CRONDEX: a web-based platform for exploring links between chromatin-related genes and neurodevelopmental disorders Authors: Javier Guerrero-Flores, Bárbara S Viegas, Marian A Martínez-Balbás, Xavier de la Cruz Published: 2026 ABSTRACT AbstractBackgroundChromatin regulatory genes are essential for orchestrating neurodevelopment by controlling DNA accessibility and the expression of genes required for neuronal differentiation and maturation. Mutations affecting these regulators—including histone modifiers, chromatin remodelers, and chromatin-binding factors—are a major cause of neurodevelopmental disorders (NDDs), many of which display overlapping yet mechanistically heterogeneous clinical features. As the catalogue of chromatin-related NDD genes has expanded, so too has the availability of phenotypic, functional, and pathway annotations. However, these data remain dispersed across multiple resources, making it difficult to integrate them in order to systematically compare genes or explore shared mechanisms. Existing platforms such as the Human Phenotype Ontology and the Monarch Initiative offer powerful phenotype–gene mapping tools, but they operate across the full genomic space and do not provide the focused framework needed to examine chromatin-related NDD genes as a coherent group.ResultsWe present CRONDEX (ChROmatin and NeuroDevelopmental disorder-related genes EXploratory platform), a user-friendly web resource developed to meet this need by supporting integrative exploration of relationships among chromatin-related genes implicated in NDDs. The database was constructed by intersecting curated chromatin (EpiFactors) and neurodevelopmental (SysNDD+Orphanet+SFARI+NDD-GeneHub) gene sets while excluding transcription factors, and enriched with annotations from Gene Ontology, KEGG, and the Human Phenotype Ontology. CRONDEX provides two complementary query modes: (i) a Gene-Based Query, which identifies genes with phenotypic profiles similar to a user-defined target using Jaccard similarity metrics; and (ii) a Criteria-Based Query, which retrieves genes matching specific phenotypic, functional, or pathway filters. Through representative examples, we show how the Gene-Based Query recovers biologically coherent relationships—from paralogous pairs such as CREBBP–EP300 to functionally convergent modules like the BAFopathies—while the Criteria-Based Query enables hypothesis-driven exploration, exemplified by the intersection of thermogenesis and chromatin-related NDDs, and by chromatin-binding genes linked to status epilepticus. By enabling rapid, integrative, and hypothesis-oriented analyses, CRONDEX facilitates the discovery of shared mechanisms across chromatin-related NDDs and supports both basic and translational research. The platform is freely accessible at https://jgf-bioinfo.shinyapps.io/CRONDEX/.ConclusionCRONDEX provides a domain-focused platform that enables phenotype-guided exploration of chromatin-related genes, supporting mechanistic hypothesis generation in NDDs.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13148-026-02176-z.Keywords: Chromatin regulation, Neurodevelopmental disorders, BAFopathies, Chromatinopathies, Histone methylation, Chromatin remodeling, Phenotypic similarity, Human Phenotype Ontology, Gene prioritization, Bioinformatics platform Background At the molecular level, chromatin is central to neurodevelopment [1, 2]. By regulating DNA accessibility and higher-order genome organization [3, 4], chromatin orchestrates the temporal and spatial expression of genes involved in neural progenitor proliferation, differentiation, and synaptic maturation. This control is executed by histones and epigenetic regulators, which form a finely tuned network that translates developmental signals into coordinated transcriptional changes [5, 6]. The complexity of this system makes it susceptible to multiple sources of disruption, which can lead to clinically significant consequences for neurodevelopment [1]. In fact, mutations in histones and epigenetic regulators have been increasingly identified as the underlying cause of a significant number of neurodevelopmental disorders (NDDs) [7–9]. For example, Rett syndrome, a severe disorder characterized by developmental regression and motor and cognitive impairment, is caused by mutations in MECP2, which encodes a transcriptional repressor involved in binding to methylated DNA and recruitment of histone deacetylases [10]. Kleefstra syndrome, marked by intellectual disability, hypotonia, and distinctive facial features, is associated with mutations in EHMT1, which encodes a histone methyltransferase responsible for the dimethylation of histone H3 at lysine 9 (H3K9me2), a key mark of transcriptional repression [11]. Rubinstein–Taybi syndrome, which presents with developmental delay, broad thumbs, and characteristic facial morphology, involves mutations in the histone acetyltransferase genes CREBBP or EP300, whose products promote transcription through histone acetylation and chromatin remodeling [12]. These findings illustrate how pathogenic, system-wide alterations in gene expression programs can result from chromatin dysregulation driven by specific genetic defects in each disorder. The increasing availability of molecular and clinical data on chromatin-related NDDs has been leveraged by recent studies to advance our understanding of these disorders, revealing shared mechanisms and potential relationships among them. By compiling data on gene function, regulatory pathways, and clinical presentation, these works highlight common features across seemingly distinct conditions. For instance, [7] analyze the convergence of multiple chromatinopathies on common neurodevelopmental pathways, discussing the resulting potential for shared therapeutic strategies. Similarly, [9] provide a comprehensive overview of epigenetic mechanisms and their links to developmental phenotypes, while [13] discuss how episignatures and clinical clustering can aid in the classification and diagnosis of chromatinopathies. In a clinical context, [14] present a five-year experience integrating genomic, phenotypic, and DNA methylation data in the care of patients with Mendelian disorders of the epigenetic machinery. These studies not only illustrate how mutations in different chromatin-related genes result in distinct syndromes—an expected consequence of their diverse enzymatic roles (e.g., methyltransferases, demethylases, acetyltransferases). They also highlight the role of chromatin as a unifying force that introduces varying degrees of overlap among the clinical phenotypes of NDDs. The presence of these shared phenotypic features adds an additional layer of complexity, underscoring the need for tools that go beyond molecular-level properties (e.g., sequence, structure, biochemical function) to identify relationships across broad and specific phenotypic categories. Such tools should help guide the interpretation of chromatin-related mechanisms in NDDs and support biologically grounded hypotheses. Here we present CRONDEX (ChROmatin and NeuroDevelopmental disorder-related genes EXploratory platform), a user-friendly tool designed to implement this paradigm and freely accessible at https://jgf-bioinfo.shinyapps.io/CRONDEX/. CRONDEX enables access to chromatin–NDD gene information through an intuitive query system (Fig. 1, Supplementary Fig. 1) that allows users to (i) input a gene of interest to find others with similar clinical phenotypes, or (ii) retrieve genes that meet specific biological criteria—such as localization to a particular cellular compartment or involvement in a defined biological pathway. In the first case, CRONDEX returns a list of genes ranked by phenotypic similarity to the query gene; in the second, it returns those genes matching the search conditions. In both cases, candidate genes are listed along with their associated annotations. Users can then refine their analysis through single-gene queries or by conducting targeted comparisons using the ‘Compare two genes’ feature. Overview of the CRONDEX interface and its two query modes. The upper portion shows the section of the landing page containing the buttons that launch the two available query modes. The lower portion illustrates how each mode operates, highlighting their respective inputs and the resulting outputs—namely, gene lists returned either by phenotypic similarity (Gene-Based Query) or by user-defined biological criteria (Criteria-Based Query) CRONDEX is, to our knowledge, the first platform dedicated to chromatin–NDD gene exploration, with a user-friendly interface that makes it accessible to users from diverse backgrounds. It streamlines the analysis of functional and phenotypic relationships among these genes, thereby enriching our understanding of chromatin-related mechanisms in NDDs. Methods Gene dataset construction The CRONDEX gene dataset was constructed to support the study of genes that are both involved in chromatin regulation and associated with NDDs. For this purpose, we used a database-driven strategy based on curated external resources rather than ad hoc manual selection. This design offers three main advantages: it makes the dataset transparent