Sister chromatid cohesion depends on ESCO2-mediated acetylation of the cohesin subunit SMC3, yet how this modification is established and maintained across the genome remains poorly understood. Here, we examined the chromatin-binding dynamics of ESCO2 and its enzymatic activity using quantitative ChIP-seq of human cells synchronously progressing through S phase. ESCO2 bound broadly along chromosomes without forming discrete peaks and preferentially localized to transcriptionally inactive regions, but was not confined to H3K9me3- or H3K27me3-enriched heterochromatin. Consistent with this binding pattern, ESCO2-dependent SMC3 acetylation showed a similarly broad distribution in these regions. In contrast, ESCO1, a paralog of ESCO2, primarily mediated SMC3 acetylation at cohesin peak sites. Notably, most of the broadly distributed, ESCO2-dependent SMC3 acetylation generated during S phase was progressively lost from chromatin after replication, whereas the fraction that persisted into late G2 accumulated in domains enriched for the repressive H3K27me3 modification. Our work identifies selective post-replicative retention of acetylated cohesin as a previously unappreciated step in shaping the genome-wide architecture of sister chromatid cohesion.

Gene transcription by RNA Polymerase II (Pol II) is highly vulnerable to transcription-blocking DNA lesions (TBLs), which, if unresolved, induce transcription stress leading to cellular decline and aging. TBLs are removed by transcription-coupled nucleotide excision repair (TC-NER). TC-NER is initiated when Cockayne syndrome group B protein recognizes lesion-stalled Pol II and promotes TC-NER complex assembly. This complex recruits TFIIH for lesion verification, followed by incision and repair to enable transcription restart. Many TBLs stall Pol II within its active site, shielding the lesion from repair and requiring Pol II processing to expose it. However, whether this occurs through Pol II backtracking or release from the DNA, and the regulatory mechanisms involved, remain poorly defined. Here, we identified the Integrator complex as a key factor in lesion-stalled Pol II processing. Interaction proteomics revealed Integrator recruitment to late-stage TC-NER complexes. Integrator's endonuclease core promotes efficient transcription recovery after TBL induction, without affecting lesion removal by TC-NER. Live-cell imaging and chromatin retention assays demonstrated that Integrator facilitates Pol II release from DNA following TC-NER. Collectively, these findings support a two-step mechanism for Pol II clearance during TC-NER, wherein TFIIH-mediated backtracking exposes the lesion for repair, followed by Integrator-mediated Pol II release to restore transcription.

tRNA fragments (tRFs) are natural, small non-coding RNAs produced by endogenous ribonucleolytic cleavage of tRNAs. tRFs play a role in many biological pathways, including regulation of cell proliferation. Specific tRNA modifications affect tRF biogenesis. Queuosine (Q) modification occurs at the wobble anticodon nucleotide of 4 mammalian tRNAs and protects these tRNAs from fragmentation. Here, we investigated cell-type and tRNA Q-modification dependencies on mammalian cell proliferation. We found that Q-modified tRNA can either increase, decrease, or make no difference in proliferation. Proliferation decreased in MCF7, a breast cancer cell line, when tRNAs were Q-modified and dependent on the modification enzyme complex, QTRT1/2. MCF7 proliferation increases upon the transfection of Q-modification-dependent tRFHis or decreases upon knockdown of endogenous tRFHis. Consistent with Q-dependent proliferation reduction, polysome profiling showed decreased translation efficiency of ribosomal proteins. Finally, protein pull-down of MCF7 lysates using tRFHis identified Musashi RNA-binding protein 2 (Msi2) as a tRFHis-binding protein. Msi2 knockdown reduced the Q-modification-dependent or tRFHis dependent cell proliferation effect. Our results illustrate a highly cell-type-dependent proliferation effect of tRNA Q-modification, suggesting multi-faceted mechanisms in studying this ubiquitous tRNA modification.

Population-scale metabolomics studies have linked metabolic measures to diseases, quantitative traits, and genetic variation, but the results are scattered across assay platforms, phenotype definitions, and analytical formats. MetabDB integrates 16 public studies spanning nuclear magnetic resonance, mass spectrometry, and lipidomics. The current release contains 2663 metabolic measures, 3524 diseases and outcomes, 3382 other genome-wide association traits, and 2025 molecular quantitative trait locus (QTL) targets. MetabDB links genomic variants, genes and proteins, metabolic measures, and diseases or traits through variant/QTL associations, fine-mapping, rare-variant or burden analyses, colocalization, Mendelian randomization, and population associations. Metabolic measures areretained in their source-reported forms, including compounds, lipid species, lipoprotein measures, ratios, percentages, and derived traits. Gene- and protein-level evidence can be examined alongside genomic signals and related disease or trait evidence for the same metabolic measure. Source information and reported analytical details are retained, including the source study, original variable or phenotype definition, assay, unit, exposure scale, analytical model, and statistics where available. Search, browse, and colocalization pages, together with gene evidence breadth and disease-association profiles, provide access to these records. MetabDB is freely available at http://www.mulinlab.org/metabdb/.

