Characterizing and decoding ultraconserved regions uncovers their regulatory significance in human brain development and disorders Authors: Yun Bai, Xuehan Zhang, Xiangting Wang, Meiting Wang, Tao Shen Published: 2025 ABSTRACT AbstractUltraconserved regions (UCRs) represent genomic elements with perfect sequence conservation across humans, rats, and mice. Though evolutionarily significant, most UCRs’ functions remain elusive. Here, by leveraging current reference genomes from 34 species spanning diverse taxonomic groups, we refine the annotation, trace the evolutionary origins, and compile an updated catalog of UCRs. The updated UCRs are mostly 200-400 base pairs long, with lower GC content and SNP density but more deleterious mutations. They first emerge in fish, growing more abundant in vertebrates. Guided by genomic context and functional analyses, we classified them into three categories: type I (within protein-coding genes) mainly encodes brain development splicing regulators; type II (within non-coding genes) associates with lncRNAs linked to brain malignancies; type III (intergenic, 46% overlapping enhancer-like elements) regulates brain development genes. Our findings underscore UCRs’ critical roles in human brain development and disorders, offering a framework for investigating their biological significance.Subject terms: Genome informatics, Evolution Introduction Ultraconserved regions (UCRs), initially identified by Bejerano et al. from the human genome, comprise 481 genomic segments ranging from 200 to 779 nucleotides in length1. These remarkable elements demonstrate perfect sequence conservation across human, rat, and mouse genomes, with extended conservation observed in dog, chicken, and fish genomes, suggesting profound evolutionary and functional significance1. Notably, a subset of UCRs exhibits transcriptional activity, giving rise to transcribed ultraconserved regions (T-UCRs)2. Emerging evidence implicates T-UCRs in diverse biological processes (BP), including colon carcinogenesis, neural progenitor cell proliferation, and cardiovascular development and function2–5. Previous classification efforts have categorized UCRs into 225 exonic and 256 non-exonic elements based on transcriptional evidence. Functional annotation analysis revealed that exonic UCR-associated genes are predominantly involved in RNA-binding and splicing regulation, while non-exonic UCR-adjacent genes are enriched in transcriptional regulatory and DNA-binding pathways1. Despite these insights, the functional characterization of UCR-related genes remains incomplete, and the biological rationale underlying the extreme conservation of most UCRs continues to elude scientific understanding. Another limitation in current UCR research stems from its reliance on the original set of 481 UCRs identified by Bejerano et al. using early genome assemblies (human: July 2003, UCSC hg16; mouse: February 2003, UCSC mm3; rat: June 2003, UCSC rn3)1,6. For instance, while recent studies have identified broader ultraconserved elements (UCEs) using updated assemblies (e.g., human: GCF_000001405.38), they still reference this foundational 481-UCR set for comparison6. The substantial advancements in sequencing technologies and bioinformatics tools have led to the development of more accurate reference genomes (human genome: GRCh38.p14; mouse genome: GRCm39; rat genome: mRatBN7.2)7. These updated genome assemblies not only provide refined sequence information but also necessitate a requirement for systematic reassessment of UCR-associated genomic features and their functional annotations. This reanalysis is particularly crucial as it may reveal previously unrecognized regulatory elements or modify our understanding of existing UCR-gene relationships. In the present study, we conducted a comprehensive study integrating evolutionary analysis, functional annotation, and genomic characterization of UCRs. Leveraging the current genome assemblies, we identified and validated 479 UCRs that maintain perfect conservation across species. The updated UCRs exhibit significantly lower mutation rates compared to flanking and random genomic regions; however, with a higher proportion of pathogenic variants, underscoring their functional importance. Phylogenetic assessment demonstrated that UCRs predominantly emerged during vertebrate evolution, suggesting their crucial roles in vertebrate-specific BP. To systematically investigate their functional significance, we developed a classification system based on genomic context: type I UCRs located within protein-coding genes, with a broad category encoding splicing regulators that modulate alternative splicing (AS) of brain development-related genes; type II UCRs are closely associated with long non-coding RNAs (lncRNAs), showing brain-specific expression patterns and involvement in tumorigenic processes; and type III UCRs overlapped with intergenic regions, with 46% overlapping enhancer-like elements and regulating genes enriched in neural development pathways. These findings not only updated the annotation and classification of UCRs but also established a foundation for exploring UCRs as potential therapeutic targets in neurodevelopmental disorders and brain-related diseases. Results Characteristics and evolutionary origins of the updated UCRs Through re-annotation of UCRs using the current reference genomes of human, rat, and mouse, we established a refined catalog of 479 conserved elements, providing an updated resource for functional genomic studies (Supplementary Data 1–3). According to the definition of UCR by Bejerano et al., two previously identified UCRs, uc.18 and uc.304, failed to meet the stringent criteria for 100% ultraconservation across human, rat, and mouse genomes and were thus excluded from subsequent analyses, as illustrated in Supplementary Fig. 1. Subsequently, we delved into the sequence characteristics of the updated UCRs, paying particular attention to their length distribution, GC content, and the density of single-nucleotide polymorphisms (SNPs). These UCRs primarily exhibited lengths within the range of 200–400 base pairs (bp), with five UCRs (uc.94, uc.190, uc.323, uc.352, and uc.382) precisely measured 200 bp in length (Fig. 1a). Notably, three longest UCRs—uc.462 (779 bp), uc.464 (770 bp), and uc.467 (731 bp)—are all located within the intron of the POLA1 gene on the X chromosome, hinting at their potential roles in DNA replication1. Next, we investigated the GC content of UCRs, a crucial factor influencing DNA stability and nucleosome positioning. Our analysis revealed a significantly lower GC content compared to their flanking regions and randomly selected genome fragments (Fig. 1b, Supplementary Data 4). This observation aligns with established evidence that reduced GC content minimizes mutation risks, particularly through diminished CpG dinucleotide deamination, a major source of spontaneous mutations8. By avoiding GC-rich motifs prone to methyl-cytosine oxidation and subsequent TpG transitions, UCRs may avoid accumulating mutations, thereby preserving their ultraconserved nature. Additionally, GC-poor regions often associate with open chromatin configurations, which may broadly facilitate regulatory element accessibility9,10, underscoring the critical role of UCRs in orchestrating genomic functions, including transcriptional regulation and chromatin organization. Sequence characterization and evolutionary origins of UCRs. a Histogram showing the length distribution of UCRs, with the majority ranging between 200 and 400 