and reproducible, because each gene can be traced back to explicit sources; it facilitates future updates as the underlying databases evolve; and it reduces the practical limitations of manual expert curation across areas that require distinct expertise, including chromatin biology, neurodevelopmental genetics, and clinical phenotyping. The core gene dataset used in CRONDEX was established (Fig. 2) by compiling genes associated with human NDDs and intersecting this list with genes related to chromatin structure and function (CHRO). Construction of the CRONDEX gene dataset. The schematic illustrates how the final set of 324 chromatin-related genes was obtained by intersecting the chromatin list from EpiFactors with the neurodevelopmental list NDD (SysNDD+Orphanet+SFARI+NDD-GeneHub) and subsequently removing transcription factors from JASPAR (hashed box) Genes classified as transcription factors (TFs) were subsequently excluded. This was a deliberate design choice to preserve the conceptual focus of CRONDEX on epigenetic regulators, here defined as proteins directly involved in shaping chromatin states and epigenomic architecture, including writers, erasers, readers, chromatin remodelers, histones, and associated chromatin complexes. Although TFs are essential regulators of neurodevelopment and often cooperate with chromatin regulators, their primary mode of action relies on sequence-specific DNA binding and transcriptional activation or repression, even when this occurs in cooperation with chromatin-modifying machinery. Including TFs would therefore substantially broaden the scope of CRONDEX and blur the distinction between transcriptional and chromatin-based regulation, reducing the specificity and interpretability of analyses centered on chromatin-related mechanisms in NDDs. We acknowledge that the boundary between these categories is not always sharp; the inclusion of TFs could be considered in future versions as the platform evolves. Formally, the final gene set was defined as (CHRO ∩ NDD) ∖ TF. The following resources and criteria were applied to define the CHRO, NDD, and TF gene sets. Neurodevelopmental disorder genes (NDD gene set) Genes associated with NDDs were compiled from four curated resources, with SysNDD used as the core source. SysNDD [15] is a curated database of genes linked to neurodevelopmental disorders; genes were selected based on documented associations with NDD phenotypes (fields: entities_ndd_phenotype and entities_ndd_phenotype_word), and entries without NDD associations were excluded. To broaden the coverage of the NDD gene set, we additionally incorporated genes from three complementary curated resources: the Orphanet [16] rare genetic intellectual disability gene list (ORPHA:183757), after excluding entries with adult onset and retaining genes with paediatric, missing, or unspecified onset information; SFARI Gene [17], focused on human genes implicated in autism spectrum disorder; and NDD-GeneHub [18], a genotype–phenotype knowledgebase for neurodevelopmental disorders. Formally, the NDD gene set was defined as (SysNDD ∪ Orphanet ∪ SFARI ∪ NDD-GeneHub). The four resources are available through the SysNDD, Orphanet/OrphaID, SFARI Gene, and NDD-GeneHub public web portals, respectively: https://sysndd.dbmr.unibe.ch/; https://id-genes.orphanet.app/ithaca/; https://gene.sfari.org/tools/; and https://ndd-genehub.org/downloads/. All were last accessed 20/05/2026. Chromatin-related genes (CHRO gene set) The list of chromatin-related genes was obtained from EpiFactors, a manually curated database cataloguing epigenetic regulators, histones, and chromatin-related protein complexes [19]. Data were retrieved from the “Histones” and “Proteins” tables, excluding protamines, and gene identifiers (GeneID) were extracted in CSV format. The database is accessible at: https://epifactors.autosome.org/ (last accessed 20/05/2026). Transcription factors (TF gene set) Transcription factors were retrieved from JASPAR, an open-access database of TF binding profiles [20]. Human TFs were obtained via SQLite queries and compared to the gene list under study. Overlapping entries were excluded to ensure TFs were removed from the final dataset. JASPAR is available at: https://jaspar.elixir.no/ (last accessed 20/05/2026). Data processing workflow Gene identifiers (HGNC symbols or GeneID) were extracted and standardized across datasets. Gene identifiers (HGNC symbols or GeneID) were extracted and standardized across datasets. Identifiers were converted between Gene Symbol and Entrez ID using dedicated R functions. Identifiers were converted between Gene Symbol and Entrez ID using dedicated R functions. The intersection of EpiFactors and NDD provided the initial list of candidate genes. The intersection of EpiFactors and NDD provided the initial list of candidate genes. TFs from JASPAR were excluded from this list. TFs from JASPAR were excluded from this list. The final gene list comprised 324 genes (Fig. 2; Supplementary Table 1). Database construction and structure The CRONDEX database (version finalized July 2025) is stored as an R object (genes_database.rds). It contains curated information on 324 chromatin-related genes associated with NDDs, including clinical phenotypes, Gene Ontology terms, KEGG pathways, gene expression data, and membership in epigenetic complexes. This set of annotations was collected using R packages and APIs on February 15, 2025. Future CRONDEX updates will follow the same pipeline, with annual re-importing of the underlying databases and annotations. Gene identifiers and descriptions Gene-level information was retrieved via the rentrez R package querying the NCBI Entrez API, including: HGNC gene symbols HGNC gene symbols NCBI Gene IDs NCBI Gene IDs Gene descriptions from NCBI annotations Gene descriptions from NCBI annotations Functional and disease annotations Annotations were collected from multiple curated resources via R packages and APIs: Phenotypes and Diseases: Human Phenotype Ontology (HPO) [21] API via httr and jsonlite. ORPHA diseases were excluded. Phenotypes and Diseases: Human Phenotype Ontology (HPO) [21] API via httr and jsonlite. ORPHA diseases were excluded. Neurodevelopmental Disorders: SysNDD, Orphanet/OrphaID, SFARI Gene, and NDD-GeneHub databases. Neurodevelopmental Disorders: SysNDD, Orphanet/OrphaID, SFARI Gene, and NDD-GeneHub databases. Modifications and Epigenetic Complexes: EpiFactors database annotations. Modifications and Epigenetic Complexes: EpiFactors database annotations. Gene Ontology (GO) Terms [22] : biomaRt package via Ensembl BioMart. Gene Ontology (GO) Terms [22] : biomaRt package via Ensembl BioMart. KEGG Pathways [23]: KEGGREST package for KEGG API. KEGG Pathways [23]: KEGGREST package for KEGG API. Spatial and cell-type-specific gene expression data were integrated from: Allen Brain Atlas [24]: Human brain RNA-seq dataset (donor H0351.2001). Allen Brain Atlas [24]: Human brain RNA-seq dataset (donor H0351.2001). Brain RNA-Seq (RRID:SCR_013736) [25]: Expression profiles across brain tissues and cell types. Brain RNA-Seq (RRID:SCR_013736) [25]: Expression profiles across brain tissues and cell types. CRONDEX implementation CRONDEX was implemented as a web application using the Shiny framework in R. The backend utilizes the tidyverse and dplyr packages for data manipulation, while visual outputs are generated with ggplot2. All data are stored as RDS objects and dynamically loaded to support responsive user interactions. The application offers two primary query modes—phenotypic similarity search and criteria-based search—through a user-friendly web interface. Jaccard index The phenotypic similarity search—referred to throughout as the Gene-Based Query—is based on pairwise comparison of genes using their HPO phenotype annotations. The Jaccard index was selected for this purpose for its interpretability, computational transparency, and natural decomposability into parent and child HPO term contributions—properties particularly suited to an exploratory tool. Its use for HPO-based phenotypic comparison is consistent with established practice, including its implementation in the Monarch Initiative platform [26]. The Jaccard index is a widely used metric to assess similarity between two sets A and B, based on the ratio of their intersection to their union. Scores range from 0 (no overlap) to 1 (identical sets) and are defined as follows: J(A,B) =|A ∩ B| / |A ∪ B|​ Specifically, in our application, the sets A and B correspond to the clinical phenotypes associated with two genes, as annotated in the Human Phenotype Ontology (HPO). In the Gene-based Query Jaccard values are used to generate relative rankings of candidate genes. Because our dataset is of moderate size (324 genes) and the depth of HPO annotations varies across genes, we emphasize biological interpretability over formal statistical testing. The similarity scores should therefore be regarded as measures that highlight potentially relevant relationships for further exploration and experimental validation. To assess whether the choice of similarity metric affects practical results, we compared Jaccard-based rankings against those produced by two