DICL: a manually curated database of ion channels and ligands as a useful platform for drug discovery targeting ion channels Authors: Wei Zhao, Xiaoqing Gong, Yanan Tian, Lingling Wang, Chenggong Fu, Xiaojun Yao, Huanxiang Liu Published: 2026 ABSTRACT AbstractIon channels represent a crucial class of drug targets. Currently, with the structural elucidation of more and more ion channels and the rapid development of artificial intelligence in drug discovery, unprecedented opportunities have emerged for designing and discovering drugs that target ion channels. However, both traditional computer-aided drug design and artificial intelligence drug design rely heavily on a large amount of effective data, particularly high-quality data. Yet, data pertaining to relevant ion channels and their ligands are notably scarce. Therefore, constructing a comprehensive and reliable database encompassing ion channels and their ligands is paramount for drug discovery efforts centered around ion channels. To address this gap, we developed database (DICL, Database of Ion Channels and Ligands) including 464,225 data records with the manually curated structure and activity information of ligands from over 6,000 peer-reviewed publications, patents and other databases. As far as we know, it represents the largest and most detailed online database dedicated to ion channels and their ligands. Additionally, the database integrates information regarding the binding sites of ligands, channelopathies, and specific experimental conditions related to biological activities for the first time. Furthermore, it features a user-friendly interface that facilitates querying, browsing, and utilization of this resource. DICL is freely available and can be accessed at https://fca_icdb.mpu.edu.mo/ICHDB/public/. Graphical abstract Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13321-026-01246-x.Keywords: Database, Ion channels, Diseases, Drug design, Online resource Introduction Ion channels are crucial for the regulation of key physiological and pathophysiological processes, which positions them as highly promising therapeutic targets [1–4]. To date, more than 60 channelopathies have been identified [5, 6], and the DisGeNET knowledge platform has cataloged 196 distinct subtypes of ion channels that are related to human diseases [7], including diabetes [8, 9], epilepsy [10, 11], hypertension [12], cancer [13, 14], and chronic pain [15]. As a result, ion channels have emerged as highly attractive targets for drug development. Statistical data reveal that 18% of small molecule drugs listed in the ChEMBL database are modulators of ion channels [16, 17], with approximately 170 drugs available on the market and over 60 undergoing clinical trials [18, 19]. Noteworthy examples include the cAMP-dependent chloride channel corrector lumacaftor for the treatment of cystic fibrosis [20, 21], calcium channel blocker isoptin for the treatment of hypertension and angina [22], and transient receptor potential channel (TRPV1) modulators, such as capsaicin, used to relieve muscle and joint pain as well as neuralgia [23]. Despite the considerable number of existing drugs and ongoing clinical trials, ion channels remain relatively underexploited for therapeutic interventions [24]. Furthermore, some ion channels—including the Voltage-dependent L-type calcium channel subunit alpha-1C (Cav1.2), the human ether-à-go-go-related gene (hERG), and the Sodium channel protein type 5 subunit alpha (Nav1.5)—are associated with cardiotoxicity [25–27]. Consequently, the development of effective and selective modulators targeting ion channels has long been a focal point of research. The limited availability of ion channel structures and the incomplete understanding of their corresponding functions significantly constrain the development of ion channel drugs. The advancement of cryo-electron microscopy (Cryo-EM) [28, 29], coupled with the introduction of AlphaFold Protein Structure Database [30], presents significant opportunities for structure-based drug design. For instance, Jiang et al. performed virtual screening (VS) of an 8-million-compound library with computed fragment maps to discover two PIEZO1 activators [31]. The advent of deep learning provides enough speed to deal with VS of billions of molecules, such as Deep Docking [32], RosettaVS [33], GENTRL [34]. Concurrently, molecular generation utilizing deep learning methodologies has introduced innovative approaches for the design of hit compounds with novel structural frameworks [35, 36]. Notably, both computer-aided drug design (CADD) and artificial intelligence-driven drug design (AIDD) rely heavily on large amounts of precise and effective data, especially AIDD, a data-driven drug design method [37–39]. Xu posited that high-quality data facilitates artificial intelligence models in recognizing significant patterns and producing corresponding predictions [40]. Therefore, the quantity and quality of data are very important factors affecting drug design based on artificial intelligence. Nevertheless, there remains a significant deficiency in comprehensive and reliable data pertaining to ion channels and their modulators. To provide the valuable information for drug design targeting ion channels, researchers have tried to collect various information related to ion channels. For instance, Gallin et al. compiled VKCDB (voltage-gated K+ channel database) consisting of 2,063 unique channel sequences derived from Bacteria (477), Archaea (18) and Eukaryotes (1568), which were used for phylogenetic and comparative analysis [41]. But in this dataset, only sequence information of ion channels was included. In 2023, Nikolay et al. released the Kalium3.0, which consolidates information of polypeptide ligands associated with potassium channels [42]. The cardiotoxicity