bp. b Raincloud plot comparing GC content among UCRs, UCR-flanking regions, and random genomic segments. UCR left: equal-length fragments located on the left side of the UCRs. UCR right: equal-length fragments located on the right side of the UCRs. Randf: random genomic fragments. c Violin plots comparing SNP density (left panel) and SNP locus density (right panel) among UCRs, UCR-flanking regions, and random genomic segments. d The cumulative distribution map demonstrates that SNPs within UCRs are enriched for low-frequency variants, with a significantly higher proportion of SNPs exhibiting AF < 0.1% compared to control fragments. e CADD score analysis revealed a significantly higher proportion of deleterious mutations in UCRs compared to control fragments. The flanking regions of UCRs also show a higher proportion of deleterious mutations than random genomic fragments (The line of UCR left is overlapped with UCR right). f Line plot showing the number of genomic regions with 80%, 90%, and 95% sequence identity to UCRs across 34 representative species. Statistical analysis was performed using a Wilcoxon rank-sum test for (b–e). To examine the implications for variation, we analyzed the density, frequency, and deleteriousness of SNPs within UCRs relative to control fragments. SNP density was lower in UCRs than in controls (Fig. 1c), consistent with reduced mutation rates under evolutionary constraint. We further analyzed allele frequencies using gnomAD v4.1 (a large-scale database with uniform genomic coverage). The results showed that 97.5% of SNPs in UCRs were low-frequency (allele frequency, AF; AF < 0.1% defined as low-frequency), compared with 92.8% in controls (Chi-square test, p < 2.2 × 10−16), and cumulative distribution analysis confirmed consistently lower frequencies of UCR-associated SNPs across all AF ranges (Fig. 1d). We then assessed deleteriousness using Combined Annotation Dependent Depletion (CADD) scores11, which revealed a notably higher proportion of deleterious mutations in UCRs compared to controls (Fig. 1e). Together, these results indicate that despite their lower SNP density, UCRs harbor a disproportionately high number of deleterious mutations. To elucidate the evolutionary origins of UCRs, we conducted a comprehensive analysis across a diverse range of species. Our study encompassed 34 representative species, spanning from invertebrates to mammals, including humans, rats, and mice. We quantified the number of sequence fragments in each species that demonstrated 80%, 90%, and 95% sequence identity with human UCR fragments. Our analysis revealed a consistent pattern: UCRs first emerge in fish and become significantly more abundant in vertebrates, implicating UCRs in the emergence and evolution of vertebrate-specific structures (Fig. 1f). This pattern also underscores the potential importance of UCRs in the evolutionary trajectory that distinguishes vertebrates from non-vertebrates. Functional significance and classification of the updated UCRs in humans To systematically investigate the functional roles of the updated UCRs in humans, we first quantified the tissue-specific index (τ) of UCR-associated genes using GTEx multi-tissue expression data (https://www.gtexportal.org/)12. Compared with all genes detected in the GTEx database, UCR-associated genes exhibited significantly lower τ values, suggesting their broad expression patterns across diverse human tissues (Fig. 2a). Notably, the broad expression patterns of UCR-associated genes were paralleled by enhanced protein–protein interaction capacities of their encoded proteins. This dual observation—widespread tissue expression coupled with relatively heightened connectivity in molecular networks—suggests that UCR-associated protein-coding genes may function as central regulatory nodes in diverse cellular systems (Fig. 2b)13. UCR-associated protein-coding genes exhibit high interactability and extensive global regulation. a Distribution of tissue-specific index (τ) of all human genes and UCR-overlapped genes annotated in the GTEx dataset. The value of τ ranges from 0 to 1. The closer a gene’s τ value is to 1, the higher its tissue specificity. b Beeswarm plot display the distribution of the number of protein–protein interactions for protein-coding genes overlapped with UCRs and random genomic fragments (Randf). c Enrichment of genes overlapping UCRs or random genomic fragments (Randf) within essential (EG) versus non-essential (NEG) gene sets. d Functional enrichment analysis reveals significant involvement of UCR-associated essential genes in the assembly of biomolecular condensates, particularly stress granules and processing bodies (P-bodies). e Pie chart illustrating the number and proportion of three groups of UCRs. Building on this observation of potential functional centrality, we next examined the essentiality of UCR-overlapping genes. We analyzed their distribution within the essentialome, defined as the set of essential genes (EGs) indispensable for cellular survival (e.g., genes whose deficiency compromises viability or fitness)13. By using non-essential genes (NEGs) as the control set, our results showed genes overlapping UCRs were significantly enriched within the essentialome compared to genes overlapping random genomic fragments (Fig. 2c). Functionally, these UCR-associated EGs showed pronounced involvement in biomolecular condensate assembly, particularly in stress granules and processing bodies (P-bodies), membraneless organelles critical for spatiotemporal gene expression regulation (Fig. 2d). Despite the substantial intersection between UCR-overlapped genes and EG sets, as well as the well-documented regulatory functions of these overlapped UCRs, the biological roles of the majority of UCRs remain largely enigmatic. To systematically investigate the functional landscape of UCRs, we categorized the updated set of 479 UCRs into three distinct groups based on their genomic context: type I UCRs (65%), which are located within protein-coding genes; type II UCRs (21%), which co-localize with non-coding genes; and type III UCRs (14%), which reside in non-coding regions lacking annotated transcripts but display strong evolutionary conservation patterns indicative of regulatory potential (Fig. 2e, Supplementary Data 1–3). This tripartite classification not only highlights the diverse genomic contexts of UCRs but also reveals their multimodal regulatory architecture, offering a structured framework for future investigations into their functional roles across different biological contexts. Type I UCRs within protein-coding genes modulate brain development by regulating alternative splicing processes Type I UCRs are located within the genomic regions of protein-coding genes, exhibiting two distinct distribution patterns: exonic and intronic regions. In detail, we identified 126 type I UCRs overlapping with exonic regions (designated as exonic UCRs) and 188 type I UCRs situated within intronic regions (termed intronic UCRs) (Fig. 3a, b, Supplementary Data 5). Functional characterization through gene ontology (GO) enrichment analysis revealed that both exonic and intronic UCR-associated protein-coding genes are significantly involved in critical BP related to brain development and alternative splicing (AS)14. To quantitatively assess potential functional divergence, we compared the enrichment of exonic versus intronic UCR-associated genes in these key processes. Genes associated with exonic UCRs were significantly more enriched in AS pathway (45/118 genes, 38%) compared to