semantic similarity measures that exploit the hierarchical structure of HPO—the Lin [27] and Jiang–Conrath [28] indices—through an exhaustive pairwise analysis of all annotated genes in CRONDEX. For Jaccard values ≥ 0.2, the conservation of top-ranked hits between Jaccard and both semantic metrics exceeded 95% across the top 1, 3, 5, and 10 positions; above Jaccard ≥ 0.3, agreement was virtually complete. These results indicate that, in the range of similarity values of practical interest, the three metrics are equivalent, and the interpretability advantages of the Jaccard index are retained without loss of ranking accuracy. A user guide to interpreting gene-based query outputs The step-by-step execution of the Gene-Based and Criteria-Based Queries is described in Supplementary Figs. 2–4. In this section, we provide a series of instructions for the use of the Gene-Based Query focusing on the outputs of this search. For each candidate gene retrieved, the output provides three types of information organised into the following blocks: gene identifiers, scores, and annotations. The scores block reports three complementary metrics. The Jaccard similarity (range 0–1) provides the primary ranking: the higher the value, the greater the phenotypic overlap with the query gene. Users can apply a threshold directly in the interface to expand or restrict the number of candidates retrieved (Supplementary Fig. 3). The median-based z-score contextualises each pairwise Jaccard value relative to the empirical distribution of all gene–gene comparisons in the dataset; values above 2.5–3 indicate that the observed similarity is less likely to reflect annotation overlap by chance. The parent/child decomposition separates the Jaccard contribution of broad HPO terms (parent terms, reflecting general clinical categories) from specific ones (child terms, reflecting precise phenotypic features); a higher child-term contribution indicates greater phenotypic specificity of the similarity. The annotations block has been designed to help users assess the reliability of each result based on the HPO annotation depth of the candidate genes—and of the query gene, whose information is displayed in the upper section of the output. This is relevant because annotation depth directly influences the reliability of similarity scores. As a general rule, comparisons supported by few HPO terms (less than 30–50) should be interpreted with caution, even when other metrics appear favorable. Therefore, before launching the Gene-Based Query, users should consider the annotation depth of their gene of interest (easily checked using the Single-Gene Query, see Supplementary Fig. 1). If this is the case, users may utilize the Criteria-Based Query instead, where combination of a few terms with annotations from GO or KEGG may provide a reduced list of potentially useful candidates. Since the Criteria-Based Query does not rank its results, the ‘Compare Two Genes’ option (Supplementary Fig. 1) can then be used to perform a more detailed similarity analysis and prioritize among the retrieved candidates. Across all metrics, results should be regarded as valuable starting points for hypothesis generation and further investigation, rather than definitive associations. Results CRONDEX enables users to explore chromatin-related NDD genes through a flexible query system based on phenotypic and functional properties. Upon accessing the landing page (Fig. 1, Supplementary Fig. 1), users encounter four main options: the two primary query modes, Gene-Based Query and Criteria-Based Query, and two additional tools for gene-to-gene comparison or single-gene inspection. Because the latter functions are straightforward and mainly provide direct access to information already available in CRONDEX, they are not discussed further here. Gene-Based Query: from paralogs to functional modules This query identifies chromatin–related NDD genes that share clinical features with a user-selected gene. CRONDEX performs this by comparing the user’s gene against all genes in the dataset using their clinical phenotypes derived from HPO annotations (Fig. 1). For each gene–gene comparison, we use a continuous similarity score (Jaccard similarity, Jas) that ranges from 0 (no shared phenotypes) to 1 (identical phenotype profiles). Once the Gene-Based Query option is selected, the search is launched by choosing a gene of interest from the Genes menu (Supplementary Fig. 2). Results are displayed as a ranked table, with the number of retrieved genes determined by the user-defined Jaccard threshold, set to 0.4 by default and adjustable when needed (Supplementary Fig. 3). As explained in the Methods, Jaccard values are not associated with a statistical model and should therefore not be interpreted as measures of statistical significance. Instead, they provide a phenotype-based ranking of genes relative to the query gene, intended to guide the assessment of biological significance. To illustrate how the Gene-Based Query operates and the types of biomedically meaningful results it yields, we present three examples that progress from straightforward paralog recovery to more complex cases of functional convergence and module-level clustering. Beyond illustrating the simplicity of this query, these examples show how it can identify relationships of genuine biomedical and mechanistic relevance. CREBBP and EP300—paralogous genes with overlapping clinical spectra The Gene-Based Query is particularly valuable when applied to paralogous genes, which can display varying degrees of phenotypic similarity independent of their sequence identity [29]. We illustrate this with CREBBP and EP300, two genes encoding histone acetyltransferases that share 58% sequence identity and catalyze the same enzymatic reaction—acetylation of histones H3 and H4 [12, 30]. When querying CREBBP in CRONDEX, EP300 appears as the top-ranking gene (Fig. 3A), with the reciprocal query yielding CREBBP as the top-ranking gene. CRONDEX also provides a network representation of the candidate list to facilitate visual interpretation. Gene-Based Query using CREBBP and EP300. A. Schematic representation of the output table returned by CRONDEX when CREBBP is used as the query gene, showing EP300 as the top-ranking hit. B. Network-style visualization of the retrieved genes, where nodes represent genes and edge thickness reflects the degree of phenotypic similarity among them; the orange node corresponds to the user-selected query gene. C. Illustration of the Jaccard similarity calculation for the CREBBP–EP300 pair, including a schematic of the formula and the position of the resulting value within the empirical distribution of pairwise similarities across all genes in CRONDEX. D. Decomposition of the Jaccard similarity into contributions from shared HPO parent terms and shared HPO child terms. Venn diagrams indicate the number of common terms at each level, while the coloured bar shows the relative contribution of each level to the overall similarity The Jas score for the CREBBP-EP300 pair is 0.835. This represents one of the highest similarity values in the dataset’s pairwise comparisons (Fig. 3C), indicating the strong phenotypic overlap between the two genes. This overlap reflects the fact that mutations in both genes cause two variants of the same disorder [12]: Rubinstein–Taybi syndrome type 1 (CREBBP) and type 2 (EP300). For further exploration, the Compare two genes option provides a breakdown of the Jas score into shared HPO parent and child terms (Fig. 3D), distinguishing broad clinical categories from more specific phenotypic manifestations. It also compares genes across clinical phenotypes, protein complexes, chromatin modifications, Gene Ontology terms, cellular expression, and brain tissue expression, with results displayed as interactive and downloadable tables and plots. KMT2D and KDM6A—distinct mechanisms, shared phenotype CRONDEX can also reveal phenotypic similarities—undetectable by conventional sequence comparisons—between non-paralog genes with distinct enzymatic functions. For example, when we query KMT2D, a well-characterized H3K4 methyltransferase [31] (Fig. 4A), CRONDEX identifies KDM6A, an H3K27 demethylase [31], as the most phenotypically similar gene in the dataset (Jas = 0.832; Fig. 4B and C), and KMT2D ranks first in the reciprocal KDM6A query. Again, this value represents one of the highest similarity values in the dataset’s pairwise comparisons (Fig. 4D). Gene-Based Query illustrating phenotypic similarity between KMT2D (methyltransferase) and KDM6A (demethylase). This case illustrates how the Gene-Based Query captures strong phenotypic similarity between genes with distinct enzymatic activities. A Structural representation (PDB code: 1EQZ) illustrating the enzymatic functions of KMT2D (green) and KDM6A (pink), highlighting their respective roles in histone modification. B Network-style visualization of genes retrieved by the Gene-Based Query, showing relationships based on shared phenotypic similarity; the orange node corresponds to the user-selected query gene. C Schematic representation of the output table returned by CRONDEX for this query, listing the top-ranking candidate