associated with hERG has garnered significant attention, leading to the publication of extensive data, with hERGCentral being one of the most widely utilized resources [43]. This database, which offered information on the chemical structures and biological activities over 300,000 compounds identified through high throughput screening [43], is currently no longer maintained and its original web interface is currently inaccessible. Furthermore, Wang et al. developed the Voltage-Gated Sodium Channels Database, which provides 6,055 data records of 3,396 compounds from 173 references [44]. Additionally, allosteric modulators represent a significant class of ion channel drug design. The ASD database includes 153 allosteric proteins of ion channels and 15,177 modulators in 2023 [45]. Users can retrieve chemical structure, biological activity (IC50/EC50), various chemical and physical properties from these databases. Some comprehensive databases, such as, BindingDB [46], PubChem database [47], ChEMBL [17], therapeutic target database (TTD) [48], also contain relevant information. Although these works can provide some useful information about some ion channels and their ligands, a multifaceted and integrated ion channel-related dataset with curated activity and structural data is still lacking [49]. Additionally, it has been brought to our attention that there are inaccuracies concerning the stereochemical assignment of specific compounds, as detailed in the supplementary information (13321_2026_1246_MOESM1_ESM.docx), although such inconsistencies are not widespread. Notably, existing databases lack detailed information on binding sites, channelopathies, and the specific experimental conditions under which biological activities were determined. Consequently, there is an urgent need for a professional and comprehensive database to integrate ion channel-related information. In this work, we developed the database of ion channels and ligands (DICL), a comprehensive database that encompasses data resources for ligands (compounds and peptides) targeting ion channels, and functional data for ion channels (Fig. 1). This database catalogues 1,750 ion channel protein subtypes across 27 major types as listed in Table 1, including Voltage-gated potassium channels (VGKCs), Transient receptor potential channels (TRP), and Nicotinic acetylcholine receptors (nAChRs). Furthermore, the DICL database includes 150,208 compound entries and 7,122 peptide records from over 6,000 peer-reviewed articles and patents, along with 306,895 compound records from the hERGCentral database. Each data record comprises over 30 distinct items of information, including biological activity or affinity metrics (IC50, EC50, Ki, Kd), experimental conditions, target information, channelopathy, binding sites of ligands, etc. These data have been manually validated and are supported by corresponding references. In summary, this database offers accurate and comprehensive information on ion channels and ligands, serving as a valuable resource for ion channel drug design. As far as we know, it represents the largest and most detailed online database dedicated to ion channels and their ligands, serving as a foundational tool for innovative drug development strategies aimed at ion channel modulation. Overview of the database of ion channels and ligands (DICL): construction process, data types, and functions Table 1All 27 types of involved ion channels and related acronymsFull nameAbbreviationVoltage-gated potassium channelsVGKCs/KvTwo-pore domain potassium channelsK2PCalcium-activated potassium channelsKCaInwardly rectifying potassium channelsKirVoltage-dependent calcium channelsVGCCs/Cav/VDCCsRyanodine receptorRyRVoltage-dependent sodium channelsVGSCs/Nav/VDSCsTransient receptor potential channelsTRPCyclic nucleotide-gated channelCNGTwo-pore channelTPCCation channel sperm-associated proteinsCatSperVoltage-gated proton channelVGHC/Hv15-hydroxytryptamine type-3 receptor5-HT3RAcid-Sensing Ion ChannelsASICsEpithelial sodium channelENaCγ-Aminobutyric acid type A receptorsγ-GABA-ARsNicotinic acetylcholine receptorsnAChRsIonotropic glutamate receptorsi-GluRsGlycine receptorsGlyRsInositol 1,4,5-trisphosphate receptorsIP3RsZinc Activated ChannelZACP2X receptorsP2XChloride channels Calcium activated chloride channelCaCC Volume regulated chloride channelsVRAC ClC familyClC cAMP-regulated epithelial cell membrane chloride (Cl-) channelCFTRGap junction channel ConnexinsGJ channel PannexinsOrai channelsOraiAquaporinsAQPsPiezo channelsPiezo All 27 types of involved ion channels and related acronyms Materials and methods Data collection The data for DICL were sourced from both literature and publicly accessible databases, as illustrated in Fig. 1. A comprehensive literature search was performed using PubMed, Google Patents, and Web of Science databases. Keywords such as ‘compound’, ‘peptide’, ‘modulator’, ‘molecule’, ‘agonist’, ‘antagonist’, ‘inhibitor’, and ‘blocker’, in conjunction with ‘Voltage-dependent calcium channel’ and other specific ion channel names, were employed to query these resources. The search in Google Patents followed a similar keyword-based strategy to identify relevant patent documents. The scope of the search included both article titles and abstracts. Candidate compounds and peptides retrieved through search were initially identified based on functional and structural information in the literature. To extract ion channel sequences and UniProt accession numbers, targeted keyword searches were conducted using terms such as “Voltage-gated potassium channels”, “Aquaporins”, “Voltage-dependent calcium channels”, “γ-Aminobutyric acid type A receptors”, “ionotropic glutamate receptors”, “Chloride channels”, et al. across UniProt [50], Protein Data Bank (PDB) [51] and GenBank [52]. Each