those linked with intronic UCRs (8/99 genes, 8%; p = 1.12 × 10−7) (Supplementary Fig. 2). Conversely, intronic UCR-associated genes showed greater enrichment in brain development pathway (33/99 genes, 33%) relative to exonic UCR-linked genes (21/118 genes, 18%; p = 6.59 × 10−3) (Supplementary Fig. 2). This quantitative analysis demonstrate a functional divergence between these two UCR categories: genes associated with exonic UCRs show predominant involvement in AS regulation, while those linked to intronic UCRs play a more substantial role in nervous system development (Fig. 3a, b). Expanding functional profiling to Molecular Function (MF) and Cellular Component (CC) categories revealed that both exonic and intronic UCRs are enriched in RNA/DNA binding terms (Supplementary Fig. 3). Exonic UCRs showed strong MF enrichment in mRNA binding, transcription factor binding, and DNA-binding transcription repressor activity, with CC terms highlighting spliceosomal complexes and cytoplasmic stress granules. Intronic UCRs exhibited similar MF enrichment but distinct CC localization in transcription regulator complexes and the nuclear periphery. These patterns suggest exonic UCRs function in RNA processing and transcriptional regulation within dynamic subcellular compartments, while intronic UCRs appear involved in spatially organized transcriptional control. Such differential localization and binding profiles further support the functional diversity of type I UCRs. Type I UCRs regulate alternative splicing events associated with brain development. a, b Gene ontology enrichment analyses of exonic UCRs (a) and intronic UCRs (b) related genes. c Chromosome map shows the overlap of type I UCRs and 46 splicing-related genes. Splicing factors from the SR and hnRNP families were annotated in red. d Heatmap shows the expression landscape of type I UCRs-overlapped splicing factors during brain development. wpc weeks post-conception, infants: 6–9 months old, toddlers: 2–4 years old, school: 7–9 years old, teenagers: 13–19 years old. e, f The barplots show the number of devAS of different types (e) and patterns (f). devAS developmentally dynamic AS, d down, u up, du down-up, ud up-down. The inclusion frequency of an alternative splicing event goes through a progressive increase during development (termed “up”), progressive decrease of the inclusion frequency (“down”), increase followed by a decrease (“up-down”) and decrease followed by an increase (“down-up”). CE cassette exon, RI retained intron, AA alternative acceptor, AD alternative donor. g Scatter plots comparing the hexamer frequencies of exons in down-regulated cassette exons (dCE) and up-regulated cassette exons (uCE) revealed an enrichment of the SRSF1 binding motif (GGA) in brain development-skipped exons. h The locations of uc.418 and uc.419 on SRSF1. To further elucidate the molecular mechanisms underlying these observations, we performed genome-wide mapping of protein-coding genes overlapping with type I UCRs and splicing-related genes annotated in the GO database as listed in Supplementary Data 6. Our results showed a significant convergence between type I UCRs and these splicing-related genes (Fig. 3c). This intersection included two pivotal splicing factor families: multiple members of the serine/arginine-rich (SR) protein family and heterogeneous nuclear ribonucleoproteins (hnRNPs) (Fig. 3c)15. Through temporal expression analysis of these UCR-associated splicing factors across human brain development—spanning from embryonic stages (4 weeks post-conception) to adulthood (63 years old)16—we observed a progressive decline in the expression of most splicing-related genes as brain maturation proceeds (Fig. 3d). To further resolve whether this temporal expression pattern is recapitulated at the cellular level and across specific neurodevelopmental trajectories, we further characterized the expression dynamics of these UCR-associated splicing-related genes across cell types and developmental stages using single-cell RNA sequencing (scRNA-seq) data spanning human brain development from the first trimester to adolescence17. As the brain matures, new cell types continuously emerge through differentiation. The developmental trajectory of glutamatergic excitatory neurons (ENs) originates from three radial glial cells, progresses through IPC-EN (IPC: intermediate progenitor cell), EN-newborn, and EN-IT-immature (IT: intratelencephalic), and ultimately differentiates into mature EN subtypes, including EN-L2_3-IT, EN-L4-IT, EN-L5-IT, EN-L6-IT, EN-L5_6-NP, EN-L5-ET, EN-L6-CT, and EN-L6b (Supplementary Fig. 4a, b). Tracking the expression of AS genes along the developmental progression of ENs revealed a consistent downward trend, consistent with aforementioned results (Supplementary Fig. 4c). Notably, this cell type-specific, stage-resolved expression pattern aligns with the temporal course of neuronal differentiation, supporting the conclusion that the decline in UCR-associated splicing factors is not merely a global phenomenon but is spatiotemporally coordinated with neurodevelopmental processes. The functional implications of this expression pattern were further investigated through a comprehensive analysis of AS events during brain development. Among various splicing events, down-regulated cassette exon (dCE) events emerged as the predominant type (Fig. 3e, f). Particularly noteworthy is the role of SRSF1, a key splicing factor and representative SR protein family member, which regulates cassette exon events through specific recognition of the GGA motif18. Our analysis demonstrated that the progressive decline in SRSF1 expression during brain development correlates with significant enrichment of its target GGA motif in exons affected by dCE events (Fig. 3g). The presence of specific UCRs (uc.418 and uc.419) within the SRSF1 genomic locus raises the intriguing, yet speculative, possibility that these conserved elements might be involved in modulating SRSF1 activity or expression to influence AS during neurodevelopment (Fig. 3h). Another key splicing factor overlapping with type I UCR is SRSF6, which provides further insight into the molecular mechanisms linking UCRs to splicing regulation. Previous studies have established that many UCRs embedded within AS genes contain premature termination codons (PTCs)19; when transcribed, these UCRs generate PTC-containing transcripts that are targeted for degradation via nonsense-mediated mRNA decay (NMD), thereby post-transcriptionally regulating protein output through modulation of splice isoform abundance. Our analysis of SRSF6 transcripts revealed that exons overlapping with type I UCR are exclusively present in a single isoform, which we identify as a “poison exon” due to its PTC-induced NMD sensitivity. To assess the regulatory relevance of this UCR-associated exon, we quantified the usage frequency of both the poison exon and two “common exons” (present in all SRSF6 isoforms) during brain development (Supplementary Fig. 4d). Compared to common exons, the usage of the poison exon exhibited marked fluctuations, particularly around birth. This dynamic shift in isoform composition indicates that, despite relatively constant declined total SRSF6 mRNA levels, the proportion of transcripts encoding functional protein is modulated by UCR-associated poison exon inclusion. Given that NMD-degraded transcripts fail to produce functional SRSF6 protein, these changes in poison exon usage might impact SRSF6 protein abundance—a critical regulator of splicing cascades in developing