OmniPath Sample Datasets for Testing and Toy Knowledge Graph Development

Dataset overview This dataset contains random samples derived from five OmniPath dataset exports: annotations, protein complexes, enzyme-substrate relationships, molecular interactions, and intercellular communication annotations. The dataset was created for software development, testing, teaching, demonstrations, workflow validation, and toy downstream examples such as building small knowledge graphs. It is intended for situations where realistic OmniPath-like tabular data structures are needed without using the full OmniPath exports. Dataset contents The data are provided as five compressed folders: annotations_sampled.zip complexes_sampled.zip enzyme-substrate_sampled.zip interactions_sampled.zip intercell_sampled.zip Each folder contains gzip-compressed TSV files sampled at different row counts. Samples were generated randomly without replacement. Sample sizes larger than the available number of source rows were not generated. Data Fields The tables presented below were generated with assissted AI due original sources lack of explicit information about how to interpret the data. This constitute another limitation about why the samples generated here should not be taken as part of real production datasets, especially in informed medical decisions or predictions (see section Limitations below). Annotations The annotations samples are derived from the OmniPath annotations export. These records describe molecular annotations associated with UniProt identifiers and gene symbols, including annotation source, label, value, entity type, and record identifier. Column Data type Description Example Constraints / Notes uniprot String Identifier assigned to the biological entity associated with the annotation. Q8NEH6 Usually a UniProt accession. Complex identifiers use the COMPLEX: prefix and may contain multiple component accessions separated by underscores. genesymbol String Gene symbol or symbolic name corresponding to the annotated entity. MNS1 Complexes use a COMPLEX:-prefixed value containing component gene symbols separated by underscores. entity_type String Type of biological entity to which the annotation applies. protein Values observed in the sample are protein and complex. source String Database or resource from which the annotation originates. UniProt_location Values observed include UniProt_location, TopDB, Vesiclepedia, HPA_tissue, MSigDB, DisGeNet, UniProt_tissue, UniProt_keyword, CytoSig, InterPro, and CSPA_celltype. label String Name of the annotation property represented by the record. location Interpretation of value depends on this field. Observed labels include location, tmregions, pmid, vesicle, organ, level, tissue, cell_type, geneset, and score. value String Value assigned to the annotation property identified by label. Cilium axoneme Stored as text even when the content represents an integer, Boolean, score, identifier, category, or publication ID. It should therefore not be globally converted to a single numeric type. record_id Integer Unique identifier assigned to the annotation record. 22179 All 100 sampled records contain a distinct value. It identifies the annotation row rather than the biological entity itself. Complexes The complexes samples are derived from the OmniPath protein complexes export. These records describe named protein complexes, their component proteins, component gene symbols, stoichiometry, source resources, references, and external identifiers. Column Data type Description Example Constraints / Notes name String Human-readable name assigned to the molecular complex. Nucleosome Optional. It is missing in 79 of the 100 sampled records. Multiple component combinations may share the same complex name. components String / Delimited list Identifiers of the molecular entities forming the complex. A0A2R8Y619_P0C0S8_P62805_P84243 Components are generally separated by underscores. Values may include UniProt accessions and identifiers from other namespaces, such as CHEBI:. An underscore-delimited value must not automatically be interpreted as one accession. components_genesymbols String / Delimited list Gene symbols or symbolic names corresponding to the entries in components. H2AC15_H2BK1_H3-3A_H4C4 Symbols are separated by underscores and normally follow the same order as components. Non-gene entities may retain namespace-prefixed identifiers such as CHEBI:. stoichiometry String / Delimited integer list Stoichiometric coefficients associated with the listed complex components. 0:0:0:0 Component coefficients are separated by colons. The number of coefficients should generally correspond to the number of components. A value of 0 may represent unspecified or unavailable stoichiometry rather than absence of the component. sources String / Delimited list Databases or resources reporting the molecular complex. SIGNOR Multiple resources are separated by semicolons, for example Compleat;CORUM. references String / Delimited list Literature references supporting the complex record. 