UniProt accession number corresponds to a unique ion channel sequence from a specific organism and all retrieved data were individually checked to ensure accuracy. Subsequently, molecular information from the literature was collated and curated using DrugFlow [53] and StoneMind Collector [54]. This included data such as IC50/EC50/Ki/Kd values, compound structures, target receptors, binding sites, and experimental conditions. The data in DICL were collected not only from peer-reviewed literature but also from patents to ensure comprehensive information. Additionally, we compiled the information of ion channels and ligands from several reputable databases (Fig. 2), including BindingDB [46], ChEMBL [17], Protein Data Bank [55], UniProt, AlphaFold Protein Structure Database [30], VGSC-DB [44], PubChem [56], and TTD [48]. In accordance with our data inclusion policy, every compound must be traceable to a primary empirical source. Therefore, entries sourced from external databases (e.g., ChEMBL, BindingDB) that are not grounded in verifiable literature are omitted from our collection. Additionally, over 300,000 compounds from hERGCentral [43] were also integrated into DICL. As the original hERGCentral database is no longer maintained and is currently inaccessible, the relevant data was sourced from a third-party repository [57]. Data sources integrated in DICL. The database integrates data from ten public resources: BindingDB, ChEMBL, Protein Data Bank (PDB), UniProt, AlphaFold Protein Structure Database, VGSC-DB, PubChem, ASD, Kalium 3.0, and TTD. (Detailed information is provided in 13321_2026_1246_MOESM6_ESM.xlsx of the Supplementary Information.) Data curation and processing In the context of compounds’ information, each record identified by a unique DICL ID comprises 35 distinct items of information. The two-dimensional structures of the compounds were generated utilizing SMILES Drawer [58]. Biological activity or binding affinity were quantitatively assessed using established biochemical and biophysical metrics, including IC50, EC50, Ki and Kd, with nanomolar (nM) as the standard unit of measurement. Binding sites were first categorized as either agonists or antagonists. Given the distinct pharmacological implications, the antagonist sites were further classified into "Inhibitor" and "Blocker" subtypes based on their precise location and mode of action. Additionally, "Allosteric" sites were separately highlighted due to their significant and distinct role in modulating ion channel function. The determination of binding site residues required corroboration by at least two independent lines of evidence. These included mutagenesis data demonstrating significant functional impact, high-resolution structural data (from X-ray crystallography or cryo-EM), or computational validation via molecular dynamics simulations or docking studies. Data on biological activity, binding affinity, targets, binding sites, and organisms were compiled from the literature. The process of data integration and standardization is shown in Fig. 3. The compound information was initially obtained from public databases, including BindingDB, ChEMBL, and TTD. For each database, we downloaded all available molecular representations—such as SMILES, SDF, and InChI—to ensure data completeness. Canonical SMILES strings served uniformly as the primary input for our standardization pipeline, while the corresponding 2D structures were employed for initial validation. All SMILES strings were subsequently verified manually using ChemDraw 19.0 (PerkinElmer Informatics, Inc.) [59]. A rigorous standardization protocol was applied, including neutralization, salt and solvent stripping, and normalization of specific functional groups. Using RDKit [60], we generated corresponding InChI, InChIKey, and 2D SDF files from the standardized SMILES. Functional classifications (e.g., antagonist, inhibitor, blocker, agonist, activator) were assigned based on the roles explicitly stated in the originating primary literature or patents. While data from curated sources like ChEMBL were considered, they did not supersede direct annotations from primary sources. Crucially, our inclusion criteria required every compound entry to be supported by a citation to a primary empirical study. Consequently, any database entries lacking such verifiable references were excluded from the final collection. The classification of each peptide was further verified by cross-referencing its sequence or identifier with the UniProt database (using UniProt Accession Numbers where available). Article DOI and PMID information were gathered from PubMed. Furthermore, eight critical properties associated with drug-likeness were predicted and incorporated into DICL. Properties including the number of heavy atoms counts (HAC), hydrogen bond acceptors (HBA), rotatable bonds (RB), topological polar surface area (TPSA), complexity, ring count and hydrogen bond donors (HBD) were calculated utilizing RDKit, whereas logP values were determined using ADMETlab 3.0 software [61]. Consequently, two separate tables were used to document the information for compounds and peptides, respectively, due to their significant differences in features. Detailed information on small molecules is available in CSV files. Data standardization and integration workflow. The process involved standardizing, harmonizing, and manually verifying the data collected from diverse sources, including original databases and published literature The DICL_Pro ID serves as a unique identifier within the database that is associated with ion channels. The UniProt accession number, name, synonyms, and target descriptions are sourced from UniProt, with certain modifications applied to the original descriptions. To optimize the functional annotations, the following