neurons. Furthermore, since SRSF6 itself regulates the splicing of other AS-related genes, this creates a feedback loop where UCR-mediated poison exon inclusion could fine-tune splicing regulatory networks during brain development. Collectively, the observed correlations—including the co-localization of UCRs with AS-related genes, the developmental decline in the expression of splicing-related genes, the association among SRSF1 downregulation, GGA motif enrichment in brain dCE events, and nearby UCRs, and the altered usage frequency of poison exon that overlaps UCR in the SRSF6 gene—provide correlative evidence supporting the hypothesis that type I UCRs may function as regulatory elements modulating AS programs during brain development. Type II UCRs in lncRNA genes serve as diagnostic biomarkers and prognostic indicators for glioma For type II UCRs, we initiated our investigation by systematically analyzing the categories of non-coding genes that overlap with these regions. Remarkably, we discovered that 99% of these overlapping transcripts are lncRNAs, as detailed in Supplementary Data 2. We designated these overlapping lncRNAs as ncUCR genes and proceeded to examine their expression profiles using the GTEx database. This analysis revealed that ncUCR genes exhibit a striking tissue-specific expression pattern, with predominant activity in brain tissues (Fig. 4a). To further explore their biological relevance, we conducted a functional enrichment analysis using Lnc2Path, which identified significant enrichment of ncUCR genes in pathways associated with the nervous system and tumor biology (Supplementary Data 7)20. Type II UCRs are associated with neurological tumors. a The heatmap shows that ncUCR genes are predominantly expressed in brain tissues. b, c The volcano plots show the differentially expressed ncUCR genes in LGG (b) and GBM (c) retrieved from TCGA-LGG, TCGA-GBM, and GTEx-cerebrum. d Forest plot of ncUCR genes associated with LGG and GBM patient survival. e The ROC curves of the logistic regression models constructed based on differentially expressed ncUCR genes for LGG (left panel) and GBM (right panel). Building on these insights, we delved into the potential connection between ncUCR genes and nervous system tumors, with a particular emphasis on glioblastoma (GBM) and lower-grade gliomas (LGG). To further validate the functional significance of type II UCRs, particularly their potential roles in tumor biology, we identified differentially expressed ncUCR genes in GBM and LGG by integrating data from TCGA and GTEx. Specifically, TCGA-GBM and GTEx-cerebrum data were collected for GBM analysis, while TCGA-LGG and GTEx-cerebrum data were collected for LGG analysis. Our analysis revealed 21 and 17 ncUCR genes that were differentially expressed in GBM and LGG, respectively (Fig. 4b, c), underscoring their potential involvement in tumorigenesis. To assess the clinical relevance of these ncUCR genes, univariate Cox regression analysis was performed, identifying a subset significantly associated with patient survival (Fig. 4d), which supports their potential as prognostic candidates. For diagnostic potential, given the limited number of features (ncUCR genes) and sample size characteristics, we adopted logistic regression for its superior interpretability and suitability for low-dimensional data21. The workflow included: (i) collecting differentially expressed ncUCR genes from TCGA-GBM, TCGA-LGG, and GTEx-cerebrum; (ii) applying LASSO regression with 10-fold cross-validation to identify key features, reducing overfitting risks; (iii) building the logistic regression model using these features; and (iv) validating it using the integrated data of the GEO cohort (GSE147352) and 20 normal tissues data from the CGGA database (Supplementary Fig. 5 and 6). In the independent validation set, the model achieved an area under curve (AUC) of 0.993 for LGG (trained on TCGA-LGG vs. GTEx-cerebrum) and 0.972 for GBM (trained on TCGA-GBM vs. GTEx-cerebrum), indicating strong discriminative ability (Fig. 4e). However, these results should be interpreted cautiously: limitations include potential batch effects in public datasets, and generalizability to clinical practice requires validation in larger, prospectively collected cohorts with standardized protocols. Thus, while ncUCR genes show promise as complementary diagnostic biomarkers (particularly in multi-gene panels), their clinical translation remains preliminary and necessitates further investigation. Type III UCRs display enhancer-like features and are associated with brain development Type III UCRs are situated in intergenic regions of the human genome and do not overlap with any coding or non-coding transcripts. To explore their potential regulatory functions, we conducted an overlap analysis between type III UCRs and candidate cis-Regulatory Elements (cCREs), which were identified by the ENCODE project through the integration of diverse omics data related to gene expression regulation22. A total of 926,535 and 339,815 cCREs were identified in humans and mice, respectively. Our analysis revealed that among 65 type III UCRs, 32 overlapped with cCREs, including 30 that overlapped with enhancer-like regions (Fig. 5a). To assess the statistical significance of type III UCR-cCRE overlap, we performed a genomic enrichment analysis based on Monte Carlo simulations. We generated a background distribution by randomly selecting 65 intergenic genomic fragments (matched to the length distribution of type III UCRs) across 10,000 iterations. The observed number of type III UCR-cCRE overlaps (38) was higher than the simulated mean (7.04 ± 2.51; one-tailed p < 2 × 10−16; 95% confidence interval: 3–12), indicating a 5.4-fold enrichment (Supplementary Fig. 7). This non-random association supports the hypothesis that type III UCRs are enriched within functional regulatory elements. We further calculated the distances between type III UCRs and the transcription start sites (TSSs) of their nearest protein-coding genes in both the human and mouse genomes (Supplementary Data 3, 8). The distances were predominantly within the range of 10,000–100,000 bp, consistent with the long-range regulatory features of enhancers (Fig. 5b). Furthermore, GO functional enrichment analysis of the protein-coding genes adjacent to type III UCRs revealed significant enrichment in nervous system development pathways in both humans and mice. The remarkable consistency of these pathways across species strongly supports the functional involvement of type III UCRs in brain development (Fig. 5c). Type III UCRs display enhancer-like features and regulate gene expression in brain development. a The left panel shows that 32 type III UCRs overlap with cCREs, while the right panel presents the type and number of cCREs overlapping with type III UCRs. dELS: distal enhancer-like signatures; pELS: proximal enhancer-like signatures. b Distance between type III UCRs and the transcription start site of their nearest protein-coding genes. The area of the curves represents the density of the distribution. The bars at the bottom of the density distribution represent the data points. Red represents the human and blue represents the mouse. c GO enrichment analyses of human and mouse type III UCRs nearest protein-coding genes. Red represents the human and blue represents the mouse. d Metagene plots and heatmaps showing H3K4me1 signal enrichment across different genomic regions in brain organoids. Sample sizes: n = 479 regions for Randf and UCRs; n = 314, 100, and 65 regions