Long non-coding RNAs (lncRNAs) have diverse biological functions, but experimentally validated evidence remains scattered across the literature and is often context dependent. EVLncRNAs 4.0 (https://www.sdklab-biophysics-dzu.net/EVLncRNAs4/) updates our manually curated database of functional lncRNAs supported by low-throughput experiments. For this release, 16 581 publications indexed between 15 May 2023 and 15 May 2026 were retrieved and manually screened. Compared with EVLncRNAs 3.0, EVLncRNAs 4.0 contains 7808 functional lncRNAs (+26%) from 199 species (+23%), 2313 diseases (+34%), 28 757 curated functional records (+21%), and 18 380 molecular interactions (+25%). EVLncRNAs 4.0 also introduces two AI-assisted functions built on the curated evidence base. AI Search uses an enhanced retrieval-augmented generation framework with adaptive selection across Quick, Expanded, and Exhaustive retrieval modes and generates answers from retrieved EVLncRNAs records with PMID-linked citations. Benchmark evaluation showed complete request execution, high retrieval coverage and full citation validity within the supplied evidence context. Functional Overview retrieves and aggregates lncRNA- and species-specific curated records before generating concise evidence-based summaries. In both functions, language models synthesize retrieved evidence rather than serve as independent sources of biological knowledge. EVLncRNAs 4.0 therefore provides an expanded, experimentally grounded resource for accessing and interpreting functional lncRNA evidence across species and biological contexts.

Homologous genes, especially those sharing close genomic locations and forming gene clusters, often have compensatory roles, therefore pose remarkable challenges for deciphering the functions and relationships of these genes. A comprehensive database of local homologous gene clusters with multimodal information would greatly facilitate such research. Here, we present LoGIC (https://omicslab.genetics.ac.cn/LoGIC/), a Local Gene-cluster Integrative Catalog that includes 20 306 local gene clusters identified from 30 representative eukaryotic species. In addition to the basic genomic features of each gene cluster, LoGIC also allows users to examine the sequence similarity of gene members within a cluster, the functional preference of clustered genes within a species, and the cross-species conservation of gene clusters. Information of tissue-wide expression profile and relationship with chromatin Hi-C structure is also available for human gene clusters. Together, LoGIC can facilitate studies of tandem gene duplication, paralog functional divergence, lineage-specific gene-family expansion, and evolutionary conservation or remodelling of homologous gene clusters across-species.

Annelids inhabit marine, freshwater, and terrestrial ecosystems and are important models for studying animal evolution, regeneration, and host-microbe interactions, yet their rapidly growing omics resources remain dispersed across repositories and publications. We developed AnnelidaDB (www.annelidadb.com), a species-centred database built on an updated phylogenetic framework that integrates genomes, mitochondrial genomes, transcriptomes, proteomes, epigenomes, single-cell datasets, host-associated symbiont genomes, and comparative synteny for Annelida. The current release covers 1121 species, including 152 species with nuclear genomes, 177 with mitochondrial genomes, 591 with transcriptome data, 5 with single-cell datasets comprising 209 802 cells, and 30 bacterial symbiont genomes. AnnelidaDB organizes these resources through 10 navigation directories and connects species, genome, gene, functional annotation, and gene-family records. Users can browse genomic assemblies, explore interactive single-cell atlases and 12 200 precomputed Oxford plots, run BLAST queries and search genes against the annelid-specific database, and submit chromosome-level data for syntenic analysis against 20 Annelida linkage groups. Sequences, annotations, and private results are available for download. As a comprehensive multi-omics resource dedicated to annelids, AnnelidaDB provides an accessible platform for comparative investigations of annelid genome evolution, adaptation, regeneration, and symbiosis.

The Structural Antibody Database (SAbDab) is a publicly available repository of experimentally determined antibody structures, first released in 2013. Explicit support for single-domain antibodies was added in 2021, with SAbDab-nano. Recently, increasing interest in antibodies has led to a proliferation of novel antibody formats, while simultaneous advances in machine learning (ML) have increased demand for standardized, high-quality structure data. Here, we present SAbDab2, re-engineered for the ML age. It introduces support for a variety of new formats, and makes it easy to retrieve and compare all known structures of a given antibody. In addition, SAbDab2 provides ready access to ML-grade structures of antibody and antibody-antigen-complexes, with standardized, versioned train/test splits. These will be updated every 6 months going forward and are available at https://zenodo.org/records/20083994. SAbDab2 itself is updated weekly and freely available at https://sabdab2.opig.stats.ox.ac.uk.

