ABSTRACT
The last chapter demonstrated how semantic similarity analysis in the Human Phenotype Ontology (HPO) can be used to implement a decision support system for clinical diagnostics. The basic strategy involved finding the diagnosis (disease) that is most similar to the query terms. This approach, however, has a number of drawbacks, including especially that it is not explicitly designed to deal with mistaken or irrelevant query terms, and it does not utilize information about the frequency of a given phenotypic abnormality among all patients with the same disease. These are both clinically important.