for type I, II, and III UCRs, respectively. e Metagene plot and heatmap showing H3K27ac signals in type III UCRs and their neighboring regions in brain tissues. To further investigate the role of type III UCRs in brain development, we analyzed the distribution of enhancer marks H3K4me1 and H3K27ac in type III UCRs and their adjacent regions in brain tissues. CUT&Tag data demonstrated that in brain organoids, the H3K4me1 signal was dramatically stronger in type III UCRs and their adjacent regions compared to Randf, all UCRs, type I UCRs, and type II UCRs (Fig. 5d). We then assessed the enrichment of H3K27ac in type III UCRs and their adjacent regions across various human organs. The results indicated that in the brain, the H3K27ac signal was most pronounced in the type III UCR regions, with a gradual decrease in the adjacent regions (Fig. 5e). In contrast, in other human tissues, the H3K27ac signal did not show a similar pattern. Notably, while stomach tissues exhibited higher absolute H3K27ac signal intensity, brain tissues displayed a distinct enrichment pattern with signal peaking specifically at type III UCR regions (Supplementary Fig. 8). This regional specificity in brain tissues, absent in other organs, supports the functional relevance of type III UCRs in neural contexts. Of note, 33 of the 65 type III UCRs do not overlap with annotated cCREs, prompting us to explore their potential regulatory signatures. Analysis of histone modifications revealed that these UCRs, despite lacking cCRE annotations, retain tissue-specific epigenetic features: H3K27ac signals are enriched within these UCRs relative to flanking regions in brain tissues (but not other tissues like stomach), while H3K4me3 (a promoter mark) shows localized upstream enrichment specifically in the brain (Supplementary Fig. 9). These patterns suggest they may function as cryptic neural-specific regulatory elements or promoter-proximal modulators, aligning with the enrichment of adjacent genes in nervous system development pathways. Collectively, our data indicate that type III UCRs contribute to brain development through both cCRE-associated and context-specific non-cCRE mechanisms. To further validate the potential of type III UCRs as enhancers, we focused on three protein-coding genes implicated in the brain development pathway, as illustrated in Fig. 5c. Specifically, we identified their nearest type III UCRs: Otp-uc.157, Pou3f3-uc.67, and Pou3f4-uc.476. To assess their regulatory potential, we conducted a comprehensive analysis of histone modification markers (H3K4me1, H3K4me2, and H3K4me3), chromatin accessibility (ATAC-seq), and CTCF binding data across multiple developmental stages in mice. Our findings revealed that uc.67, uc.157, and uc.476 exhibit dynamic patterns of histone modifications and chromatin accessibility, consistent with the characteristics of active enhancers (Supplementary Fig. 10). Notably, these UCRs showed enrichment of H3K4me1 and H3K4me2 markers typically associated with enhancer activity, as well as high chromatin accessibility during critical periods of brain development (Supplementary Fig. 10). Furthermore, the presence of CTCF binding sites near these UCRs suggests potential roles in chromatin looping and long-range regulatory interactions (Supplementary Fig. 10). Additionally, ecHi-C data revealed that the regions where uc.67, uc.157, and uc.476 are located can interact with the promoter regions of Pou3f3, Otp, and Pou3f4, respectively (Supplementary Fig. 11). The Otp gene is known to regulate the fate, migration, and terminal differentiation of hypothalamic neurons23–25, while Pou3f3 is a central regulator of gene expression programs in neocortical progenitors critical to determine brain size during evolution26. Notably, Pou3f4 plays a crucial role in embryonic development, particularly in the formation of the neural tube as well as the paraventricular and supraoptic nuclei of the hypothalamus27. Furthermore, Pou3f4 upregulates the expression of Gli1, promoting the differentiation and development of neural stem cells28. Collectively, these results strongly support that uc.67, uc.157, and uc.476 possess enhancer-like properties and may play pivotal roles in modulating the expression of their associated genes during brain development. This evidence not only reinforces the enhancer feature of type III UCRs but also highlights their potential contribution to the intricate regulatory networks governing neurodevelopment. Discussion UCRs are among the most intriguing features of the human genome, exhibiting perfect sequence conservation across human, rat, and mouse. Despite their evolutionary conservation, the functional roles of many UCRs remain enigmatic. In this study, we confirmed the ultraconservation of UCRs in the current versions of the human, mouse, and rat genomes, with the exception of uc.18 and uc.304, which no longer meet the criteria for UCR. Phylogenetic analysis revealed that UCRs emerged with chordates during evolution and reached their peak in mammals, suggesting their critical role in the development of vertebrate-specific traits1. Furthermore, the significantly lower mutation rate of UCRs compared to control regions underscores the essential importance of sequence fidelity in these conserved elements29. This evolutionary conservation is consistent with the functional enrichment of UCR-associated genes in pathways related to brain development, highlighting their role in preserving fundamental neurodevelopmental programs across species30. To systematically explore their diverse regulatory functions, we classified UCRs into three distinct types: type I (located within protein-coding genes), type II (associated with non-coding RNAs), and type III (situated in intergenic regions). This tripartite classification provides a comprehensive framework for understanding the biological significance of UCRs and their contributions to gene regulation and disease. Type I UCRs, which constitute the majority of the updated catalog, are predominantly located within protein-coding genes and are enriched in pathways related to mRNA metabolism and brain development. Our findings highlight their critical role in regulating AS, a process essential for generating protein diversity during neurodevelopment. Notably, the progressive downregulation of SRSF1, a key splicing factor overlapped with uc.418 and uc.419, during brain development was associated with the enrichment of GGA sequences in exons affected by dCE events. This suggests that type I UCRs may fine-tune the splicing of genes critical for brain development, a finding supported by recent studies demonstrating the importance of AS in neurogenesis31,32. Furthermore, the convergence of broad tissue expression, enhanced protein–protein interaction network connectivity, and essentiality strongly suggests that genes associated with UCRs, particularly those protein-coding ones, function as core components of fundamental cellular operating systems. These conserved loci encode regulators of deeply ancient and indispensable processes foundational to cellular life, such as core transcriptional, translational, and stress-responsive pathways (including biomolecular condensate assembly)33. Importantly, as exemplified by their critical role in fine-tuning neurodevelopmental splicing revealed in this study, these core regulatory machineries are not merely static “housekeeping” elements but are dynamically repurposed and intricately regulated to execute complex developmental programs. Moreover, the dysregulation of such