modifications were undertaken: (1) removal of internal cross-reference identifiers (e.g., [EC:1.1.1.1], [GO:0005634]) for readability; (2) condensation of redundant or overly technical phrasing into core functional statements for clarity; and (3) supplementation with literature-curated, clinically relevant protein-disease relationships to enhance biological context. The target structures are sourced from the PDB database or AlphaFold Protein Structure Database [30], while the corresponding binding site data are collected from peer-reviewed literature. All files in DICL are available for free download. The DICL accommodates the complexity of compound-target interactions by systematically organizing the data. Since a single compound may interact with multiple targets under varied experimental conditions, each compound’s data is categorized into distinct, method-specific records based on the experimental methodology, target, organism, and other relevant parameters. Users can access the details page of each record to find more information. Additionally, the activity value serves as the foundation for both qualitative and quantitative analyses. It is important to note that research findings from various groups may sometimes conflict, leading to discrepancies that can pose challenges for users. Therefore, we meticulously verified and amended these data entries based on the reliability and recency of the supporting research. Our evaluation criteria included the completeness of methodological descriptions, the use of orthogonal validation assays, the presence of appropriate controls, and the rigor of statistical analyses. In essence, we refined and standardized the data according to the factual information to improve user engagement. Development of the DICL Web platform The website was built using the CodeIgniter 4.6 framework, with CSS and JavaScript for front-end development, and PHP 8.1.2 and MySQL for back-end functionality, all hosted on Apache2.The three-dimensional structures of ion channels were visualized using 3Dmol.js for online protein visualization. The two-dimensional structures of compounds were displayed using the SMILES Drawer. All technical operations of DICL adhered to the established standards. Results and discussion The comparison between DICL and other related databases To highlight the novel information in our database, we conducted a comparison between the DICL database and other related databases (as shown in Table 2). Firstly, ChEMBL encompasses a broad range of small molecule activities assessed in phenotypic and biological assays, totally including approximately 111,890 records related to ion channels [17]. BindingDB primarily focuses on experimentally measured protein-small molecule binding affinities, comprising approximately 148,570 records related to ion channels sourced from peer-reviewed publications and patents [46]. Kalium 3.0 integrates data on polypeptide ligands associated with potassium channels, encompassing a total of 4,108 records [42]. Unlike the databases mentioned above, the hERGCentral database provides information on the biological activities of 318,941 compounds related to hERG, identified through high-throughput screening [43]. However, it is currently inaccessible. To construct a more comprehensive and reliable database encompassing ion channels and their ligands, here, we integrated these databases and other scattered data from peer-reviewed publications and patents to construct the DICL database. The DICL database comprises 150,208 compound entries and 7,122 peptide records sourced from more than 6,000 peer-reviewed articles and patents, in addition to 306,895 activity data (hERGCentral database) points obtained from the third part source [57]. DICL is an integrated resource encompassing ion channels and their associated binding ligands (including peptides and small molecules), covering not only potassium channels but also Voltage-dependent calcium channels (VGCCs), Voltage-gated sodium channels (VGSCs), and Chloride channels. Due to this broad scope, DICL currently represents the most comprehensive publicly available database dedicated to ion channels and their ligands. Secondly, the DICL database incorporates 1,750 subtypes of 27 different types of ion channel proteins and their related information from 165 species. Additionally, beyond compiling bioactivity, affinity, and molecular property data, DICL also incorporates key functional annotations—including ligand binding sites, associated channelopathies, and detailed experimental conditions. Therefore, DICL represents the most comprehensive and detailed online database of ion channels and their ligands. Table 2The comparison of DICL and other related databasesDatabase nameTotal data recordsData typesSourcesIon channel numberhERG-APDbase [62]1NTPercentage inhibition rates at 10 µMPRPNThERGCentral [43]2318,941 entriesPercentage inhibition rates at (1 µM and 10 µM), CSLab-specific screening data by themself1 subtypeBindingDB [46]3148,570 entriesIC50, EC50, Kd, Ki, Function, CS, Chemical and physical propertiesPublic Databases, PRP; Patents497 subtypesHERB [63]4348 entries IC50, diseasePublic Databases,190 subtypesVGSC-DB [44]56,055 entriesBioactivity; Function, CS, binding sites, Chemical and physical propertiesPRP, Public Databases21 subtypesKalium3.0 [42]64,108 entriesSequence, Experimental methods, IC50, EC50, CC50, KiPRP, Public Databases182 subtypesTTD [48]735,341 entriesChemical and physical propertiesPRP, Public Databases242 subtypesASD2023 [45]815,177 entriesCSPublic Databases, PRP182 subtypesChEMBL [17]9111,890 entriesCS, IC50, EC50, Kd, Ki, Chemical and physical properties, FunctionPublic Databases, PRP470 subtypesDICL464,225 entries (7,122 peptides; 457,103 compounds)IC50, EC50, Kd, Ki; Function, specific experimental conditions and methods; binding sites of ligands; channelopathies, Chemical and physical properties, CS,PRP; Public Databases; Patents1750 subtypesThis table compares the availability of a standardized set of core data types across databases. The list is not exhaustive but represents key data types of interest in this domain. The presence or absence of a specific type is not an indicator of a database's overall value, as each resource has its own unique scope and focus. Total Data Records: the number of protein–ligand pairs. Ion Channel Number: The number of subtypes includes distinct entries for orthologous subtypes from different species, which have been provided in the Supplementary Information [13321_2026_1246_MOESM5_ESM.xlsx]. NT: Not Mention; PRP: Peer-reviewed Publications. CS: Compound Structure. 