HPO and Disease Phenotype Databases

  • Phenotypes are the observable characteristics of an organism resulting from interactions between its genetic makeup, environmental influences, and biological processes. In human disease research, phenotypes include physical abnormalities, clinical signs, symptoms, laboratory findings, developmental features, and changes in organ function. Examples include short stature, intellectual disability, congenital heart defects, seizures, hearing impairment, and abnormal blood glucose levels. Phenotype information is essential for understanding genetic diseases because individuals with the same disorder may display different clinical features, while unrelated disorders may produce similar symptoms. Bioinformatics databases and ontologies provide standardized ways to describe, organize, compare, and analyse these characteristics.
  • The Human Phenotype Ontology (HPO) is one of the most important resources for representing human disease phenotypes. It provides a standardized vocabulary of clinical abnormalities and phenotypic features, assigning a unique identifier and structured definition to each term. HPO was developed to address the problem of inconsistent clinical terminology across medical records, scientific publications, genetic databases, and diagnostic laboratories. For example, similar clinical findings may be described using different expressions by different clinicians or research groups. HPO provides standardized terms that allow these descriptions to be represented consistently and analysed computationally. It is widely used in rare-disease research, clinical genetics, genomic diagnostics, phenotype matching, and genotype–phenotype analysis. <Cite refs={[“turn624709search8″,”turn624709search3”]}/>
  • Each HPO term has a stable identifier in the format HP:nnnnnnn, together with a preferred name, definition, synonyms, and relationships to other ontology terms. For example, the term Atrial septal defect has the identifier HP:0001631. Other terms describe abnormalities involving the nervous system, cardiovascular system, skeletal system, eyes, ears, skin, metabolism, and other biological systems. These identifiers make it possible to exchange phenotype information without relying only on free-text descriptions. HPO terms can be used in clinical databases, research datasets, variant interpretation pipelines, and computational tools that compare patient phenotypes with known disease profiles.
  • HPO is organized as a hierarchical ontology, commonly represented as a directed acyclic graph (DAG). Broad categories are connected to more specific clinical features through relationships such as is_a and other formally defined relations. For example, a specific cardiac abnormality can be placed beneath a broader cardiovascular-system abnormality. This hierarchical structure allows computational systems to recognize relationships between specific and general phenotype descriptions. If a patient has a highly specific clinical feature, an algorithm can also consider its broader parent terms when comparing the patient’s phenotype with disease annotations. This is particularly useful when patients have incomplete clinical descriptions or when different databases describe similar findings at different levels of detail.
  • HPO includes terms describing many different types of clinical abnormalities. These include morphological abnormalities, abnormalities of organ function, neurological findings, biochemical abnormalities, developmental abnormalities, growth disturbances, and clinical course. Some terms describe the presence of a feature, while others can represent its absence or modification through annotation qualifiers. Additional information may describe the age of onset, frequency, severity, laterality, or other characteristics. These details are important because the same phenotype may have different diagnostic significance depending on when it develops, how severe it is, and whether it occurs in all affected individuals.
  • An important component of the HPO project is the Human Phenotype Ontology Annotations (HPOA) dataset. HPOA connects HPO terms with diseases and, in related annotation files, with genes and inheritance information. Disease–phenotype annotations describe the clinical features associated with a particular disorder. Gene–phenotype and gene–disease association files provide additional relationships that can support computational analysis. The annotations may include evidence codes, references to scientific publications, frequency information, age of onset, sex qualifiers, and other contextual details. This allows researchers to move beyond a simple list of symptoms and examine how frequently particular features are reported in a disease and what evidence supports each annotation. <Cite refs={[“turn624709search10”]}/>
  • Disease–phenotype annotations are not equivalent to direct gene–phenotype relationships. A disease may be associated with many phenotypic features, and a gene may cause more than one disease with different clinical presentations. A gene–phenotype association may therefore need to preserve the disease context through which the relationship was established. This distinction is important in computational analysis because directly combining all phenotypes associated with a gene can create an overly broad profile. HPOA and related resources help preserve the connections among diseases, genes, phenotypes, inheritance patterns, and supporting evidence.