9584200 Optional and missing in 90 of the 100 sampled records. Values generally represent PubMed identifiers; multiple references may be semicolon-delimited. identifiers String / Delimited list External database identifiers assigned to the complex. SIGNOR:SIGNOR-C371 Optional and missing in 58 sampled records. Each value normally uses the form database:identifier; multiple identifiers are separated by semicolons. Enzyme-substrate relationships The enzyme-substrate samples are derived from the OmniPath enzyme-substrate export. These records describe enzyme-substrate relationships, substrate modification sites, modification types, source resources, literature references, curation effort, and taxonomy identifiers. Column Data type Description Example Constraints / Notes enzyme String Protein accession assigned to the enzyme responsible for the modification. P11362 The same enzyme may occur in multiple records and may modify several substrates or residues. enzyme_genesymbol String Gene symbol corresponding to the enzyme. FGFR1 Gene-symbol capitalization should be handled consistently during downstream processing. substrate String Protein accession assigned to the modified substrate. P11362 An enzyme may also act on itself, so enzyme and substrate can contain the same accession. substrate_genesymbol String Gene symbol corresponding to the substrate protein. FGFR1 The same substrate may occur in multiple rows when different sites or modifications are reported. isoforms String / Delimited list Protein isoform numbers for which the modification site is applicable. 1;8;19 One or more isoform numbers are separated by semicolons. The column should remain textual because it may contain multiple values. residue_type String One-letter amino-acid code identifying the modified residue. Y Values observed in the sample are Y, T, S, and K, corresponding to tyrosine, threonine, serine, and lysine. residue_offset Integer Position of the modified amino-acid residue within the substrate sequence. 605 Interpreted together with substrate, isoforms, and residue_type. Sequence-position conventions may depend on the referenced protein isoform. modification String Type of enzymatic modification assigned to the substrate site. phosphorylation Values observed in the sample are phosphorylation, acetylation, and dephosphorylation. sources String / Delimited list Databases or resources reporting the enzyme–substrate relationship. KEA;PhosphoSite;ProtMapper;SIGNOR;SIGNOR_ProtMapper;phosphoELM Multiple resources are separated by semicolons. Resource names containing suffixes such as _ProtMapper or _MIMP represent source-specific or mapped evidence. references String / Delimited list Source-qualified literature references supporting the modification record. KEA:12601080;ProtMapper:12601080;SIGNOR:12601080;phosphoELM:12601080 Optional and missing in 21 sampled records. Entries generally use resource:reference format and are separated by semicolons. A value such as dbPTM:- indicates that a source is present without a conventional publication identifier. curation_effort Integer Aggregate measure of the curation or evidence effort associated with the record. 4 Non-negative integer. It should be interpreted as metadata about evidence aggregation, not as a biological effect size or confidence probability. ncbi_tax_id Integer NCBI Taxonomy identifier assigned to the organism of the enzyme–substrate relationship. 9606 Values observed in the sample are 9606, 10090, and 10116, representing human, mouse, and rat. Interactions The interactions samples are derived from the OmniPath molecular interactions export. These records describe interactions between biological entities, including source and target identifiers, gene symbols, directionality, stimulation and inhibition flags, consensus annotations, source resources, references, evidence metadata, taxonomy identifiers, and entity types. Column Data type Description Example Constraints / Notes source String Identifier assigned to the source node of the interaction. P52926 Usually a UniProt accession, but values may also identify complexes or other entities. Complex identifiers use the COMPLEX: prefix. target String Identifier assigned to the target node of the interaction. Q13315 Together with source, identifies the pair of entities associated with the interaction. source_genesymbol String Gene symbol or symbolic name corresponding to the source entity. HMGA2 Complexes may be represented by underscore-delimited component symbols. target_genesymbol String Gene symbol or symbolic name corresponding to the target entity. ATM Unknown mappings may be represented by values such as UNK. is_directed Integer / Boolean flag Indicates whether the interaction has a defined direction from the source entity to the target entity. 1 1 indicates a directed interaction; 0 indicates an undirected interaction. is_stimulation Integer / Boolean flag Indicates whether the interaction is assigned a stimulatory or activating effect. 1 1 indicates stimulation; 0 indicates that stimulation is not assigned. A zero does not necessarily mean inhibition. is_inhibition Integer / Boolean flag Indicates whether the interaction is assigned an inhibitory effect. 0 1 indicates inhibition; 0 indicates that inhibition is not assigned. consensus_direction Integer / Boolean flag Indicates whether the assigned interaction direction is supported as the consensus direction. 1 1 indicates consensus support for the direction; 0 indicates no consensus direction. consensus_stimulation Integer / Boolean flag Indicates whether stimulation is the consensus regulatory effect. 1 1 indicates consensus stimulation; 0 indicates no consensus stimulation. consensus_inhibition Integer / Boolean flag Indicates whether inhibition is the consensus regulatory effect. 