PhenoTips is an open source, web-based system for collecting and analyzing structured phenotype information during the patient encounter, and providing various forms of decision support to clinicians. This deep phenotyping tool and its underlying database was developed at the University of Toronto in collaboration with several other Toronto area health care centers (1), and incorporates the Human Phenotype Ontology (HPO) as a controlled phenotype terminology and knowledge source.
Background and Goals
The emergence of low-cost genomic sequencing technologies is poised to revolutionize the practice of clinical genetics. Integrating this data into the clinical analysis pipeline will enable genotype-to-phenotype (G2P) associations to be made, and support the development of new and personalized therapies for a large number of Mendelian disorders. A critical barrier to exploiting this data is the fact that patient phenotype information is not amenable to computational and statistical analyses required to establish and validate G2P associations. This is particularly true for many rare and orphan disorders, where identification of patient cohorts with unifying phenotypic commonalities is hindered by a lack of standardized approaches for documenting physical findings in a clinical setting. Here, only the most basic phenotype data are captured in structured formats (e.g. height, weight, some lab data), while the most diverse, rich, and compelling physical findings are recorded as free-text. PhenoTips was developed to provide a tool for recording clinical phenotype observations using the controlled terminology provided by the Human Phenotype Ontology. The system is designed to be seamlessly integrated into the clinician workflow to facilitate data capture. Its backend that can support both clinical and research related needs by providing structured, anonymized data for research using powerful computational approaches.
PhenoTips consists of a web interface for data entry and analysis, a Jetty web server, and a HyperSQL Java database back-end to storing data. It can be downloaded at the project website , and installed and run on any machine with Web browser support (PCs, mobile devices, tablets). The system enables intuitive and integrated data entry (see below), and allows browsing, filtering, and querying for phenotypic data by various criteria. Search is supported by the Solr platform , which indexes all of the Human Phenotype Ontology including its synonyms and encoded semantic relationships. The software supports various user roles including view only, contributor, data administrator, and server administrator.
Features and Functionality
The web-based user interface is designed around the patient encounter, with sections to collect data related to demographic information, family history, medical history, measurements, clinical symptoms and physical findings (ie phenotypes), diagnoses, and genetic testing. What sets PhenoTips apart from other clinical exam support tools is its use of ontology driven features for capturing phenotype information using a standardized vocabulary, and its support by underlying knowledgebases and analytical tools to provide extensive clinical decisions support.
Standardized Phenotype Data Collection
PhenoTips was developed to be used in real-time to capture structured and standardized physical finding data during the patient encounter. Current data collection practices in clinical systems document phenotype findings in free-text, which results in idiosyncratic and variable wordings for the same concept, inconsistent spellings and abbreviations, and typographical errors. PhenoTips circumvents these issues by using the HPO (2) as a standardized vocabulary for describing physical findings and other patient data. The HPO is the most complete and widely used vocabulary providing over 10,000 well-defined human phenotype classes (aka 'terms') and associated synonyms, as well as many semantic relationships between these classes and related biomedical entities. The complexity of the ontology is well-hidden by an intuitive user interface that supports rigorous and complete data entry, allowing rich and computable phenotype profiles to be captured during the patient encounter (Figure 1 download). One such feature is the ability to configure pre-defined phenotype lists that collect the most relevant findings in a given clinical setting - for example, common cardiac phenotypes. Terms in these views can be expanded to view more specific sub-terms, and additional data about selected terms, such as age of onset or pace of progression, can be entered using dynamic widgets that are also supported by the underlying ontology. In addition, search boxes embedded in each pre-defined category allow ontology-guided searching that exploits the synonyms and hierarchical structure of the HPO to guide data entry. Precise and rigorous data entry is further supported by advanced features including the ability to indicate that each phenotype is either present ('Y'), assessed but not found ('N'), or unexplored ('NA') - allowing important documentation of relevant findings that are absent in a patient. Finally, a field to enter free-text descriptions of findings can be used include additional information, as well as feed back requests to the HPO when a needed term is not found. This is particularly important as some areas of the HPO are lacking in terms for some domains/specialties (e.g. oncology, prenatal markers).
Knowledge-Guided Decision Support
In addition to the above clinical decision support features related to standardized documentation forms and templates, PhenoTips provides more advanced, automated support including relevant data presentation and reference information and guidance. Some examples are listed below.
- Automated data plotting and interpretation: PhenoTips can plot quantitative measurements of patient height, weight, and various body part size measurements, as well as generate growth curves, percentiles, and Z-scores for measurements entered over time. Any extreme values will trigger selection of any corresponding phenotype from HPO by the system - for example, head size under first percentile will result in a "microcephaly" (HP:0000252) annotation in the patient record.
- Automated phenotype suggestion: The Online Mendelian Inheritance in Man (OMIM) knowledgebase is leveraged by PhenoTips to identify additional phenotypes that may be associated with those entered for a patient. As data is entered, an advanced similarity ranking algorithm calculates and presents a ranked list to the clinician to suggest additional phenotypes that should be assessed, to ensure a more complete record is obtained.
- Automated diagnosis support: PhenoTips also implements OMIM-based similarity algorithms to provide a ranked list of the top 20 genetic disorders that best match the complete patient phenotype profile. This approach analyzes the clinical features of OMIM disorders that have been mapped to HPO terms (2), and also processes data regarding their frequency in the general population, and features asserted not to occur in a given disorder. This functionality addresses a key task clinicians identified as one of the most difficult and time-consuming challenges in the clinical workflow.
Support for Clinical and Translational Research
The comparison of genomic variation data across patients is key to identifying the genetic basis of Mendelian disorders. Particularly for rare diseases, this requires the sharing of data across geographically dispersed and technologically diverse hospitals. The PhenoTips system can support this activity by collecting standardized, machine-computable phenotype data, and exporting/exchanging this data in a customizable and de-identified manner. In this way, a comprehensive catalog of phenotype-genotype correlations can be built to address some of the most critical open challenges in clinical genetics research. Ontology-coded phenotype data is also being applied in translational research, to identify model systems for researching disease mechanisms and therapies. This work is enabled by interspecies anatomy and phenotype ontologies such as Uberon (3) and UberPheno (4), which support ontology-based algorithms to calculate the "semantic similarity" of phenotype profiles between human patients and animal or cell-based research models. Efforts such as the Monarch Initiative  are developing advanced ontologies, algorithms , and tools to enable integrated analysis of semantically coded phenotype data from systems such as PhenoTips, model organism databases, and other public genotype-phenotype resources. This work will allow researchers to discover hidden connections between genes, phenotypes, and disease, and identify new systems in which to study human disorders.
- Girdea, Marta, et al. "PhenoTips: Patient Phenotyping Software for Clinical and Research Use." Human mutation (2013). PMID:23636887
- Robinson, Peter N., et al. "The Human Phenotype Ontology: a tool for annotating and analyzing human hereditary disease." The American Journal of Human Genetics 83.5 (2008): 610-615. PMID:18950739
- Mungall, Christopher J., et al. "Uberon, an integrative multi-species anatomy ontology." Genome Biol 13.1 (2012): R5.b PMID:22293552
- Köhler, Sebastian, et al. "Construction and accessibility of a cross-species phenotype ontology along with gene annotations for biomedical research."F1000Research 2 (2013). F1000 2013
Submitted by Matthew Brush