The interpretation of missense variants remains a major challenge in clinical genetics. "Meta-domains" aggregate population and pathogenic variation across homologous Pfam domain instances in the human proteome, providing per-residue context for interpreting variants of uncertain significance (VUS). Our 2019 implementation, MetaDome, is widely used and named in clinical variant-classification guidelines. Here, we present the MetaDome 2027 update, featuring a comprehensively updated dataset and GRCh38 support. The redesigned pipeline enables incremental updates of GENCODE, UniProtKB/Swiss-Prot, Pfam, gnomAD, and ClinVar while maintaining 100% sequence-identity gene-to-protein mapping. Annotated Pfam domain instances grew 14.9% from 71 419 to 82 069 and meta-domain-eligible Pfam families (≥2 human occurrences) by 73.3% from 3334 to 5778; Pfam domains are annotated to 92% of human proteins. Approximately 43% of mapped protein-coding nucleotides (14.3 million in GRCh38, 13.8 million in GRCh37) are in a meta-domain; in GRCh38, 67.9% (37 692 of 55 548) of pathogenic or likely pathogenic ClinVar missense variants fall at such a position. We show how MetaDome helped reclassify a de novo missense VUS in RALA and identify 52 463 ClinVar missense VUS for which meta-domains supply otherwise unavailable pathogenic evidence. MetaDome is freely available at www.metadome.app.

Proteolysis-targeting chimeras (PROTACs) represent an emerging therapeutic modality that harnesses the ubiquitin-proteasome system to induce targeted protein degradation (TPD). Their event-driven and catalytic mechanisms provide new opportunities to target previously undruggable proteins. However, the rapid expansion of TPD technologies, including molecular glues (MGs) and X-functional targeting chimeras (XTACs), together with the increasing application of artificial intelligence (AI) in degrader design, highlights the need for comprehensive and continuously updated data resources. Here, we present PROTAC-DB 4.0, an expanded knowledgebase covering major targeted degradation and induced-proximity technologies. Compared with the previous version, PROTAC-DB 4.0 contains 15 502 PROTAC entries, together with 24 798 MG entries and 272 XTAC entries, with expanded experimental annotations, patent-derived compounds, and modality-specific data fields. A major advancement in PROTAC-DB 4.0 is the integration of a large language model and retrieval-augmented generation-based natural language query system, which connects structured database records with scientific literature to enable intuitive knowledge discovery across PROTACs, MGs, and XTACs. By combining comprehensive data curation with AI-assisted information retrieval, PROTAC-DB 4.0 provides an accessible platform to support degrader discovery, data-driven modeling, and rational design of next-generation targeted degradation and induced-proximity therapeutics. PROTAC-DB 4.0 is freely accessible at http://cadd.zju.edu.cn/protacdb/.

The current peptide renaissance-driven by metagenomic and proteomic mining, molecular de-extinction, protein language models, and generative or evolutionary design-is producing millions of candidates, but useful machine learning depends on experimentally anchored, context-rich labels. The Database of Antimicrobial Activity and Structure of Peptides (DBAASP) is a manually curated resource linking peptides and peptidomimetics to quantitative antimicrobial, antibiofilm, anticancer, hemolytic, and mammalian-cell cytotoxicity measurements and their experimental conditions. DBAASP v4 contains >25 000 entries and extends the resource from a sequence-centered database to a structure-aware repository. Major additions include systematic SMILES representations, enhanced annotation of intrachain bonds and ring architectures, dedicated antibiofilm, and synergy records, PubChem links, >9200 molecular-dynamics trajectories, and new prediction services. The condition-resolved data support supervised learning, strain-, and target-specific models, model calibration, and surrogate fitness functions for genetic algorithms. SMILES representations further enable modeling of noncanonical peptides and peptidomimetics that cannot be adequately described by sequence alone. DBAASP therefore complements large peptide catalogues by providing the experimentally curated activity and structure layers, needed to convert sequence information into testable predictions, design, and optimization of biologically active compounds. DBAASP v4 is freely available at https://dbaasp.org.