1: hERG-APDbase (cannot be accessed at present): http://www.grt.kyushuu.ac.jp/hergapdbase/; 2: hERGCentral (cannot be accessed at present): http://www.hergcentral.org; 3: Binding DB: https://www.bindingdb.org/rwd/bind/index.jsp; 4: HERB: http://herb.ac.cn/; 5: VGSC-DB: http://cadd.zju.edu.cn/vgsc/; 6: Kalium 3.0: http://kaliumdb.org/; 7: https://idrblab.org/ttd/; 8: https://mdl.shsmu.edu.cn/ASD/module/mainpage/mainpage.jsp; 9: https://www.ebi.ac.uk/chembl/. All data displayed in Table updated to November 2025 The comparison of DICL and other related databases This table compares the availability of a standardized set of core data types across databases. The list is not exhaustive but represents key data types of interest in this domain. The presence or absence of a specific type is not an indicator of a database's overall value, as each resource has its own unique scope and focus. Total Data Records: the number of protein–ligand pairs. Ion Channel Number: The number of subtypes includes distinct entries for orthologous subtypes from different species, which have been provided in the Supplementary Information [13321_2026_1246_MOESM5_ESM.xlsx]. NT: Not Mention; PRP: Peer-reviewed Publications. CS: Compound Structure. 1: hERG-APDbase (cannot be accessed at present): http://www.grt.kyushuu.ac.jp/hergapdbase/; 2: hERGCentral (cannot be accessed at present): http://www.hergcentral.org; 3: Binding DB: https://www.bindingdb.org/rwd/bind/index.jsp; 4: HERB: http://herb.ac.cn/; 5: VGSC-DB: http://cadd.zju.edu.cn/vgsc/; 6: Kalium 3.0: http://kaliumdb.org/; 7: https://idrblab.org/ttd/; 8: https://mdl.shsmu.edu.cn/ASD/module/mainpage/mainpage.jsp; 9: https://www.ebi.ac.uk/chembl/. All data displayed in Table updated to November 2025 The statistics and analysis of data in DICL To present detailed information from DICL, we conducted statistical analyses of the data. It should be noted that the biological activity values from hERGCentral are reported exclusively as percentage inhibition rates, rather than standardized quantitative measures (e.g., IC50, EC50, or Ki). Due to this unit incompatibility with the other datasets, direct comparative analysis was not methodologically sound. Consequently, these data were excluded from further analysis. Ion channel-based statistics and analysis The DICL includes a curated collection of approximately 1,750 ion channel protein subtypes from 165 species (Supplementary Information [13321_2026_1246_MOESM4_ESM.xlsx]). These subtypes are systematically classified into 27 major ion channel types based on the authoritative framework provided by the IUPHAR/BPS Guide to Pharmacology [64]. As depicted in Fig. 4, over 200 unique sequences of VGKCs and Aquaporins (AQPs) are included. Furthermore, more than 100 unique sequences have been collected for the following channel types: VGCC, γ-Aminobutyric acid type A receptors (γ-GABA-ARs), Ionotropic glutamate receptors (i-GluRs), Chloride channels, and Gap junction channel (GJ channel). Additionally, the database contains more than 50 unique sequences for several key ion channel families, including VGSCs, Inwardly rectifying potassium channels (Kir), Calcium-activated potassium channels (KCa), TRP, and nAChRs. Moreover, Fig. 4 illustrates that the types of ion channels found in humans, rats, and mice collectively represent nearly 50% of the total identified sequences. It is noteworthy that the types of VGKCs, VDCCs, γ-GABA-ARs, i-GluRs, KCa, Kir, and TRP identified are relatively abundant, with VGKCs being particularly prominent, encompassing nearly 50 subtypes. Statistics of ion channels in DICL. The number of 27 types of ion channels across humans, rats, and mice as a percentage of the total number of that ion channel type Compound-based statistics and analysis Except the data from hERGCentral database, this current version of DICL comprises 150,208 data entries related to 72,420 compounds gathered from 5,960 peer-reviewed publications or patents. Figure 5A illustrates the distribution of binding affinity or biological activity for compounds in the database across ion channels. There are over 10,000 records for VGKCs, VGSCs, nAChRs, i-GluRs, and TRP. Notably, the records for VGKCs exceed 20,000, accounting for more than 20% of the total data. This prevalence may be attributed to the increased focus on the cardiotoxicity issues associated with hERG. In contrast, there are only a few dozen records for Cyclic nucleotide-gated channel (CNG), Voltage-gated proton channel (VGHC), Two-pore channel (TPC), Piezo channels, and Epithelial sodium channel (ENaC). A statistically significant divergence in modulator efficacy was observed across ion channel families. Over 70% of compounds exhibited antagonistic effects on VGKCs, VGSCs, i-GluRs, and TRP channels. In contrast, more than 50% of compounds acted as agonists on γ-GABA-ARs and nAChRs. The observed target-class bias likely stems from a combination of factors. This includes practical aspects such as the composition of screening libraries and historical