00523 to KBK, MP, and BV), SKAN Research Trust (SKAN/002/208/2021/01481 to KBK, MP, BV and SJ), DBT/Wellcome Intermediate Fellowship (IA/CPHI/20/1/505266 to BV), MEXT/JSPS KAKENHI Grants (JP25K02377, JP22H04998, JP24H02163 to KA), and AMED Grants (JP24gm6910012, JP24wm0625210 to KA). SJ acknowledges the Department of Biotechnology (Government of India) grant entitled “Accelerator program for discovery in brain disorders using stem cells” (ADBS) for funding (BT/PR17316/MED/31/326/2015). Data availability The DD, AD and SZ risk-related genes analyzed in this study have been made available as Supplementary Data File 1 and Supplementary Data File 7. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Kalyani B. Karunakaran, Email: kalyanithepebble@gmail.com Biju Viswanath, Email: bijuv1@gmail.com. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04198-w. References Associated Data Supplementary Materials Supplementary Information Supplementary Data Files Data Availability Statement The DD, AD and SZ risk-related genes analyzed in this study have been made available as Supplementary Data File 1 and Supplementary Data File 7.