genes. D Illustration of the Jaccard similarity calculation for the KMT2D–KDM6A pair, including the schematic formula and the position of the resulting value within the empirical distribution of pairwise similarities across all CRONDEX genes This similarity, nearly identical to that observed between the paralogs CREBBP and EP300 (0.835), reflects the fact that mutations in both genes cause markedly similar clinical presentations of Kabuki syndrome (type 1 for KMT2D and type 2 for KDM6A) [31]. Importantly, this phenotypic overlap is not easy to infer from their enzymatic activities alone: KMT2D functions as a transcriptional activator through H3K4me3 deposition, whereas KDM6A removes the repressive H3K27me3 mark—distinct mechanisms of transcriptional regulation that nevertheless yield overlapping developmental phenotypes. This case illustrates how, for researchers exploring H3K4- or H3K27-related pathways, a simple CRONDEX query can rapidly reveal valuable functional relationships that conventional approaches are likely to miss. BAFopathies—functional modules with overlapping clinical spectra The previous examples demonstrated that Gene-Based queries can identify biomedically relevant candidates based on phenotypic similarity. We conclude this section with a case where such similarity stems from dysfunction within the same molecular machine: the BAFopathies [32]. This group of closely related neurodevelopmental disorders—including multiple forms of Coffin–Siris syndrome and Nicolaides–Baraitser syndrome—is characterized by intellectual disability, speech delay, coarse facial features, and other congenital anomalies. These disorders share a common cause: disruptive mutations in subunits of the BAF chromatin-remodeling complex (Fig. 5A), which regulates nucleosome positioning and is essential for neural development, cell differentiation, and DNA repair. Gene-Based Query illustrating the BAFopathies case. A Three-dimensional structure of the human BAF chromatin-remodelling complex (PDB code: 6LTJ), in which only the eleven subunits associated with BAFopathies and present in CRONDEX are coloured. Arrows indicate the names of these subunits within the complex. B Network-style visualization of genes retrieved by the Gene-Based Query using ARID1A as the seed gene. Nodes represent genes, edge thickness reflects phenotypic similarity, and the large orange node corresponds to the user-selected query gene. C Bar plot summarizing the performance of Gene-Based Queries seeded with each of the eleven BAFopathy-associated genes. For each gene, bar height indicates the number of BAFopathy genes recovered among the top ten ranked results of the corresponding query, highlighting the phenotypic coherence of the BAF module. The number inside each bar indicates the total number of HPO phenotypes associated with each gene CRONDEX contains eleven genes encoding canonical BAF (cBAF) subunits that are associated with established BAFopathies (ARID1A, ARID1B, DPF2, SMARCA2, SMARCA4, SMARCB1, SMARCC1, SMARCC2, SMARCD1, SMARCD2, and SMARCE1) [32–34]. Remarkably, a query seeded with any one of them typically retrieves the other ten among its top results (Fig. 5B and C). For instance, using ARID1A (Coffin–Siris syndrome type 2, Fig. 5B) as a query returns, within the top ten positions, eight of the ten cBAF genes. The only top-ranked gene outside this set is ARID2, a BAFopathy-associated gene that belongs to the PBAF (polybromo-associated BAF) complex, where it replaces ARID1A or ARID1B as the defining ARID subunit. The four top-ranking genes (Jas 0.629–0.766) all belong to the cBAF set and share its core neurodevelopmental phenotype, despite not being ARID1A paralogs. ARID1B (Jas 0.476; Supplementary Fig. 5A), an ARID1A paralog (49% sequence identity) and cause of Coffin–Siris syndrome type 1, appears only at the sixth position. Three exceptions arise from the cBAF-seeded queries: SMARCA2, SMARCC1, and SMARCD2 (Fig. 5C). For SMARCC1 and SMARCD2, sparse HPO annotations (2 and 30 terms, respectively) likely limit the reliability of the similarity scores. SMARCA2, despite its extensive annotation (152 HPO terms), diverges because it is associated with Nicolaides–Baraitser syndrome, a disorder related to Coffin–Siris syndrome within the broader BAFopathy spectrum [35]. This clinical coherence is a major strength of our phenotype-driven approach. By contrast, a STRING database query using SMARCA2 yields an interaction network in which phenotypic relationships between genes are not explicitly reflected (Supplementary Fig. 5B). CRONDEX therefore provides a disease- and phenotype-aware perspective on gene relationships, offering valuable complementary information to interaction-based resources such as STRING. Disease coherence and discovery potential of the Gene-Based Query In this section, we go beyond the examples presented above and provide a broader view of how the Gene-Based Query connects chromatin–NDD genes across the entire dataset, and of the type of relationships it can reveal that conventional sequence-based approaches would miss. We first examined how chromatin–NDD genes organize when compared by phenotypic similarity. To this end, we performed an exhaustive pairwise comparison of all annotated genes in CRONDEX and visualised the result as a dendrogram (Supplementary Fig. 6). Visual inspection of the dendrogram reveals that genes associated with the same or closely related disorders tend to cluster together without any disease label being used as input. This is the case, for instance, for genes causing Coffin–Siris syndrome, Rubinstein–Taybi syndrome, Kabuki syndrome, Cornelia de Lange syndrome, and Fanconi anemia, each forming a coherent cluster within the dendrogram. Notably, the Cornelia de Lange-related cluster groups NIPBL, HDAC8, and BRD4—a cohesin loader, an SMC3 deacetylase, and a bromodomain acetylation reader, respectively—based exclusively on shared phenotypic features. This type of relationship, connecting genes that act through different chromatin-based mechanisms but converge on a common clinical presentation, is not readily recoverable by sequence comparison alone. This last point is directly illustrated by Fig. 6, which plots, for each gene pair, the strength of phenotype-based similarity against pairwise protein sequence identity. Multiple pairs with high phenotype-based similarity fall below the 40% sequence-identity threshold commonly associated with functional inference. Together, these results suggest that the Gene-Based Query captures biologically meaningful relationships that sequence similarity cannot reveal, supporting its utility for hypothesis generation in the chromatin–NDD field. Phenotypic similarity vs. sequence identity for the top-ranked Gene-Based Query hits in CRONDEX. Each point represents a gene retrieved as the top-ranked hit of a Gene-Based Query performed with another gene as input. The y-axis shows the pairwise sequence identity (%) between both genes, and the x-axis shows the Z-score of their phenotypic similarity, reflecting how much the observed similarity exceeds background levels. The horizontal dashed line marks the 40% sequence identity threshold below which functional similarity cannot be reliably inferred from sequence alone. Coloured points correspond to potential discoveries, i.e., queries where the retrieved gene would not have been identified as a candidate of interest using sequence comparisons. Selected genes are labelled Criteria-Based Query: from phenotype filters to mechanistic insights While the Gene-Based Query is suited to identifying genes with broadly similar phenotypic profiles, the Criteria-Based Query allows users to focus on selected phenotypes or mechanistic features. This is useful when specific clinical features of interest might be masked within the full HPO annotation set used for Jaccard-based comparisons. Users can combine HPO terms with other annotations, such as Gene Ontology categories or KEGG pathways, to retrieve genes sharing both clinical and mechanistic properties. Launching this search is straightforward: after selecting the Criteria-Based Query option, the user defines the desired criteria using menus on the left panel (Supplementary Fig. 4) and presses the Query button to retrieve results. In contrast to the Gene-Based Query, the output is an unranked list of genes meeting the selected criteria (Fig. 1). To illustrate how the Criteria-Based Query operates in practice, we present three examples that demonstrate its simplicity and flexibility. These examples show how even a small number of well-chosen criteria can reveal biomedically meaningful gene relationships. Finding chromatin genes that connect histone methylation and intellectual disability To provide a first illustration of the results obtained with the Criteria-Based Query, we selected an example that highlights how this query offers a general approach to explore the relationship between chromatin-related genes and NDDs, distinct from the more gene-centered perspective implemented in the Gene-Based Query. In this example, we searched for chromatin-related genes associated with intellectual disability, a frequent clinical phenotype in NDDs—for