pharmaceutical focus (e.g., the high number of TRP channel antagonists related to analgesic development). At the same time, the bias may also reflect the inherent physiological and pharmacological characteristics of different ion channel families, such as variations in the accessibility or druggability of key regulatory sites. The uneven distribution of reported compounds across different subtypes of ion channels indicates that the function of some ion channels needs to be explored and underscores the urgent need for innovative drug development targeting the unexploited targets. In Fig. 5B, we can clearly observe that the data in database is more concentrated on high and moderate bioactivity (the value of IC50 < 100 nM and IC50 from 100 nM to 10,000 nM) for ion channels. One contributing factor is that high-throughput screening (HTS) studies frequently generate large compound libraries, however, the resulting publications often omit inactive or weakly active compounds (the value of IC50 > 100,000 nM and from 10,000 nM to 100,000 nM) from their reported datasets. It is noteworthy that extensively studied targets—including VGKCs, VGSCs, VGCCs, i-GluRs, and TRP channels, each with over 5,000 activity records—constitute a major focus of pharmacological research and are therefore particularly well-represented in the database (the detailed data and 2D figure shown in the 13321_2026_1246_MOESM2_ESM.xlsx of Supplementary Information). This pattern reflects both sustained research interest and comprehensive functional characterization. The Distribution of Compounds Targeting on Ion Channels. A The number of compounds targeting each ion channel type in the database. (PAM: Positive Allosteric Modulator, NAM: Negative Allosteric Modulator, Others: Corrector and Potentiator) B Distribution of compounds with different IC50 values, categorized into four ranges: < 100 nM, 100 nM–10000 nM, 10,000–100000 nM, and > 100,000 nM. Circle size corresponds to the number of compounds within each activity range; colors represent the logarithm of that count Additionally, the molecular properties of compounds in DICL with bioactivity values below 10 μM were statistically analyzed. This analysis was based on Lipinski’s Rule of Five [65] (Fig. 6A–D) and rotatable bond count (Fig. 6E). Nearly 78% of the active compounds exhibit logP values between 1 and 5, while over 80% of the compounds have molecular weights ranging from 200 to 500 Dalton. For 95% of the compounds, the number of HBA is fewer than 10, while 99% of the compounds have fewer than 5 HBD, and 93% of the compounds have fewer than 10 RB. Overall, these data provide valuable information for discovery and design of compounds targeting ion channels. Property distributions of active compounds: A logP, B molecular weight (MW), C number of hydrogen bond acceptors (HBA), D number of hydrogen bond donors (HBD), and E number of rotatable bonds (RB) Peptide-based statistics and analysis DICL included 7,122 data records for peptide ligands related to ion channels from peer-reviewed publications, encompassing 2,195 peptides from 703 references. Figure 7A illustrates the number of peptides interacting with several types of ion channels. There are nearly 3,500 records for VGKCs, 1,200 records for VGSCs, and 1,000 records for nAChRs, collectively accounting for 84% of the total dataset. Inhibitors, blockers, and antagonists represent nearly 93% of dataset. This finding also elucidates, to some extent, the propensity of these known sites to engender antagonists. In Fig. 7B, the percentages of peptides with high and moderate bioactivity for VGKCs (the value of IC50 < 100 nM and the value of IC50 ranging from 100 nM to 10,000 nM) are approximately 98.85% and 1.15%, respectively. For VGSCs, the percentages of peptides with high and moderate bioactivity are nearly 99.14% and 0.84%, respectively. For nAChRs, the percentages of peptides with high and moderate bioactivity are nearly 94.22% and 5.78%, respectively. Detailed data are provided in the 13321_2026_1246_MOESM3_ESM.xlsx of Supplementary Information. These data will provide valuable insights for the discovery and design of active peptides targeting ion channels. The distribution of peptides targeting on ion channels. A The number of peptides targeting each ion channel type in the database. B Distribution of their IC50 in different range, categorized into four ranges: < 100 nM, 100 nM–10000 nM, 10,000–100000 nM, and > 100,000 nM. The size of the circle corresponds to the number of peptides exhibiting the respective biological activity. The different color of the circle corresponds to log (the number of peptides exhibiting the respective biological activity) Data querying, download, structural display and user manual Querying DICL offers users a search (Fig. 8A) function for retrieving information on compounds and peptides. Users can select the search category from the "All Categories" drop-down menu for quick searches (Fig. 8B). For compound retrieval, available search options include "SMILES", "Organism", "IUPAC Name", "Ion Channel Name", "References", "PubChem CID", "InChIKey", "Function", "Value", and "DICL ID". The "DICL ID" serves as a unique identifier within the database for ligands targeting ion channels. The "Value" field includes numerical data such as Ki, EC50, IC50, and Kd. The "SMILES" search pattern is text-based, and users may input an isomeric SMILES string either obtained from PubChem or derived from the molecular structure. When a SMILES query is submitted, the system invokes RDKit to parse the input string into a molecular graph, determine aromaticity and bond orders, and generate a canonical isomeric SMILES representation. This normalized form is then directly compared to the canonical SMILES strings stored in the database, guaranteeing accurate