  • HPO is regularly updated as new clinical features are described and existing terminology is refined. The ontology and annotation datasets are available for download in machine-readable formats, including OBO and OWL for ontology structure and tabular files for phenotype annotations. These files can be integrated into Python, R, and other bioinformatics workflows. The HPO website and ontology services provide term browsing, searching, definitions, and links to related resources. Researchers should record the HPO release date and annotation version used in an analysis because terms, definitions, relationships, and disease annotations can change over time. The September 2026 HPO release is listed in ontology repositories, illustrating the importance of version-specific data management. <Cite refs={[“turn624709search1″,”turn624709search3”]}/>
  • The Monarch Initiative is an important complementary resource that integrates genotype, phenotype, disease, and biological knowledge across humans and model organisms. It connects information from multiple databases using standardized identifiers and ontologies, allowing researchers to explore relationships among genes, diseases, phenotypes, pathways, and species. Monarch supports phenotype-based analysis, disease discovery, cross-species comparisons, and knowledge graph exploration. Its infrastructure incorporates HPO, the Mondo Disease Ontology, model organism phenotype ontologies, and other biological resources. This integration is particularly useful when investigating rare diseases or comparing human disease features with experimental findings in animal models. <Cite refs={[“turn624709search4″,”turn624709search10”]}/>
  • The Mondo Disease Ontology (MONDO) provides a standardized and harmonized representation of human diseases. Different databases may use different names, identifiers, and classifications for the same disorder. Mondo addresses this problem by integrating and reconciling disease concepts from resources such as OMIM, Orphanet, the Disease Ontology, and other biomedical databases. It provides stable MONDO identifiers and structured relationships among disease entities. Mondo is particularly useful for connecting disease records across resources and avoiding inconsistent disease terminology during large-scale data integration. It should be distinguished from HPO: Mondo describes diseases, while HPO describes the phenotypic features observed in individuals with those diseases. <Cite refs={[“turn624709search4″,”turn624709search9”]}/>
  • Online Mendelian Inheritance in Man (OMIM) is a major resource for human genes and genetically determined phenotypes. It provides detailed descriptions of inherited disorders, associated genes, molecular mechanisms, clinical features, and scientific literature. OMIM is especially important for Mendelian disease research and the investigation of genotype–phenotype relationships. Its disease and phenotype information contributes to HPO annotations, while HPO provides standardized terms that make OMIM-related clinical features more suitable for computational analysis. OMIM is primarily a curated knowledgebase of genes and genetic phenotypes rather than a general-purpose clinical terminology ontology.
  • Orphanet is a major resource dedicated to rare diseases and orphan drugs. It provides information about rare-disease definitions, clinical characteristics, epidemiology, associated genes, inheritance patterns, diagnostic resources, and specialist centres. Each rare disease is assigned an ORPHAcode, which supports consistent identification across research and healthcare systems. Orphanet contributes disease information and phenotype annotations to HPO-related resources. It is particularly valuable for rare-disease analysis because many disorders have overlapping symptoms, different historical names, and limited published clinical information.
  • NCBI MedGen is a resource for human diseases, disorders, and phenotypes that integrates information from multiple biomedical databases. It provides disease descriptions, identifiers, synonyms, and links to related resources such as ClinVar, OMIM, GeneReviews, and other NCBI databases. MedGen supports searches by disease name, gene, clinical feature, and related concepts. It is useful for connecting clinical terminology with genetic and genomic records, particularly when different databases use different names for the same disorder.
  • DECIPHER is a genomic and phenotypic resource used extensively in the study of rare and undiagnosed genetic disorders. It connects genomic variants, including copy-number changes and sequence variants, with clinical phenotypes described using standardized terminology. Researchers can investigate genes and genomic regions associated with developmental disorders, intellectual disability, congenital abnormalities, and other rare conditions. DECIPHER supports phenotype-based comparison and the interpretation of genomic findings in the context of clinical observations. Its data-sharing framework also supports collaboration among clinical and research communities, subject to appropriate consent and access arrangements.
  • GeneReviews provides expert-authored reviews of inherited disorders, including their clinical features, molecular genetics, diagnosis, inheritance, management, and genetic counselling. It is particularly useful for understanding how a disease phenotype develops and how clinical features relate to underlying molecular mechanisms. GeneReviews is not a formal phenotype ontology or a structured disease-association database, but it provides detailed clinical context that complements HPO, OMIM, Orphanet, and ClinVar.
  • The Disease Ontology (DO) is another resource for the standardized representation of human diseases. It provides a structured vocabulary of disease concepts and relationships that supports data integration and computational analysis. Disease Ontology terms can be connected with genes, pathways, publications, and other biomedical entities. It is frequently used in gene-set enrichment analysis and disease association studies. Mondo and Disease Ontology have related but distinct roles: Mondo emphasizes cross-resource disease harmonization, while DO provides its own structured disease classification and annotation framework.