evolutionarily constrained core processes is frequently implicated in pathogenesis, including cancer34,35, highlighting their critical importance beyond development. These insights position UCRs, especially those overlapped with protein-coding genes, as central regulators at the nexus of cellular survival, fundamental BP, developmental plasticity, and potentially, disease mechanisms. Type II UCRs, which account for 21% of the catalog, are predominantly associated with lncRNAs. Our analysis revealed that these lncRNAs are highly expressed in brain tissues and are significantly enriched in pathways related to neurodevelopment and tumor biology. Notably, several ncUCR genes, including DLX6-AS1 (ENSG00000231764) and LINC00609 (ENSG00000257585), showed associations with patient survival in our LGG cohort. DLX6-AS1 exhibited a marginal association with increased risk, while LINC00609 was associated with a protective effect. Previous studies have suggested roles for DLX6-AS1 and LINC00609 within broader regulatory networks contributing to tumor progression in various cancers, including glioma36–38, aligning with the growing evidence implicating lncRNAs in brain malignancy pathogenesis. However, the precise functional roles of these specific ncUCR genes in glioma require further experimental validation. The identification of ncUCR genes associated with glioma prognosis highlights their potential value as diagnostic or prognostic biomarkers. Future studies should explore the molecular mechanisms underlying these associations and evaluate the therapeutic potential of targeting these lncRNAs. Type III UCRs, representing 14% of the catalog, are located in intergenic regions and exhibit features of enhancer-like elements. Our analysis revealed that 50% of type III UCRs overlap with cCREs, including enhancer-like regions identified by the ENCODE project. The enrichment of enhancer marks (H3K4me1 and H3K27ac) in type III UCRs in brain tissues further supports their role as active regulatory elements. Additionally, the interaction between type III UCRs and promoter regions of brain development-related genes, such as Otp, Pou3f3, and Pou3f4, underscores their regulatory potential. These findings are consistent with recent studies demonstrating the role of enhancers in orchestrating gene expression during neurodevelopment39,40. The conservation of type III UCRs across species further highlights their evolutionary importance in brain development. These insights suggest that type III UCRs act as conserved functional elements that regulate the expression of genes critical for brain development. The evolutionary conservation and functional significance of UCRs in brain development invite comparison with another class of genomic elements: Human Accelerated Regions (HARs). HARs are genomic sequences that, despite being highly conserved across non-human mammals, have undergone rapid evolutionary acceleration in the human lineage. These regions are thought to play a pivotal role in the development of human-specific traits, particularly in the nervous system41. Interestingly, approximately 50% of HARs are implicated in regulatory functions during human brain development42, mirroring the functional enrichment of UCRs in neurodevelopmental pathways. While UCRs are characterized by their extreme conservation across species, HARs represent regions where accelerated evolution in humans may have driven the emergence of uniquely human cognitive abilities. Despite these differences, both UCRs and HARs converge in their critical roles in nervous system development, suggesting a complementary relationship: UCRs may preserve fundamental neurodevelopmental programs shared across species, while HARs may provide the evolutionary substrate for human-specific neural innovations43. This interplay between conservation and acceleration highlights the dual evolutionary forces shaping the genomic landscape of brain development and function. However, while our study provides an updated catalog of UCRs defined by the human-mouse-rat triad—ensuring comparability with foundational work—it is important to acknowledge the evolutionary limitations of this framework. The original definition of UCRs, established when genomic resources were more limited, relies on three species with relatively narrow evolutionary divergence: mouse and rat are closely related rodents, and the evolutionary distance from human to these rodents, while meaningful, does not capture the full breadth of vertebrate diversity. Broader vertebrate comparisons offer clues to this context: UCRs first emerge in fish and expand in vertebrates, suggesting their co-option into vertebrate-specific traits and implying that their conservation likely extends beyond the human-mouse-rat triad. Notably, this pattern of early emergence in fish is not merely an artifact of our mammalian-centered definition. Cross-species analyses have identified UCR-like elements in non-mammalian vertebrates (e.g., zebrafish, frog) and non-rodent mammals (e.g., dog), confirming that these sequences predate the mammalian radiation44,45. Their functional relevance across these lineages is further supported by the conservation of associated pathways: for example, orthologs of UCR-linked splicing regulators (e.g., SR family genes) in drosophila and zebrafish play critical roles in neurodevelopment, mirroring their functions in mammals46–48. The functional conservation across distantly related species suggests UCRs (and their analogs) contribute to shared neurodevelopmental programs, not exclusively mammalian ones. This broader context also highlights that the strict 100% identity threshold in a small number of species may exclude sequences with functional significance conserved across deeper evolutionary time. Looking forward, expanding the UCR catalog will require leveraging the wealth of high-quality genomes now available, spanning invertebrates to non-rodent mammals. Such an approach would enable the identification of elements conserved across greater evolutionary distances, potentially uncovering ancient regulatory mechanisms predating mammalian divergence. Refining conservation thresholds—balancing stringency with inclusivity—will also be critical: while 100% identity ensures extreme constraint, moderate relaxations (e.g., 80–95%) may capture functionally important sequences that tolerate minor variations without losing biological activity. Nevertheless, the present study provides a comprehensive update on the characteristics, evolutionary origins, and functional significance of UCRs. The tripartite classification of UCRs offers a robust framework for future investigations into their diverse roles in gene regulation and disease. Our findings highlight the critical contributions of UCRs to brain development and disorders, paving the way for novel therapeutic strategies and diagnostic tools. By elucidating the molecular mechanisms underlying UCR function, this study advances our understanding of the genomic basis of brain development and provides a foundation for future research in genomics and medicine. Methods Acquisition of UCR genomic locations UCR genomic locations were initially downloaded from the hg16 genome (available at https://users.soe.ucsc.edu/~jill/ultra.html). To update these coordinates to the GRCh38.p14 genome, the NCBI Remap tool was utilized. In the Remap interface, the Assembly-Assembly function was selected, with Homo sapiens as the source organism, GRCh34 as the source assembly, and GRCh38.p14 as the target assembly. The UCR site information file