structure-based retrieval irrespective of variations in the original input. For peptide retrieval, the available options differ and include "Sequence", "Organism", "Ion Channel Name", "References", "Function", "Value", and "DICL ID". Currently, sequence searching relies on SQL substring matching, which retrieves entries containing the input fragment anywhere in the text fields. Specifically, this approach performs an exact match query on the sequence column using the SQL equality operator ("=") with search keyword, returning only records that exactly match the search keyword. Additionally, users can select one or more keywords from options such as ion channel name, organism, function, Ki, EC50, IC50 and Kd to perform an advanced search (Fig. 8C). Upon completion of the search, users are redirected to a results page that displays the DICL ID, Ion Channel, Compound function, Species, Compound Activity, and References (Fig. 8D). The data querying function of DICL. A Schematic of the search entry point. B Quickly search function for the database. C Advanced search function for the database. D The search results page. E Download function for search result from the database Ligand data display Users can readily access essential compound information and click on the ' DICL ID' to navigate to the corresponding detail page, which provides comprehensive data and properties for the selected compound. The detailed page of a compound contains three sections: Basic Information (Fig. 9A), Experimental Data (Fig. 9B) and Computed Properties (Fig. 9C). The Basic Information section presents detailed data about a compound, including the SMILES, IUPAC name, target, InChI, and more. The Experimental Data section includes PubChem CID, MW, IC50/EC50/Ki/Kd values, organism, binding site, and article PMID or DOI. Additionally, we utilized ADMETlab3.0 and RDKit to calculate the properties of compounds, which are displayed in the Computed Properties section. The detailed page for peptides only contains Basic Information and Experimental Data, in contrast to the detailed page of compounds. Ligand Information Page in DICL Database. A Basic compound information. B Experimental pharmacological data (IC50, EC50, Ki, Kd) assay details. C Computed molecular properties (logP, TPSA, HBD/HBA, etc.) Manual and download The ‘Manual’ page provides a step-by-step tutorial for users on how to manipulate, query, and browse the database (Fig. 10A). It is essential to acknowledge that the collections in DICL do not encompass all data related to ion channels. As shown in Fig. 8E, users can select one or more query results to download in batch in CSV file format. Furthermore, user could directly download the ligand information for a specific target or all targets by clicking the ‘Download Link’ (Fig. 10B). The manual and download function of DICL. A Manual overview of DICL, which includes the following five sections: Overall Layout, Database Construction, Data Types, Browsing, and Search & Download. B Download function of all targets data or specific target data Protein data display As shown in Fig. 11A, 27 types of ion channels were contained in DICL, including VGKCs, VGCCs, VGSCs, γ-GABA-ARs, nAChRs, i-GluRs, Acid-Sensing Ion Channels (ASICs), 5-hydroxytryptamine type-3 receptor (5-HT3R), ENaC, TPC, VGHC, Ryanodine receptor (RyR), CNG, Chloride channels, Orai channels, Piezo, GJ channel, AQPs, Glycine receptors (GlyRs), KCa, K2P, Kir, Inositol 1,4,5-trisphosphate receptors (IP3Rs), TRP, and Cation channel sperm-associated proteins. These channels are categorized into three groups: voltage-gated (light yellow), ligand-gated (light blue), and other ion channels (light red). Information Page for Ion Channels in DICL. A The table lists 27 types of ion channels (e.g., VGKCs, VDCCs, VGSCs, γ-GABA-ARs, nAChRs, i-GluRs, ASICs, 5-HT3R). B The Ion Channel Details page is divided into three sections: Summary (structural composition, subunit function, and regulatory mechanisms), Binding Sites (inhibitor, activator, blocker, and allosteric sites), and Target List (subtypes under each ion channel category). C The Detailed Information page provides comprehensive data on each ion channel, including its function, sequence, organism, PDB location, related diseases, and other relevant information The summarized information page of the protein (Fig. 11B) serves as the primary repository for protein information in DICL. This page is organized into three distinct sections: Summary, Binding Sites, and Target List. The three-dimensional structure of the protein is displayed in an interactive window, enabling users to rotate and zoom the model. The Binding Sites section is categorized into four classifications according to the function of ligands, namely inhibitor, agonist, blocker, and allosteric modulator. DICL ID, Protein Name, Gene name, Organism, and the UniProt accession number are provided in the Target List. Users can click on the ICDB_Pro ID to access the detailed information page (shown in Fig. 11C). This page also offers the detailed information regarding the source, sequence, location, function, related disease, the PDB ID, and organism, thereby facilitating users to understand the involved ion channel quickly. Conclusions In this study, we developed the DICL, a comprehensive database that integrates data resources for ion channels and their ligands (compounds and peptides). DICL encompasses more than 1,750 subtypes ion channel proteins, categorized into 27 different types of ion channels. Meanwhile, it contains 464,225 data records, including 150,208 data records of compounds sourced from 5,960 peer-reviewed publications and patents, 306,895 data records of compounds sourced from the hERGCentral database, as well as 7,122 data records of peptides from 708 peer-reviewed publications and