Support for subjects with hierarchical structure (disease, ECO)

[kcs3]

Here is the original proposal text:

Use hierarchies in search

Overview of current search methods: Search currently matches directly on individual words, or quoted phrases, that appear in any of the controlled vocabularies. This includes both the official titles as well as any synonyms or aliases provided by the relevant ontologies. Examples include gene symbols and their aliases, tissue names and their synonyms, Evidence Ontology terms and their aliases, etc. All “subjects” hit in a search are returned as individual rows in the search results, with each row containing a link to observations using any matched term or its alias or synonym. In a second phase, search currently also attempts to identify observations that contain ALL the terms that were searched on. Therefore, relying exclusively on exact-text matching is a limitation of the current searching paradigm. While inclusion of synonyms and aliases helps (e.g., searching for “Non-Hodgkin” will match the subject entry “follicular lymphoma” because “Follicular Non-Hodgkin Lymphoma” is an alias of the corresponding NCI thesaurus term), many useful “hits” will be missed. For example, observations involving Burkitt’s lymphoma, a subtype of Non-Hodgkin Lymphoma, will escape discovery as they will not match the search term “Non-Hodgkin”, either directly or through an alias.

Implementing hierarchies: To rectify this deficiency, we propose to extend the database schema of the Dashboard to allow for explicit representation of hierarchical relationships. Specifically, if hierarchies are found suitable for direct incorporation in the Dashboard, we propose to include knowledge of the term hierarchy of matched terms in the results. This involves looking for matches not only of terms directly, but also to their parents or children, depending on whether a more permissive or restrictive search is desired

[kcs3]

Prior related issue is #265.

[kcs3]

Implemented in #487