example, present in 151 of the 324 genes in our dataset. Simultaneously, we required functional linkage to the regulation of histone marks that control gene-expression programs, more specifically those involved in histone methylation. At the user level, this query can be expressed in different ways within CRONDEX. For instance, we can use the following filters: Phenotype: Intellectual disability (HP:0001249) Phenotype: Intellectual disability (HP:0001249) Modification: Histone methylation Modification: Histone methylation Modification, in this case, specifically refers to the Epifactors annotation field of the same name (see Methods). This query retrieves thirty-two genes (Fig. 7; Supplementary Table 2), encoding both methylases and demethylases. This broad output reflects the annotation strategy used in Epifactors, where, for histone methylation, epigenetic regulators are grouped under the same annotation regardless of their specific functional role (e.g., writer, eraser, reader). All thirty-two genes have been implicated in NDDs featuring intellectual disability and provide a first framework for exploring the different facets of histone methylation in neurodevelopment and cognition. Criteria-Based Query identifying chromatin regulators involved in histone methylation and intellectual disability. A Schematic representation of a nucleosome highlighting the N-terminal tails of histones H3 and H4 and the lysine residues subject to methylation and demethylation. NDD-associated histone methyltransferases (“writers”) and histone demethylases (“erasers”) are indicated with green and red arrows, respectively. B Summary of genes retrieved by the Criteria-Based Query combining two terms: intellectual disability and histone methylation–related annotations. The outer box represents genes retrieved using EpiFactors histone methylation annotations, grouped into three boxes corresponding to writers (green), erasers (yellow), and readers & others (magenta). The inner green box highlights the subset of these genes that are retrieved when using the KEGG lysine degradation pathway (hsa00310), illustrating the overlap and complementarity between the two annotation frameworks Alternatively, a conceptually equivalent query can be implemented using the following filters: Phenotype: Intellectual disability (HP:0001249) Phenotype: Intellectual disability (HP:0001249) Pathway: Lysine degradation (KEGG: hsa00310) Pathway: Lysine degradation (KEGG: hsa00310) Pathway, in this case, refers to a specific category in the KEGG database. We selected the Lysine degradation pathway because histone lysine residues—such as H3K4, H3K9, and others—are the sites of methylation marks that underpin the epigenetic control of transcription. In KEGG, the enzymatic reactions responsible for these histone lysine modifications are grouped under the Lysine degradation pathway (hsa00310). This query retrieves twelve histone-lysine methyltransferases (Fig. 7; Supplementary Table 2). Functionally, these include members of the KMT2/SETD1 (COMPASS) family of histone H3K4 methyltransferases [36], as well as other methyltransferases such as EHMT1, EZH2, KMT5B, and NSD1/NSD2, which establish repressive or elongation-associated modifications (H3K9, H3K27, H4K20, and H3K36). As before, all these genes have been implicated in NDDs whose clinical features include intellectual disability. All twelve genes identified through the KEGG query were also retrieved in the Epifactors-based search (Supplementary Table 2). The latter additionally includes multiple classes of chromatin regulators, such as demethylases (e.g., KDM3B, KDM4B, KDM5A, KDM6A, PHF8) and readers (e.g. PHF21A, MECP2), consistent with its broader functional scope. These differences reflect the distinct annotation frameworks of the two databases: KEGG classifies enzymes according to their reaction chemistry, whereas Epifactors curates proteins based on their biological roles in chromatin regulation. Together, the two views reveal complementary aspects of the same biological process. The resulting gene sets, easily obtained through the Criteria-Based Query, provide a useful starting point for studying histone methylation and its contribution to NDDs—for instance, by comparing the expression programs regulated by each protein. Using phenotypic and mechanistic constraints to explore chromatin-related status epilepticus The Criteria-Based Query can also be used to collect information relevant to the study of clinical phenotypes of interest. For example, it can help determine whether a given phenotype appears across multiple NDDs under specific molecular constraints, thereby supporting subsequent comparative studies. Here, this is exemplified with the clinical phenotype status epilepticus. In Rett syndrome caused by MECP2 mutations [37], around 20–25% of patients experience status epilepticus [38], a severe and prolonged seizure type associated with significant clinical risk [39], making it important to better comprehend how this complication arises. To explore potential mechanistic commonalities and distinctions, we then used the Criteria-Based Query to identify chromatin-related NDD genes associated with this phenotype. To this end, we applied a Criteria-Based Query using: Phenotype: Status epilepticus (HP:0002133) Phenotype: Status epilepticus (HP:0002133) Gene Ontology: Chromatin binding (GO:0003682) Gene Ontology: Chromatin binding (GO:0003682) In addition to status epilepticus—the phenotype of interest—we included the GO category chromatin binding (GO:0003682) as a mechanistic constraint. This filter restricts the search to non-enzymatic chromatin regulators that, like MECP2 (a methylated DNA-binding protein), physically bind chromatin without directly catalysing chemical modifications—essentially readers, remodelers, or structural chromatin factors—thereby enabling closer mechanistic parallels with MECP2. The four genes retrieved (Fig. 8) are MECP2; CHD2, an ATP-dependent chromatin remodeler; ACTL6B, a neuronal BAF subunit; and SMARCA2, the ATPase core of the BAF complex. Criteria-Based Query identifying chromatin-binding genes associated with status epilepticus. Left block: Schematic representation of the four genes retrieved by the Criteria-Based Query—MECP2, CHD2, ACTL6B, and SMARCA2—illustrated as simplified protein–chromatin interactions. The diagram shows generic nucleosome representations with the corresponding proteins depicted in contact with chromatin, reflecting different non-enzymatic mechanisms of chromatin regulation. Right block: Schematic overview of neuronal phenotypes linked to status epilepticus. Neuronal hyperexcitability is shown as the central phenotype associated with MECP2, with surrounding illustrations indicating major contributing features linked to the other retrieved genes, including reduced synaptic inhibition (CHD2), altered dendritic and synaptic maturation (ACTL6B), and impaired activity-dependent transcription (SMARCA2). Together, the figure highlights how distinct chromatin-binding proteins converge on a shared clinical outcome Although their molecular roles differ, mutations in these genes converge on the shared outcome of neuronal hyperexcitability [40], the basis of the persistent seizures observed in MECP2-related Rett syndrome (Fig. 8) [41]. Notably, mutations in each gene can be linked to a prominent component of this neuronal phenotype (Fig. 8): reduced synaptic inhibition (CHD2) [42], immature dendritic and synaptic architecture (ACTL6B) [43], or impaired activity-dependent transcription (SMARCA2) [44]. Although these contributions may overlap, the major effect associated with each gene provides a distinct entry point for understanding specific components of the status epilepticus phenotype through comparative studies. Linking thermogenesis and chromatin-related NDDs The following example is intentionally exploratory. Rather than presenting a confirmed biological link, it illustrates how the Criteria-Based Query can support cross-domain explorations connecting chromatin regulation and neurodevelopment to other biological processes—in this case, energy metabolism. Recent work has revealed that chromatin regulators not only control developmental programs but also dynamically shape cellular energy metabolism through stimulus-dependent transcriptional responses [45, 46]. To obtain a first view of how this connection might intersect with chromatin-related NDDs, we focused on thermogenesis, a metabolic process of heat production that is most prominently active in brown and beige adipocytes. For this example, we used the KEGG thermogenesis pathway (hsa04714) as a reference. Within CRONDEX, this question can be formulated using a simple Criteria-Based Query: Phenotype: Intellectual disability (HP:0001249) Phenotype: Intellectual disability (HP:0001249) Pathway: Thermogenesis (hsa04714) Pathway: Thermogenesis (hsa04714) As before, intellectual disability was chosen because of its frequency among NDDs. The query retrieves 12 genes, eight of which (Fig. 9) encode core subunits belonging to the ATPase and Base modules of the BAF chromatin-remodelling complex [47]. This complex, whose disruption underlies multiple BAFopathies [32], is also