patents. To the best of our knowledge, DICL is the most comprehensive and detailed online database dedicated to ion channels and their corresponding ligands. Serving as a foundational resource for innovative drug development, the database provides a user-friendly interface and extensive information to support querying, browsing, and in-depth analysis of ion channels, their ligand interactions, and associated functions. The constructed platform provides a valuable contribution to the development of effective drug design strategies targeting ion channels and the drug discovery related to ion channels. DICL serves as a valuable platform offering extensive information on ion channels and their ligands, providing a solid foundation for the development of AI-based drug design methods specifically tailored to ion channels and advancing target-based drug discovery. The accurate and comprehensive data on ion channels and their ligands can be leveraged to develop machine learning and AI-based models for identifying binding sites, discovering agonists and antagonists, and elucidating selective mechanisms of ion modulation as well as toxicology prediction models since cardiotoxicity is closely related with some ion channels [66–68]. DICL contains 18,823, 2,481, and 1,045 raw data entries for hERG, Nav1.5, and Cav1.2, respectively. These data can provide strong support for the development and evaluation of cardiotoxicity prediction models. Moreover, it also serves as a reference for the design of experimental methodologies. However, several issues remain to be addressed. Firstly, some compounds have been excluded from the database due to the unavailability of precise biological activity values. Secondly, while we strive to minimize potential errors throughout this process, we acknowledge that some mistakes due to oversight may still occur. Additionally, although the current version of the platform provides full access to all core database content and supports essential search and download functions, several aspects can be further optimized to better serve advanced research needs and enhance usability. Future improvements will focus on the following areas: (1) System Performance and Scalability. While basic queries perform adequately, complex searches or bulk data operations can experience delays. We are optimizing backend query processing and implementing advanced caching mechanisms to significantly accelerate data retrieval and support larger-scale user interactions. (2) User Workflow and Customization Features. The site currently does not support user-specific features such as saving search conditions, storing personal molecule sets, or annotating records. We are developing user management capabilities that will allow researchers to track their query history, create custom compound lists, and share results with collaborators. (3) Enhanced Visualization and Interactive Tools. Although the platform includes a structure viewer, its functionality remains limited to basic visualization. (4) Development of Chemical Search Functionality. Implementing a true chemical search system—capable of substructure or similarity queries based on molecular fingerprints—would represent a highly valuable addition. However, robust deployment of such functionality requires substantial computational backend development and integration. Establishing a comprehensive chemical search system, including both substructure and similarity capabilities, is therefore an important objective for future versions of the database. In summary, we are working to integrate more advanced and interactive tools for exploring protein–ligand interactions, analyzing binding poses, and visualizing structure–activity relationships. In the future, we plan to maintain and expand DICL by periodically incorporating additional ion channels and their corresponding ligands, sourced from newly published research. We anticipate that DICL will serve as a significant resource and a robust tool for the study of ion channels, as well as for the innovative design of ion channel modulators. We also encourage users to provide feedback to help us improve the platform. Supplementary Information Supplementary material 1 Supplementary material 2 Supplementary material 3 Supplementary material 4 Supplementary material 5 Supplementary material 6 Acknowledgements Thanks to the Information Technology Department of Macao Polytechnic University for supporting this website network. Author contributions Wei Zhao: Conceptualization, methodology, collected the data, constructed the website, analyzed the data, writing—original draft and visualization; Lingling Wang: methodology, verified the data, and visualization; Xiaoqing Gong: Collected the data, writing—original draft; Yanan Tian: Verified the data, writing—original draft; Chenggong Fu: collected the Data, Xiaojun Yao: Ideological guidance, conceptualization; Huanxiang Liu: Ideological guidance, project administration, writing—review and editing, and funding acquisition. Moreover, all authors read and approved the manuscript. Funding This work is supported by Macao Science and Technology Development Fund (No. 0043/2023/AFJ) and Macao Polytechnic University (No. RP/FCA-02/2023). Data availability The data that support the findings of this study are openly available in [https://fcaicdb.mpu.edu.mo/ICHDB/public/]. All considered accession numbers are included in the Supplementary Information. Declarations 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 Xiaojun Yao, Email: xjyao@mpu.edu.mo. Huanxiang Liu, Email: hxliu@mpu.edu.mo. References Associated Data Data Availability Statement The data that support the findings of this study are openly available in [https://fcaicdb.mpu.edu.mo/ICHDB/public/]. All considered accession numbers are included in the Supplementary Information.