  • The Experimental Factor Ontology (EFO) provides standardized terminology for experimental variables, biological characteristics, diseases, and other concepts used in biomedical research. It is widely used in resources such as the GWAS Catalog and other data-integration platforms. EFO can help represent disease traits, experimental conditions, and phenotypic measurements in a consistent manner. Its broader scope makes it useful for integrating phenotype and disease information with genomic association studies and other experimental datasets.
  • Clinical terminology systems such as SNOMED CT and the International Classification of Diseases (ICD) also provide important disease and clinical-feature classifications. SNOMED CT supports detailed clinical terminology in electronic health records, while ICD is widely used for disease classification, reporting, and healthcare statistics. These systems serve different purposes from HPO. HPO is designed to describe phenotypic abnormalities in a form suitable for biomedical research and computational analysis, whereas SNOMED CT and ICD support broader clinical documentation and classification. Mapping among these systems can support data exchange, but mappings should be checked carefully because concepts may differ in specificity or meaning.
  • Several phenotype databases and ontologies are available for model organisms. The Mammalian Phenotype Ontology (MP) describes phenotypic abnormalities observed in mice and is important for connecting experimental animal models with human disease. The Zebrafish Phenotype Ontology (ZP) describes phenotypes observed in zebrafish, while the Worm Phenotype Ontology (WBPhenotype) supports phenotype annotation in Caenorhabditis elegans. The FlyBase resource provides phenotype information for Drosophila melanogaster, and the Plant Ontology (PO) describes plant anatomical structures and developmental stages, alongside related plant phenotype resources. These resources allow researchers to investigate whether genes and biological pathways produce similar phenotypic effects across species.
  • Cross-species phenotype analysis is supported by ontology mappings and integrated resources such as Monarch. A gene associated with a human neurological disorder may have a corresponding orthologue in mice or zebrafish. If disruption of that orthologue produces related neurological abnormalities, the model organism may provide experimental evidence about the gene’s biological function. However, phenotypes are not always directly equivalent across species. Differences in anatomy, development, physiology, and experimental conditions must be considered. Cross-species comparison therefore requires carefully defined ontology relationships and biological interpretation rather than simple matching of similar words.
  • HPO and disease phenotype databases support several important applications in bioinformatics. One major application is phenotype-driven disease diagnosis. A patient may present with several clinical features, such as developmental delay, seizures, abnormal muscle tone, and distinctive facial characteristics. These features can be encoded using HPO identifiers and compared with phenotype profiles of known diseases. Computational tools can rank candidate disorders according to the similarity between the patient’s clinical presentation and the recorded disease phenotypes. This approach is particularly useful in rare-disease diagnosis, where individual disorders may be difficult to recognize from clinical examination alone.
  • Phenotype similarity analysis is based on the semantic relationships among ontology terms. Two phenotypes may be considered related even when their names are different, provided that they share a meaningful ancestor in the ontology. Information-content methods can assign greater weight to specific and uncommon features than to broad features shared by many diseases. Algorithms can combine the similarity scores of multiple patient phenotypes and compare them with disease-associated phenotype profiles. This helps identify candidate disorders even when a patient does not display every feature traditionally associated with a disease.
  • Tools such as Phenomizer support disease ranking using clinical features represented with HPO terms. Exomiser combines phenotype similarity with genomic information to prioritize candidate genes and variants. Phen2Gene uses phenotype information to rank candidate disease-associated genes. These tools can be used in different stages of rare-disease investigation, depending on whether the input consists primarily of clinical features, genomic variants, or both. Their rankings are computational predictions and should be reviewed alongside variant quality, inheritance, functional evidence, and clinical assessment. <Cite refs={[“turn624709search5″,”turn624709search8”]}/>
  • Another important application is genotype–phenotype association analysis. Researchers can investigate which phenotypic features are associated with particular genes, variants, or molecular pathways. This can reveal the clinical consequences of different genetic mechanisms and help explain why mutations in the same gene may produce different disorders. Disease phenotype annotations can also support the identification of genes that have not previously been associated with a particular clinical presentation. Such analyses are valuable in rare-disease research, functional genomics, and the discovery of new disease mechanisms.
  • HPO and related databases are also used in variant prioritization. Whole-exome and whole-genome sequencing can identify thousands of variants, many of which are unlikely to explain the patient’s condition. By combining phenotype similarity with predicted variant consequences, inheritance patterns, population frequencies, and gene–disease evidence, bioinformatics tools can reduce the number of candidate variants requiring detailed investigation. For example, a rare variant in a gene associated with a specific neurological disorder may receive a higher priority if the patient displays several characteristic neurological features. Nevertheless, phenotype similarity alone cannot establish that a variant is pathogenic.