deployed in adipocytes to support thermogenesis by remodeling chromatin at heat-producing genes such as UCP1 [48]. To assess whether the identified BAF complex components are expressed in adipocytes, we examined publicly available expression data from the Human Protein Atlas [49]. Seven of the eight proteins showed high protein-level expression in adipocytes, whereas ARID1A was detected at the RNA level. Notably, independent studies have implicated ARID1A in adipogenesis [50]. Criteria-Based Query linking thermogenesis and chromatin-related neurodevelopmental disorders. Schematic representation summarizing the relationship between chromatin regulation, intellectual disability, and thermogenesis. The BAF chromatin-remodelling complex is illustrated in association with a nucleosome, with individual subunits corresponding to genes retrieved by the Criteria-Based Query. A leftward arrow indicates the link to intellectual disability and NDDs, while a rightward arrow indicates the association with thermogenesis, providing a simplified, linear representation of the KEGG thermogenesis pathway The four remaining genes (ACTB, ACTL6B, KDM3B, and RPS6KA3), all implicated in NDDs featuring intellectual disability, illustrate different degrees of potential coupling between neurodevelopment and metabolism. Their protein products are expressed in adipocytes, according to the Human Protein Atlas [49], except for ACTL6B. ACTL6B is a neuronal-specific paralog of the adipocyte-active BAF subunit ACTL6A. The inclusion of ACTL6B in the thermogenesis pathway reflects KEGG’s practice of assigning related proteins, such as paralogs, to the same functional module. Among the remainder, perhaps ACTB is the clearest case, as it belongs to the ARP module of the BAF complex [47], and also supports cellular processes that can facilitate mitochondrial positioning [51], required during thermogenesis. KDM3B is a paralog of KDM3A (JMJD1A), a histone demethylase that directly activates thermogenic gene expression [52]. This relationship, plus its expression in adipocytes, is suggestive of a potential role in energy metabolism, although confirming such a role would require experimental validation. Similar considerations apply to RPS6KA3: it is expressed in adipocytes, RSK signalling has been implicated in thermogenesis, and RPS6KA3 is one of several RSK family members that could potentially contribute to this process. However, paralog-specific evidence for RPS6KA3 in thermogenesis is currently lacking. This example illustrates both the opportunities and caveats of exploratory, criteria-based analyses. A simple combination of “intellectual disability” and “thermogenesis” highlights a coherent subset of chromatin regulators—most prominently the BAF complex—whose functions span neurodevelopment and energy metabolism, while also revealing how pathway annotation conventions (e.g., paralog grouping) can influence outputs, calling for careful data analysis. Discussion In this work, we present CRONDEX as a tool for examining how chromatin-related genes relate to NDDs. The platform uses straightforward gene-based and criteria-based queries that enable users to navigate the chromatin-related gene space and support comparative analyses and the development of mechanistic hypotheses. The usage examples presented here were chosen not only to illustrate how CRONDEX operates, but also to show how it can be applied to the exploratory tasks outlined above, where it can yield observations of biological relevance. This reflects the structure of the underlying database, which restricts all searches to genes jointly implicated in chromatin regulation and NDDs. Such a focused design increases the likelihood that retrieved candidates are biologically relevant and reduces the risk of spurious matches. The nature of the queries themselves also contributes to this behavior. The Gene-Based Query is conceptually analogous to familiar sequence-similarity searches such as BLAST [53], in which a target sequence is used to identify related entries within a database. Here, similarity is defined in terms of shared phenotypic annotations, allowing users to explore clinical proximity among chromatin-NDD genes—an approach whose value for early-stage hypothesis generation is supported by the biological coherence observed across the examples presented. The Criteria-Based Query allows users to formulate broader, concept-driven questions about the chromatin–NDDs relationship. In this mode, the query target is a set of selected gene characteristics that collectively reflect a specific biological interest—for example, the role of demethylation in intellectual disability or the potential links between NDDs and energy metabolism. The output is an unranked list of genes that meet the chosen criteria, which can then be used to extend analyses within a given topic or to suggest new research questions. For instance, the thermogenesis–NDD example presented here—arguably the most exploratory case—highlights how such queries can raise clinically and biologically relevant questions, such as whether metabolic vulnerabilities may contribute to phenotypes observed in disorders like Coffin–Siris or Nicolaides–Baraitser syndromes. A practical scenario where this approach may be useful is the mechanism-informed review of candidate genes emerging from clinical sequencing. When a candidate chromatin-related gene has been identified in a patient with an NDD phenotype, the Gene-Based Query can be used to retrieve other chromatin–NDD genes with similar HPO profiles. The phenotypes associated with these neighbouring genes may then highlight clinical features for targeted re-evaluation, as clinically appropriate, consistent with reverse phenotyping approaches in which molecular findings guide reassessment of potentially overlooked features [54, 55]. In this way, CRONDEX can help import phenotypic knowledge from better-characterized chromatin-related disorders into the interpretation of less familiar genes or cases. The Criteria-Based Query can serve a complementary role in analogous scenarios—particularly when the number of available HPO terms is limited—by allowing users to combine clinical phenotypes with functional annotations to retrieve a focused set of candidates (see below). It should be noted, however, that CRONDEX is designed as a research tool and its results should be interpreted as exploratory leads, not as diagnostic conclusions. The construction of the CRONDEX gene set introduces certain limitations worth discussing explicitly. We address them in turn, starting with dataset coverage and continuing with the reliability of Gene-Based Query results. Dataset coverage. CRONDEX was built using specialised curated databases rather than manual literature curation, which would require expert filtering across two largely independent fields—chromatin biology and neurodevelopmental disease—a task difficult to standardise and keep up to date. Database-driven curation ensures expert oversight, facilitates reproducibility, and supports periodic updates. To estimate the coverage achieved with this approach, we compared the CRONDEX gene set against a list of chromatin-NDD genes (Supplementary Table 3) compiled from three independent reviews [7, 9, 14], yielding an estimated coverage of 91.6%. This coverage, combined with the planned periodic updates to the underlying databases, makes CRONDEX a practical starting point for rapid exploration of the chromatin-NDD gene space—one that can be complemented by more targeted literature searches when needed. A related design choice concerns the exclusion of transcription factors. Although TFs frequently cooperate with chromatin regulators and some operate in a mechanistic grey zone, their inclusion would broaden the scope of CRONDEX beyond its intended focus on chromatin-based regulation. This boundary is acknowledged as a deliberate decision, and the inclusion of TFs could be considered in future versions of the platform. Looking further ahead, the extension of CRONDEX to non-coding regulatory regions would be a valuable direction for future development. This would require a region-based rather than gene-based representation, as well as a system for associating phenotypic annotations with genomic regions, and remains a longer-term prospect as the necessary resources and standards continue to develop. Reliability of Gene-Based Query results. In any exploratory process, knowing how reliable the results are is essential for deciding how to use them. In the Gene-Based Query, differences in HPO annotation depth across genes can affect the value and reliability of Jaccard scores—as illustrated by the SMARCC1 and SMARCD2 cases above. A formal statistical framework would help capture this and other sources of uncertainty in a simple p-value; however, the moderate size of the CRONDEX gene set prevents this approach. This limitation is partly offset by the focused scope of the resource, which reduces the risk of recovering non-meaningful hits by restricting searches to genes with related clinical and mechanistic profiles. Additionally, the Gene-Based Query output provides users with tools to assess result reliability directly: a z-score