0 1 indicates consensus inhibition; 0 indicates no consensus inhibition. sources String / Delimited list Databases or resources reporting the interaction. SPIKE;SPIKE_LC Multiple resources are separated by semicolons. references String / Delimited list Source-qualified literature references supporting the interaction. SPIKE:21339738;SPIKE_LC:21339738 Optional and missing in 43 sampled records. Entries generally use resource:reference format and are separated by semicolons. omnipath Boolean Indicates whether the interaction belongs to the OmniPath core interaction dataset. True Boolean membership flag. It does not independently describe the biological mechanism. kinaseextra Boolean Indicates whether the interaction belongs to the additional kinase-interaction dataset. False Boolean dataset-membership flag. ligrecextra Boolean Indicates whether the interaction belongs to the additional ligand–receptor dataset. False All 100 sampled values are False; this does not prove the complete dataset contains no True values. pathwayextra Boolean Indicates whether the interaction belongs to the additional pathway-interaction dataset. True Boolean dataset-membership flag. mirnatarget Boolean Indicates whether the record belongs to the microRNA–target interaction dataset. False Boolean dataset-membership flag. dorothea Boolean Indicates whether the interaction belongs to the DoRothEA transcription-factor regulon dataset. False Boolean dataset-membership flag. DoRothEA-specific metadata may be absent when this value is False. collectri Boolean Indicates whether the interaction belongs to the CollecTRI transcriptional-regulation dataset. False Boolean dataset-membership flag. collectri2 Boolean Indicates whether the interaction belongs to the secondary or updated CollecTRI dataset represented by this export. False Boolean dataset-membership flag. The exact distinction from collectri should be preserved from the producing system rather than inferred from the flag name alone. tf_target Boolean Indicates whether the record represents a transcription-factor–target relationship. False Boolean classification or dataset-membership flag. lncrna_mrna Boolean Indicates whether the interaction represents a long-noncoding-RNA–mRNA relationship. False All sampled values are False; the full dataset may contain other values. tf_mirna Boolean Indicates whether the interaction represents a transcription-factor–microRNA relationship. False All sampled values are False; the full dataset may contain other values. small_molecule Boolean Indicates whether a small molecule participates in the interaction record. False All sampled values are False; the full dataset may contain other values. dorothea_curated Boolean / Nullable Indicates whether the DoRothEA relationship is supported by curated evidence. True Optional and present in only 33 sampled records. Missing values generally indicate that DoRothEA-specific metadata is not applicable or unavailable. dorothea_chipseq Boolean / Nullable Indicates whether the DoRothEA relationship is supported by ChIP-seq evidence. False Optional and present in only 33 sampled records. dorothea_tfbs Boolean / Nullable Indicates whether the DoRothEA relationship is supported by transcription-factor binding-site evidence. False Optional and present in only 33 sampled records. All non-null values in the sample are False. dorothea_coexp Boolean / Nullable Indicates whether the DoRothEA relationship is supported by co-expression evidence. False Optional and present in only 33 sampled records. dorothea_level String / Nullable DoRothEA confidence level assigned to the transcription-factor–target interaction. D Optional and present in only 33 sampled records. Values observed are A, B, and D; confidence categories should be treated as ordered labels rather than ordinary text. type String Broad biological-regulation category assigned to the interaction. post_translational Values observed are post_translational, transcriptional, and post_transcriptional. curation_effort Integer Aggregate measure of curation or evidence effort associated with the interaction. 2 Non-negative integer. It should not automatically be interpreted as a normalized confidence score. extra_attrs JSON-formatted string Resource-specific attributes associated with the interaction evidence. {"SPIKE_effect":"1","SPIKE_mechanism":"Transcription Regulation"} Contains a JSON object serialized as text. Keys vary by contributing resource and record. Empty objects are represented as {}. Parse with a JSON parser rather than splitting the string manually. evidences JSON-formatted string Structured evidence object describing supporting and conflicting records by resource. {"id_a":"P52926","id_b":"Q13315","positive":[...],"negative":[],"directed":[],"undirected":[]} Contains nested JSON with entity identifiers and evidence grouped into categories such as positive, negative, directed, and undirected. Each evidence item may contain a resource, references, dataset, provenance route, and additional attributes. ncbi_tax_id_source Integer NCBI Taxonomy identifier assigned to the organism of the source entity. 9606 Values observed in the sample are 9606, 10090, and 10116. entity_type_source String Type of biological entity represented by the source node. protein Values observed are protein, complex, and mirna. ncbi_tax_id_target Integer NCBI Taxonomy identifier assigned to the organism of the target entity. 