from hg16 was uploaded, and default settings were retained to generate the corresponding UCR locations on GRCh38.p14. Validation of conservation of UCRs Based on the updated UCR locations in GRCh38.p14, the nucleotide sequences of 481 UCRs were extracted using bedtools (v.2.31.1) getfasta (bedtools getfasta -fi reference_genome.fa -bed UCR_locations.bed -name -fo UCR_sequneces.fasta)49. These sequences were then aligned to the reference genomes of rat (mRatBN7.2) and mouse (GRCm39) using Blast (Basic Local Alignment Search Tool, v.2.14.1) (blastn -query - -db UCR.blastdb -outfmt ‘6 qseqid sseqid pident length mismatch gapopen qstart qend sstart send qseq sseq evalue bitscore’ » “$OUTPUT_DIR/$OUTPUT_FILE”)50. A UCR was considered ultraconserved if at least one alignment site exhibited 100% sequence identity. Selection of random genomic fragments Random genomic fragments were selected using the sample function in RStudio. First, a chromosome was randomly chosen from chromosomes 1–22, X, and Y. Next, a fragment length was randomly selected within the UCR length range (200–779 bp). Finally, a starting position was randomly determined within the selected chromosome’s length range. A fragment’s position was defined by the chromosome, starting position, and fragment length. Calculation of GC content The getfasta function from bedtools (v.2.31.1) was used to extract the nucleotide sequences of UCR flanks and random genomic fragments. UCR flanks of equal length to UCRs were used as controls. In RStudio, the readDNAStringSet function from the Biostrings (v.2.66.0) package was employed to read the FASTA sequences, and the letterFrequency function was used to count the occurrences of A, G, C, and T. GC content was calculated as the ratio of (G + C) to the total nucleotide count (A + G + C + T)51. The gghalves package (v.0.1.4) was used to plot a raincloud plot to show the distribution of GC context52. The ggsci package (v.3.2.0) was used to select the color palette53. Acquisition of SNP, SNP frequency and CADD score SNP data were retrieved in batches using the biomaRt (v.2.54.1). Parameters were set as follows: Mart to ENSEMBL_MART_SNP, dataset to hsapiens_snp, and the coordinates of UCRs, UCR flanks, and random genomic fragments were input separately. SNPs were filtered based on supporting evidence, retaining only those from TOPMed, gnomAD, ExAC, 1000 Genomes, HapMap, ESP, Cited, Frequency, and Phenotype_or_Disease. CADD is a tool for scoring the deleteriousness of single-nucleotide variants, multi-nucleotide substitutions, as well as insertion/deletion variants in the human genome11. SNP frequency data from geomAD were retrieved using the Variant Effect Predictor (VEP) web tool provided by Ensembl, with the gnomAD (genomes) allele frequency option enabled54. CADD scores were likewise obtained via the Ensembl VEP web tool, with the CADD option enabled and default settings retained. Evolutionary analysis A total of 34 species representing key evolutionary milestones were selected, and their reference genomes were downloaded using the ncbi-datasets-cli tool (v.15.20.0). The reference genome versions for each species used in this study were listed in Supplementary Data 9. UCR sequences were aligned to these genomes using blastn, and the number of sequences with 80%, 90%, and 95% similarity to UCRs was calculated. Results were visualized as a line graph. Calculation of tissue-specific index Tissue-specific index (τ) for UCR-overlapping genes was calculated using expression data from the GTEx database. The τ ranges from 0 (low specificity) to 1 (high specificity), with values closer to 1 indicating higher tissue specificity. The τ was calculated using the following formula: The analysis of UCR-overlapping protein-coding genes To assess whether protein-coding genes overlapped by UCRs contribute to key BP through extensive protein–protein interactions, we utilized a comprehensive dataset of experimentally validated molecular interactions reported in the ref. 13. We then quantified the number of proteins interacting with protein-coding genes overlapped by UCRs and random genomic fragments. Additionally, the essentialome, the phase-separated condensates assembled by protein-coding genes, and the BP participated by phase-separated condensates were retrieved from the ref. 13. The ggupset package (v.0.4.1) was used to generate an UpSet diagram depicting phase separation particles assembled by protein-coding genes and the BP in which these particles participate55. Gene ontology enrichment analysis GO enrichment analysis was performed using the Metascape tool (https://metascape.org/) with a false discovery rate (FDR) threshold of <0.0514. Expression data of genes associated with alternative splicing during brain development Data on the expression patterns of genes associated with AS during brain development were obtained from the EvoDevoApp database (http://evodevoapp.kaessmannlab.org)16. ScRNA-seq data spanning the period from the first trimester to adolescence were obtained from Human Developing Brain scMultiome Atlas (https://cell.ucsf.edu/snMultiome/)17. The analysis of developmentally dynamic alternative splicing events Data on developmentally dynamic alternative splicing (devAS) events were obtained from Evo-devo: Alternative splicing database (https://apps.kaessmannlab.org/alternative-splicing)56. The number of devAS in different tissues and organs, as well as the proportion of devAS types in the brain, were quantified. To calculate the frequency of hexamers in brain dCE, a BED file containing dCE event-related exons was obtained, and exon sequences were retrieved using the getfasta function of bedtools (v.2.31.1), then saved as FASTA files. In RStudio, the readDNAStringSet function from the Biostrings package (v.2.66.0) was used to read the FASTA sequences, and the oligonucleotideFrequency function was employed to calculate the occurrence frequency of each hexamer in the sequences. The parameters were set as follows: width = 6, step = 6. Finally, the occurrences of each hexamer across all variable exon sequences were summed to obtain the overall hexamer frequencies. Analysis of exon usage frequency of SRSF6 Exon usage frequency data during brain development were retrieved from the BrainSpan database (based on GENCODE v10). Using the GENCODE v10 genome annotation file, we identified common exons (present in all transcripts) and poison exons (present only in a subset of transcripts and overlapping with UCRs). The usage frequencies of the target exons were then extracted from the frequency dataset. Differential expression analysis of type II UCRs-related lncRNAs RNA-seq data from TCGA-GBM, TCGA-LGG, and GTEx-cerebrum were obtained from the TCGA-TARGET-GTEx project in the UCSC Xena database. Normal tissue data from GTEx, used as controls, included only samples from the Brain-Cortex and Brain-Frontal Cortex (BA9). Differential expression analysis was conducted using the DESeq2 package (v.1.38.3). Genes with an adjusted p < 0.05 and |log2FoldChange| ≥ 1.0 were defined as differentially expressed genes. Furthermore, differentially expressed lncRNAs overlapping with type II UCRs were identified. Data visualization was performed using ggplot2 (v.3.5.0), heatmap (v.1.0.13), and ggpubr (v.0.6.1) packages. Univariate Cox regression analysis was conducted using the ezcox package (v.1.0.4). Construction and validation of a logistic regression model Expression data for TCGA-GBM, TCGA-LGG, and GTEx-cerebrum samples were obtained from the TCGA-TARGET-GTEx project hosted by the UCSC