to gauge how much the observed similarity deviates from random comparisons; the parent/child decomposition of the Jaccard score, to detect imbalanced contributions; and HPO term counts, to identify rankings based on few terms. We also explored whether ignoring the HPO hierarchy in Jaccard computations is a practical disadvantage relative to more complex ontology-aware measures. Comparing Jaccard rankings against those of the Lin [27] and Jiang–Conrath [28] semantic indices showed (Supplementary Fig. 7) that conservation of top-ranked hits was virtually complete above Jaccard values of 0.3. This confirms that the Jaccard index recovers essentially the same information as more complex measures, without sacrificing interpretability. For cases where the Gene-Based Query yields unreliable results or is a priori non-advisable due to scarce HPO annotations available for a gene, or where the identity of the target gene is unclear, the Criteria-Based Query offers a valuable approach. By combining a few clinical phenotypes with additional GO or KEGG annotations, it is possible to rapidly reduce the number of candidates to a manageable set. To illustrate this, we performed a simulation using ARID1A as a test case, randomly sampling increasing numbers of its HPO phenotypes across 100 iterations, with and without an additional GO or KEGG constraint (Fig. 10). Using phenotypes alone, the average number of candidates drops sharply from ~ 43 with a single phenotype to ~ 8 with three, and continues to decrease gradually thereafter. Adding a single GO or KEGG term—here, myoblast differentiation (GO:0045663) or thermogenesis (hsa04714)—further reduces the candidate list from as few as two-four phenotypes onwards. These results confirm that a small number of well-chosen criteria is sufficient to obtain a focused and manageable candidate list, which can then be explored in more detail using the other CRONDEX query options. Effect of query complexity on the number of candidates retrieved by the Criteria-Based Query. The figure shows the average number of genes retrieved by the Criteria-Based Query as a function of the number of HPO phenotype terms used, based on 100 iterations of random sampling from the HPO annotations of ARID1A. Four situations are shown: HPO terms alone (Baseline), and with the addition of a single GO term—myoblast differentiation (GO:0045663)—a single KEGG pathway—thermogenesis (hsa04714)—or both combined. Using phenotypes alone, the number of candidates drops sharply with the first few terms and continues to decrease gradually thereafter. Adding a single GO or KEGG constraint further reduces the candidate list, particularly from two to four phenotypes onwards, supporting the idea that a small number of well-chosen criteria can be sufficient to obtain a focused and manageable set of candidates In addition to these methodological constraints, it is useful to consider how CRONDEX relates to existing resources that support gene or phenotype exploration. Several widely used platforms provide complementary information, but their aims differ from the type of phenotype-aware, domain-focused queries implemented here. Sequence-based tools such as BLAST [53] are invaluable for identifying homologs and for studying evolutionary or structural relationships, but sequence similarity does not reliably reflect functional or clinical proximity. As seen in the BAFopathy [32] examples, informative candidates are often not homologous to the query gene and are therefore invisible to sequence-driven searches. At the level of structural and interaction data, STRING [56] is noteworthy because it allows queries based on pathways or disease terms and returns results in the form of protein–protein interaction networks. However, it does not support combinations of filters such as those implemented in our Criteria-Based Query. On the clinical annotation side, portals such as HPO [21] and the Monarch Initiative [26] address the phenotype–gene mapping problem, and Monarch’s Phenotype Similarity Search can generate ranked gene lists based on different similarity metrics. At present, however, this feature remains under development and, like the resources above, operates across the full genomic space rather than within a mechanistically coherent subset of chromatin-related NDD genes. These platforms also lack features such as inspection of gene-specific phenotype lists, decomposition of similarity scores, or the ability to formulate complex multi-criteria queries. Conclusions CRONDEX represents the first web-based resource specifically designed for the study of chromatin-related genes in the context of NDDs. By streamlining the discovery process from hours of manual curation to minutes of interactive queries, CRONDEX enables researchers to rapidly generate biologically grounded hypotheses about chromatin dysregulation in neurodevelopment. The tool's intuitive interface makes these analyses accessible to researchers regardless of their computational background, potentially fostering new collaborations between clinicians, molecular biologists, and bioinformaticians in the chromatin-NDD field. CRONDEX is freely available via an interactive RShiny application at https://jgf-bioinfo.shinyapps.io/CRONDEX/. The platform will be updated on a regular basis to incorporate emerging clinical and molecular data as the field advances, including newly curated chromatin regulators, SysNDD, Orphanet, SFARI, and NDD-GeneHub entries, HPO annotations, and updated KEGG and GO releases as they become available. Supplementary Information Supplementary file 1: Figure 1. Landing page of the CRONDEX web application.Screenshot of the CRONDEX landing page showing the main interface layout. Two panels in the upper section provide access to the two primary query modes (Gene-Based Query and Criteria-Based Query), while a third panel in the lower section links to additional exploration options, including gene-to-gene comparison and single-gene information. Supplementary file2: Figure 2. (A) Step-by-step execution of a Gene-Based Query in CRONDEX. Three consecutive screenshots illustrating the workflow for performing a Gene-Based Query. Arrows indicate the required user actions at each step: selection of the query gene from the gene menu, confirmation of the selected gene, and activation of the Perform search option to show the results. (B) Output of the Gene-Based Query. For each candidate gene retrieved, the output provides three blocks of information: gene identifiers, scores, and annotations. The scores block reports: the Jaccard similarity (range 0–1), the median-based z-score, and the parent/child decomposition of the Jaccard score. The annotations block reports: the total number of HPO phenotypes for the candidate gene, and its breakdown into parent and child contributions. Supplementary file3: Figure 3. Adjustment of the Jaccard similarity threshold in the Gene-Based Query. Two consecutive screenshots illustrating how the Jaccard index threshold can be modified to control the number of genes retrieved and displayed by the Gene-Based Query. Lowering or increasing the threshold allows users to visualize a broader or more restrictive set of phenotypically similar genes, respectively. Supplementary file4: Figure 4. Step-by-step execution of a Criteria-Based Query in CRONDEX. Series of consecutive screenshots illustrating the workflow for performing a Criteria-Based Query. Arrows indicate the required user actions at each step, including the selection of phenotypic, functional, or pathway-based criteria, and the execution of the query to retrieve the corresponding gene list. Supplementary file5: Figure 5. Comparison between phenotype-based and interaction-based views of BAFopathy-related genes. (A) Screenshot of the Gene-Based Query results obtained using ARID1A as the seed gene, showing only the top-ranked hits retrieved based on phenotypic similarity. This query corresponds to the BAFopathies example discussed in the main text. (B) Protein–protein interaction network retrieved from the STRING database for BAFopathy-associated genes. In contrast to the phenotype-driven relationships highlighted by CRONDEX, this interaction-based view does not explicitly reflect the clinical coherence observed among BAFopathy genes. Supplementary file6: Figure 6. Hierarchical clustering of CRONDEX genes based on phenotypic similarity. Dendrogram of all annotated genes in CRONDEX, constructed from pairwise Jaccard similarity scores computed over their HPO phenotype annotations. The inset on the left indicates the region of the dendrogram shown in the expanded view on the right. In the expanded view, several examples of disease-related gene clusters are highlighted, with those corresponding to cases discussed in the main text shown in colour: Coffin–Siris syndrome (red), Rubinstein–Taybi syndrome 1 and 2 (green), Kabuki syndrome 1 and 2 (blue), Cornelia de Lange syndrome, and Fanconi anemia. Kleefstra syndrome genes form a separate cluster visible in the full dendrogram. Supplementary file7: Figure 7. Jaccard vs. Semantic Similarity Metrics. Each panel shows the conservation of top-ranked hits — top 1, 3, 5, and 10, respectively — between the