9606 Values observed in the sample are 9606, 10090, and 10116. entity_type_target String Type of biological entity represented by the target node. protein Constant value protein in the 100 sampled records. Intercell The intercell samples are derived from the OmniPath intercellular communication export. These records describe intercellular communication roles and annotations, including category, parent category, source database, UniProt identifier, gene symbol, entity type, consensus score, transmitter and receiver roles, secreted status, and plasma-membrane-related properties. Column Data type Description Example Constraints / Notes category String Intercellular communication or cellular-location category assigned to the entity. transmembrane Values observed include transmembrane, peripheral, plasma_membrane, secreted, cell_surface, growth_factor, and ligand. parent String Parent category in the intercellular annotation hierarchy. transmembrane May be equal to category or represent a broader category from which the record inherits. database String Database or computational resource providing the annotation. UniProt_topology Values observed include UniProt_topology, UniProt_keyword, Phobius, GO_Intercell, LOCATE, Ramilowski_location, OmniPath, and UniProt_location. scope String Indicates whether the annotation describes a general category or a more narrowly defined category. generic Values observed are generic and specific. aspect String Indicates whether the annotation describes localization or biological function. locational Values observed are locational and functional. source String Indicates whether the record comes directly from one resource or represents a combined annotation. resource_specific Values observed are resource_specific and composite. This field is conceptually different from database. uniprot String Identifier assigned to the annotated biological entity. O43909 Usually a UniProt accession. Complex records may use a complex identifier according to the producing system’s conventions. genesymbol String / Nullable Gene symbol corresponding to the annotated entity. EXTL3 Optional and missing in 5 of the 100 sampled records. Absence may occur when a symbol cannot be assigned or when the entity is a complex. entity_type String Type of biological entity represented by the record. protein Values observed are protein and complex. consensus_score Integer Aggregate score representing the level of support for the intercellular annotation across resources or evidence records. 7 Non-negative integer. The sample contains 14 distinct values. It should not be treated as a probability unless the producing system defines such a conversion. transmitter Boolean Indicates whether the entity is classified as a transmitter in intercellular communication. False True indicates transmitter classification; False indicates that the classification is not assigned. receiver Boolean Indicates whether the entity is classified as a receiver in intercellular communication. False True indicates receiver classification; False indicates that the classification is not assigned. secreted Boolean Indicates whether the entity is classified as secreted. False True indicates a secreted entity. This flag may summarize several underlying annotations. plasma_membrane_transmembrane Boolean Indicates whether the entity is classified as a transmembrane component of the plasma membrane. False True indicates plasma-membrane transmembrane classification. It is more specific than a general transmembrane category. plasma_membrane_peripheral Boolean Indicates whether the entity is classified as peripherally associated with the plasma membrane. False True indicates peripheral plasma-membrane association rather than membrane-spanning localization. Data quality The files are random samples from OmniPath exports. No additional biological curation, correction, normalization, deduplication, or validation was performed after sampling. The samples preserve the original column structure of each OmniPath export. Some fields may contain delimited values, boolean-like flags, references, provenance information, or structured metadata strings inherited from the original source tables. Samples were generated without replacement. The same source row should not appear more than once within a given sampled file, although the same biological entity or relationship may occur across different sample sizes or dataset folders. OmniPath-derived data statement All rights to the source data belong to the creators and maintainers of OmniPath. These sampled files are derived from OmniPath exports and are not intended to create, mirror, or redistribute a separate authoritative database from OmniPath data. The samples preserve real values from OmniPath source tables, but they are incomplete subsets generated for testing, development, examples, demonstrations, teaching, and toy downstream workflows such as building small knowledge graphs. These samples should not be treated as final biomedical data resources, used for medical decisions, or presented as substitutes for the original OmniPath datasets. Users requiring authoritative OmniPath data should consult the original OmniPath resources, documentation, and release information. Limitations This dataset should not be used for: clinical interpretation biomedical decision-making biological discovery experimental validation prediction of regulatory mechanisms pathway enrichment conclusions interaction-confidence assessment claims about complete OmniPath coverage claims about complete biological annotation coverage substitution of the original OmniPath datasets The sampled datasets are incomplete by design. Absence of a record from a sample does not imply absence from OmniPath or from biological knowledge.