Xena platform and used as the training set. Differentially expressed lncRNAs overlapping with type II UCRs between tumor and normal cerebrum samples were identified using the DESeq2 package (v.1.38.3), with significance thresholds set at a Benjamini–Hochberg adjusted p < 0.05 and |log2FoldChange| ≥ 1. To minimize model overfitting, least absolute shrinkage and selection operator (LASSO) regression was performed with 10-fold cross-validation using the glmnet package (v.4.1-10), and the optimal penalty parameter (lambda) was chosen based on the minimum mean cross-validated error. The selected features were then used to construct a multivariable logistic regression model to classify tumor versus normal samples in the training set. Model validation was performed using an independent dataset integrating the GEO cohort GSE147352 with 20 normal brain tissue samples from the CGGA database. Discriminative ability was evaluated using the area under the receiver operating characteristic curve (ROC) by the pROC package (v.1.19.0.1). Overlap between type III UCRs and human candidate cis-Regulatory Elements Genomic locations of human cCREs were downloaded from the SCREEN database (http://screen.encodeproject.org/). Overlaps between cCREs and type III UCRs were identified using the intersect function in bedtools (v.2.31.1). Identification of protein-coding genes adjacent to type III UCRs Human gene annotations were obtained from the ENSEMBL database in GTF format. A bed file listing protein-coding gene names and their TSSs was generated using homemade code. The closest function in bedtools (v.2.31.1) was used to identify the nearest upstream and downstream genes to type III UCRs, then further determine the nearest protein-coding genes. ChIP-seq analysis Dataset GSE183627 contains CUT&Tag experimental data for H3K4me1 in brain organoids. We downloaded the BAM files and generated metagene plots and heatmaps using DeepTools (v.3.5.5)57. Additionally, ChIP-seq data for H3K27ac across various human tissues and organs were obtained from the European Genome-Phenome Archive (EGA) under accession number EGAD00001004335. Metagene plots and heatmaps for these data were also generated using DeepTools (v.3.5.5). Statistics and reproducibility All statistical analyses were conducted using R (v.4.2.1). The analysis of variance (ANOVA) was applied to assess differences among multiple groups, whereas the Wilcoxon rank-sum test was used to compare two groups. Specifically, the p for Figs. 1 and 2a, b were calculated using the Wilcoxon rank-sum test, and the p for Fig. 2c was calculated using the chi-squared test. Statistical significance was defined at a threshold of p < 0.05. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1–9 Reporting Summary Acknowledgements This work was supported by the National Natural Science Foundation of China (32300450), the 2023 Wuhu Science and Technology Plan Project (2023jc14), and the Outstanding Innovative Research Team for Molecular Enzymology and Detection in Anhui Provincial Universities (2022AH010012). The authors also thank the Nebula computing platform of the University of Science and Technology of China (USTC). Author contributions T.S., M.T.W., and X.T.W. conceived the project and designed the research. Y.B. performed the majority of the analysis with contributions from T.S., X.H.Z., M.T.W., and X.T.W. T.S. and Y.B. drafted the manuscript with input from all authors, whereas M.T.W. and X.T.W. revised the manuscript. Peer review Peer review information Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editors: Cesar Fortes-Lima & Rosie Bunton-Stasyshyn. A peer review file is available. Data availability The data on the expression patterns of alternative splicing genes during brain development were obtained from the EvoDevoApp database (http://evodevoapp.kaessmannlab.org)16. The scRNA-seq data related to brain development from the first trimester to adolescence were obtained from the Human Developing Brain scMultiome Atlas (https://cell.ucsf.edu/snMultiome/)17. Data on developmentally dynamic alternative splicing (devAS) events were obtained from Evo-devo: Alternative splicing database (https://apps.kaessmannlab.org/alternative-splicing)56. RNA-seq data from TCGA-LGG, TCGA-GBM, and GTEx used for analyses of differentially expressed ncUCR genes, survival-associated ncUCR genes, and logistic regression model training were obtained from the TCGA-TARGET-GTEx project in the UCSC Xena database (https://xenabrowser.net/). RNA-seq data used for LR model validation included 18 LGG samples, 85 GBM samples, and 15 normal brain tissues obtained from GSE147352 in the GEO database (https://www.ncbi.nlm.nih.gov/geo), and 20 normal brain tissues obtained from the CGGA database (https://www.cgga.org.cn/). Data on exon usage frequency were obtained from the BrainSpan database (https://www.brainspan.org/). CUT&Tag data used in this study were obtained from GSE183627 in the GEO database (https://www.ncbi.nlm.nih.gov/geo), and ChIP-seq data were obtained from EGAD00001004335 in the EGA database (https://ega-archive.org/). All other data are available from the corresponding author on reasonable request. Code availability The code used to generate the results in this study is publicly available at the GitHub repository (https://github.com/byebye1034/1_UCR_project), and the same version of the code has been deposited in Zenodo (10.5281/zenodo.17373904)58. 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 Meiting Wang, Email: stucomputer0@163.com. Tao Shen, Email: stao@ahnu.edu.cn. Supplementary information The online version contains supplementary material available at 10.1038/s42003-025-09115-3. References Associated Data Supplementary Materials Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1–9 Reporting Summary Data Availability Statement The data on the expression patterns of alternative splicing genes during brain development were obtained from the EvoDevoApp database (http://evodevoapp.kaessmannlab.org)16. The scRNA-seq data related to brain development from the first trimester to adolescence were obtained from the Human Developing Brain scMultiome Atlas (https://cell.ucsf.edu/snMultiome/)17. Data on developmentally dynamic alternative splicing (devAS) events were obtained from Evo-devo: Alternative splicing database (https://apps.kaessmannlab.org/alternative-splicing)56. RNA-seq data from TCGA-LGG, TCGA-GBM, and GTEx used for analyses of differentially expressed ncUCR genes, survival-associated ncUCR genes, and logistic regression model training were obtained from the TCGA-TARGET-GTEx project in the UCSC Xena database (https://xenabrowser.net/). RNA-seq data used for LR model validation included 18 LGG samples, 85 GBM samples, and 15 normal brain tissues obtained from GSE147352 in the GEO database (https://www.ncbi.nlm.nih.gov/geo), and 20 normal brain tissues obtained from the CGGA database (https://www.cgga.org.cn/). Data on exon usage frequency were obtained from the BrainSpan database (https://www.brainspan.org/). CUT&Tag data used in this study were obtained from GSE183627 in the GEO database (https://www.ncbi.nlm.nih.gov/geo), and ChIP-seq data were obtained from EGAD00001004335 in the EGA database (https://ega-archive.org/). All other data are available from the corresponding author on reasonable request. The code used to generate the results in this study is publicly available at the GitHub repository (https://github.com/byebye1034/1_UCR_project), and the same version of the code has been deposited in Zenodo (10.5281/zenodo.17373904)58.