Jaccard index and two semantic similarity metrics (Lin and Jiang–Conrath) as a function of the Jaccard similarity threshold. In CRONDEX, hits with Jaccard values below the user-defined threshold are not returned. Conservation is expressed as the fraction of hits shared between the Jaccard ranking and each semantic metric. Above a Jaccard threshold of 0.2, conservation of top-ranked hits between Jaccard and both semantic metrics exceeds 95%; above 0.3, agreement is virtually complete. Supplementary file8: Table 1. Chromatin-related neurodevelopmental disorder genes included in CRONDEX. Table containing the 324 genes in the CRONDEX database, with columns indicating gene symbol, NCBI Gene ID, and gene name. Supplementary file9: Table 2. Genes retrieved by the Criteria-Based Query combining intellectual disability and histone methylation–related annotations. Table listing the 32 genes retrieved by the Criteria-Based Query illustrated in Figure 7. Columns indicate the gene symbol, functional classification (writer, eraser, or reader/other), the target or associated molecular feature (e.g. modified lysine residue or related function), and whether the gene is annotated (1) or not (0) in the KEGG lysine degradation pathway (hsa00310). Supplementary file10: Table 3. Benchmark gene set used to estimate CRONDEX coverage. List of 119 chromatin-NDD genes compiled from three independent reviews [7, 9, 14], used to estimate CRONDEX coverage. For each gene, the table indicates the review source where it was identified, its presence in CRONDEX, and the source URL. Acknowledgements The authors thank Aitana Díaz for technical support. Author contributions X.d.l.C. and M.A.M.-B. conceived and designed CRONDEX, defined the validation strategies, performed the data analyses, and drafted the manuscript. J.G.F. developed CRONDEX, implemented the analyses, and prepared all figures and tables. B.S.V. extensively tested CRONDEX across multiple usage scenarios. All authors read and approved the final manuscript. Funding X.d.l.C. acknowledges support from the Ministerio de Ciencia, Innovación y Universidades de España through grant PID2022-142753OB-I00, co-funded by the European Regional Development Fund (FEDER). M.A.M.-B. acknowledges support from the Ministerio de Ciencia, Innovación y Universidades de España through grant PID2021-125862NB-I00. J.G. acknowledges a predoctoral fellowship (PREP2022-000566) from the Ministerio de Ciencia, Innovación y Universidades de España. Data availability The datasets supporting the conclusions of this article are included within the article (and its additional files). CRONDEX is publicly accessible via standard web browsers and is currently deployed via an interactive RShiny application: https://jgf-bioinfo.shinyapps.io/CRONDEX/. The complete source code for CRONDEX, including application logic and example data, is openly available on GitHub at: https://github.com/JavierGuerreroF/CRONDEX. This repository enables transparency, reproducibility, and local deployment of the application. No restrictions are applied to the use of CRONDEX by non-academics. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. 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 Marian A. Martínez-Balbás, Email: mmbbmc@ibmb.csic.es Xavier de la Cruz, Email: xavier.delacruz@vhir.org. References Associated Data Supplementary Materials Supplementary file 1: Figure 1. Landing page of the CRONDEX web application.Screenshot of the CRONDEX landing page showing the main interface layout. Two panels in the upper section provide access to the two primary query modes (Gene-Based Query and Criteria-Based Query), while a third panel in the lower section links to additional exploration options, including gene-to-gene comparison and single-gene information. Supplementary file2: Figure 2. (A) Step-by-step execution of a Gene-Based Query in CRONDEX. Three consecutive screenshots illustrating the workflow for performing a Gene-Based Query. Arrows indicate the required user actions at each step: selection of the query gene from the gene menu, confirmation of the selected gene, and activation of the Perform search option to show the results. (B) Output of the Gene-Based Query. For each candidate gene retrieved, the output provides three blocks of information: gene identifiers, scores, and annotations. The scores block reports: the Jaccard similarity (range 0–1), the median-based z-score, and the parent/child decomposition of the Jaccard score. The annotations block reports: the total number of HPO phenotypes for the candidate gene, and its breakdown into parent and child contributions. Supplementary file3: Figure 3. Adjustment of the Jaccard similarity threshold in the Gene-Based Query. Two consecutive screenshots illustrating how the Jaccard index threshold can be modified to control the number of genes retrieved and displayed by the Gene-Based Query. Lowering or increasing the threshold allows users to visualize a broader or more restrictive set of phenotypically similar genes, respectively. Supplementary file4: Figure 4. Step-by-step execution of a Criteria-Based Query in CRONDEX. Series of consecutive screenshots illustrating the workflow for performing a Criteria-Based Query. Arrows indicate the required user actions at each step, including the selection of phenotypic, functional, or pathway-based criteria, and the execution of the query to retrieve the corresponding gene list. Supplementary file5: Figure 5. Comparison between phenotype-based and interaction-based views of BAFopathy-related genes. (A) Screenshot of the Gene-Based Query results obtained using ARID1A as the seed gene, showing only the top-ranked hits retrieved based on phenotypic similarity. This query corresponds to the BAFopathies example discussed in the main text. (B) Protein–protein interaction network retrieved from the STRING database for BAFopathy-associated genes. In contrast to the phenotype-driven relationships highlighted by CRONDEX, this interaction-based view does not explicitly reflect the clinical coherence observed among BAFopathy genes. Supplementary file6: Figure 6. Hierarchical clustering of CRONDEX genes based on phenotypic similarity. Dendrogram of all annotated genes in CRONDEX, constructed from pairwise Jaccard similarity scores computed over their HPO phenotype annotations. The inset on the left indicates the region of the dendrogram shown in the expanded view on the right. In the expanded view, several examples of disease-related gene clusters are highlighted, with those corresponding to cases discussed in the main text shown in colour: Coffin–Siris syndrome (red), Rubinstein–Taybi syndrome 1 and 2 (green), Kabuki syndrome 1 and 2 (blue), Cornelia de Lange syndrome, and Fanconi anemia. Kleefstra syndrome genes form a separate cluster visible in the full dendrogram. Supplementary file7: Figure 7. Jaccard vs. Semantic Similarity Metrics. Each panel shows the conservation of top-ranked hits — top 1, 3, 5, and 10, respectively — between the Jaccard index and two semantic similarity metrics (Lin and Jiang–Conrath) as a function of the Jaccard similarity threshold. In CRONDEX, hits with Jaccard values below the user-defined threshold are not returned. Conservation is expressed as the fraction of hits shared between the Jaccard ranking and each semantic metric. Above a Jaccard threshold of 0.2, conservation of top-ranked hits between Jaccard and both semantic metrics exceeds 95%; above 0.3, agreement is virtually complete. Supplementary file8: Table 1. Chromatin-related neurodevelopmental disorder genes included in CRONDEX. Table containing the 324 genes in the CRONDEX database, with columns indicating gene symbol, NCBI Gene ID, and gene name. Supplementary file9: Table 2. Genes retrieved by the Criteria-Based Query combining intellectual disability and histone methylation–related annotations. Table listing the 32 genes retrieved by the Criteria-Based Query illustrated in Figure 7. Columns indicate the gene symbol, functional classification (writer, eraser, or reader/other), the target or associated molecular feature (e.g. modified lysine residue or related function), and whether the gene is annotated (1) or not (0) in the KEGG lysine degradation pathway (hsa00310). Supplementary file10: Table 3. Benchmark gene set used to estimate CRONDEX coverage. List of 119 chromatin-NDD genes compiled from three independent reviews [7, 9, 14], used to estimate CRONDEX coverage. For each gene, the table indicates the review source where it was identified, its presence in CRONDEX, and the source URL. Data Availability Statement The datasets supporting the conclusions of this article are included within the article (and its additional files). CRONDEX is publicly accessible via standard web browsers and is currently deployed via an interactive RShiny application: https://jgf-bioinfo.shinyapps.io/CRONDEX/. The complete source code for CRONDEX, including application logic and example data, is openly available on GitHub at: https://github.com/JavierGuerreroF/CRONDEX. This repository enables transparency, reproducibility, and local deployment of the application. No restrictions are applied to the use of CRONDEX by non-academics.