Methanogenic environments harbour complex prokaryotic consortia that drive the anaerobic conversion of organic matter into methane and carbon dioxide. Beyond being key to the global carbon cycle, such processes have been steered into powerful biotechnologies for energy production in bioreactors. Decoding the functional and taxonomic landscape of these communities requires comprehensive surveys, including bacteriophages and plasmids, which, however, remain vastly underrepresented in current databases. Here, we present the microbial Catalogue of Anaerobic Digestion (CAnDi), a database comprising 36 480 prokaryotic genomes spanning 11 149 species recovered from 1256 metagenomic samples collected across bioreactors worldwide. The reconstructed metagenomes also encode ∼1.9 million viral and over 130 000 plasmid sequences. Identity clustering revealed ∼536 000 candidate viral species, of which 85% are novel relative to existing databases, as well as over 15 000 unique plasmids. Clusters of viral and plasmid proteins were generated de novo, totalling 1 721 182 and 123 813 distinct protein families, respectively. Together, these resources substantially advance previous systematic approaches to characterize anaerobic digestion microbiomes, providing an expanded foundation for investigating the ecological and evolutionary roles of the virome and plasmidome in methane-producing ecosystems. CAnDi is accessible via https://www.candi-db.org/.

Multitarget drug discovery has gained considerable traction in recent years, particularly for the treatment of complex diseases. Despite the growing demand for dedicated, well-curated resources to advance polypharmacology research and rational multitarget drug discovery, such resources remain severely limited. Here, we present MultiTargetDB, a database of multitarget ligands integrated with comprehensive biological annotations and measurement-level provenance. At the small molecule level, MultiTargetDB integrates chemical structures (including molecular scaffolds), their drug development stages, and target information, whereas at the target level, MultiTargetDB incorporates protein domains and families, functional associations, and pathway-disease annotations. The current release contains 173 977 multitarget molecules, 6496 protein targets, and 474 749 standardized quantitative bioactivity records, complemented by 46 234 high-confidence protein functional associations, 370 pathways from the Kyoto Encyclopedia of Genes and Genomes (KEGG), and 2493 diseases. MultiTargetDB supports multitarget ligand retrieval, target-combination analysis, quantitative activity profiling, inspection of source-level evidence, and biological interpretation of molecule-target-pathway-disease relationships, with a natural-language query assistant facilitating flexible database exploration. The integrated quantitative, pharmacological and biological data further provide a basis for drug-repurposing exploration, computational modeling, and machine-learning applications. MultiTargetDB v1.0 is freely accessible at https://multitargetdb.inflygent.com/.

Disease progression is driven by coordinated changes in tissue microenvironments, cellular composition, cell states, and transcriptional programmes. Single-cell and single-nucleus RNA sequencing enable these changes to be studied at cellular resolution. However, public disease-associated datasets are often organized within individual studies and annotated using heterogeneous clinical, pathological, or molecular staging systems, limiting comparison across disease stages and datasets. Here, we present PathoEvo (https://biomedlab-ai.com/pathoevo/), a disease-associated single-cell and single-nucleus transcriptomic database centred on disease progression. PathoEvo integrates 7 452 971 cells or nuclei from 981 samples and 59 datasets, covering 18 diseases and 27 progression frameworks. Through manual curation, unified processing, metadata harmonization, cell-type annotation, and stage mapping, PathoEvo organizes heterogeneous datasets into standardized resources organized around disease progression. Its data and analytical functions span three complementary levels, allowing users to examine changes in cellular composition associated with disease stage and cell states, characterize sample-level molecular variation across disease progression, and investigate gene expression dynamics across relevant cell populations and samples. Structured search and browsing further facilitate dataset discovery and navigation of stage-resolved information across diseases and staging systems. PathoEvo provides an integrative resource for investigating pathological evolution and comparing cellular and molecular patterns of disease progression across datasets at single-cell resolution.

Atomic-resolution nucleosome array structures have provided invaluable information on eukaryotic genome packing. Although these arrays are often obtained for regular fibers with designed sequences, such information complements in vivo genome organization data and modeling studies. Here, we analyze such systems with a focus on chromatin organization. We highlight both robust and unique features, as well as suitable average values for coarse-grained modeling, such as the relative orientation and positioning of neighboring nucleosomes, and opening of the linker DNA. Our associated web resource (https://www.biomath.nyu.edu/chromstruct/) shares these collected systems and provides tools for their analysis. The heterogeneous features exhibit a wide range of geometrical properties that depend on linker DNA length, core histone protein variants, auxiliary protein binding, and fiber interdigitation. Yet experimental fiber geometries can be reproduced reasonably using average parameters, as demonstrated here by simulations with our mesoscale model for regular fibers and as reported separately for life-like systems with epigenetic marks. Such analysis may also be useful for modular design of chromatin fibers. The structural characterization of these chromatin building blocks provides new insights into the regulation of chromatin architecture and has implications for the modular design of chromatin fibers.