  • Disease phenotype resources also support cohort analysis and research data harmonization. Clinical studies may collect patient information using different terminology, free-text descriptions, and institutional coding systems. Mapping these records to HPO or other standardized ontologies makes it possible to compare patients across hospitals, research projects, and countries. Researchers can identify shared clinical features, define disease subgroups, investigate phenotype frequencies, and discover relationships between clinical characteristics and genomic findings. Standardized phenotype annotation is therefore important for building large, reusable rare-disease datasets.
  • An important standard for exchanging structured phenotype and disease information is the Phenopacket Schema, developed through the Global Alliance for Genomics and Health (GA4GH) community. A Phenopacket can represent information about an individual or biological sample, including phenotypic features, disease information, biospecimens, genomic interpretations, and related clinical metadata. HPO terms can be used to encode phenotypic features within a Phenopacket. This supports standardized exchange of phenotype and genomic information between diagnostic laboratories, research groups, databases, and computational tools. Phenopackets are particularly valuable for improving interoperability and reproducibility in rare-disease research. <Cite refs={[“turn624709search4”]}/>
  • The integration of phenotype databases with genomic resources is essential for comprehensive disease analysis. HPO can be combined with OMIM, Orphanet, ClinVar, ClinGen, DECIPHER, gnomAD, and gene annotation resources to connect clinical findings with genetic evidence. Mondo can help harmonize disease identifiers, while gene nomenclature resources provide standardized gene symbols and identifiers. Functional annotation databases such as Gene Ontology, Reactome, and KEGG can provide additional information about biological processes and pathways associated with candidate genes. Together, these resources allow researchers to move from clinical observations to disease concepts, candidate genes, genetic variants, and possible molecular mechanisms.
  • Despite their importance, phenotype databases have several limitations. Clinical features may be incompletely documented, and different clinicians may use different terminology to describe similar findings. Some phenotypes are subjective or difficult to measure, while others vary with age, disease progression, treatment, and environmental conditions. A patient may not display all the features associated with a particular disorder, and the frequency of a phenotype may differ among families or populations. Disease annotations may also be based on small case series or limited published evidence. These factors can affect phenotype matching and disease ranking.
  • Another challenge is the interpretation of absent phenotypes. A feature not recorded in a patient’s clinical profile should not automatically be interpreted as absent. It may simply not have been assessed or documented. This distinction is important when computational systems compare patient profiles with disease annotations. Explicitly recording negative findings, when clinically evaluated, can improve phenotype-based analysis. Age of onset, disease stage, severity, frequency, and sex-specific information can also influence interpretation and should be preserved where available.
  • Ontology updates and terminology mappings introduce additional challenges. HPO terms may be revised, replaced, or reorganized as knowledge develops. Disease identifiers may also change as resources harmonize their classifications. Different databases may use different levels of specificity, and a mapping between two terms does not always imply complete equivalence. Researchers should therefore record ontology versions, annotation releases, and mapping sources. When integrating phenotype information, it is important to preserve original clinical descriptions alongside standardized identifiers so that the original meaning can be reviewed.
  • Reproducible phenotype-based bioinformatics requires careful documentation of the HPO release, disease annotation version, ontology mapping files, software versions, similarity algorithms, and filtering thresholds. Researchers should record which clinical features were included, which were explicitly absent, and whether age of onset or frequency qualifiers were available. Patient-level phenotype data must also be handled according to consent, privacy, and data-sharing requirements. Because clinical phenotypes can be identifying when combined with genomic information, appropriate data protection and access controls are essential.
  • Overall, HPO and disease phenotype databases provide essential infrastructure for organizing and analysing clinical characteristics in human disease. HPO supplies standardized phenotype terminology, while HPOA connects phenotypic features with diseases and genes. Resources such as Monarch, Mondo, OMIM, Orphanet, MedGen, DECIPHER, GeneReviews, Disease Ontology, and model organism phenotype databases provide complementary information. Tools such as Phenomizer, Exomiser, and Phen2Gene use these resources to support phenotype matching, disease prioritization, and genomic diagnosis. Together, they enable researchers to connect clinical observations with genetic variation, disease mechanisms, and biological pathways, making phenotype-based analysis an important part of modern bioinformatics and precision medicine.
Author:

Leave a Reply

Your email address will not be published. Required fields are marked *