In an effort to make the full potential of the NCBI Web services and underlying databases more available to users, the NCBI has begun a long-term project to improve the relevance and usefulness of search results. This effort is called the Discovery Initiative. A primary goal of the Discovery Initiative is to promote the discovery of previously hidden relationships in the large amount of pre-calculated similarity data and pre-compiled links between different molecular and literature databases available at the NCBI. Changes in search interfaces and result displays are being phased in gradually and include the appearance of database ads, alternative search suggestions, and various sensors that will bring to the surface results in other databases that may be more relevant to the search. Particular Discovery components will appear in a context-specific manner ultimately producing the most relevant result possible. To help with this effort NCBI is also designing Web interfaces and links so that the effectiveness and popularity of these changes can be measured and studied. In turn, the results of these studies will be used to improve NCBI’s services even more.
Discovery Components in PubMed
Click on image to enlarge
Figure 1. The new PubMed display for document summaries for a search with TPH1 (tryptophan hydroxylase 1). The gene symbol TPH1 triggers the Gene Sensor that appears at the top of the display. The Gene Sensor provides a direct link to the human gene record that has associated molecular data and a curated set of literature citations. Direct links to genes of the same name in rat, mouse, and a full search in the Gene database are also available. The right-hand discovery column shows the top five related searches listed under “Also try …”, the result set obtained by a title search with the term highlighted in the title, and “Recent Activity”, the last five page views or searches performed. The Review tab at the top of the results displays only the review articles when selected.
Click on image to enlarge
Figure 2. The Abstract Plus page in PubMed. Discovery components in the right hand column are the pre-computed Related Articles and Recent Activity. The Related Articles component highlights review articles and provides a link to all 100 related items in PubMed and a separate link to the ten related reviews.
Many Discovery-related changes can already be seen in the PubMed database, and more will be coming soon. Current Discovery components appearing on PubMed results pages may include Related Queries, Title Searches, a Review tab, the Gene Sensor, and Recent Activity as shown in . The Abstract Plus page view may include the top five Related Articles with review articles highlighted, Patient Drug Information and Recent Activity ().
The Related Queries component shown under “Also try …” in the search results is a completely new kind of feature offering suggested queries from the most popular PubMed queries that contain the current search term. Using these suggested queries may provide more precise results than the current search.
Certain Discovery components improve and make more obvious previously existing pathways that are powerful but may have been cryptic before. For instance, the results available for ‘Title Searches’ and the ‘Review’ tab have been available by field-limited searches (term[Title], term AND review[Publication Type]). But, despite its usefulness, only a small minority of PubMed searchers use fielded searching.
Likewise, exposing the top five related articles the Abstract Plus view provides a more obvious alternative to the Related Articles available in the Links menu. Related Articles has been removed from the summary view of the search results, but is still available in the Abstract Plus record view. The new feature that highlights review articles here is a popular enhancement called ‘Recent Activity’ that provides a gateway to the broader literature relevant to a particular field. The Recent Activity component partially replaces the functions of the History tab allowing navigation to previous searches and record views in PubMed. Unlike the History tab however, Recent Activity provides access to searches and record views in other Entrez databases.
Other Discovery components such as the Gene Sensor, triggered when a gene symbol is used in a search, show results from other databases that may be more directly relevant than those from the current database. In some cases these sensors may provide useful results when the current database does not. For example, searching with the human gene symbol STK40 produces no results in PubMed itself. The Gene Sensor, however, reports results for the search of STK40 in the Gene database and provides a link to the human, mouse, rat, and the complete search results in Gene. These gene records have links to a variety of molecular data including mRNA and genomic sequences, genomic regions and maps, expression information, homologs in other species, and in the case of the human gene, a literature citation when PubMed found none.
Upcoming Changes
Many new Discovery-related changes will be appearing over the next several months. Discovery components will be released initially to a fraction of users so that NCBI can measure the effect and popularity of changes to the system. Some changes will appear first in PubMed and will be implemented in other databases as appropriate. The Recent Activity component, for example, was recently ported to some of the molecular databases after a test period in PubMed. In some cases, components will be database-specific such as the Taxonomy report that appears in the right-hand column of Entrez sequence database search results. In all cases the goal of these changes is to improve the usability and quality of results obtained from NCBI services.
Summary
The NCBI Website is the premier portal to biomedical literature and molecular biology data. Interconnecting the literature and these data is an enormously rich set of similarity and linkage relationships where previously unknown connections are waiting to be uncovered. While access to these connections has always been possible, it’s clear that many visitors are not enjoying the full potential of the system. Recent and upcoming changes to the NCBI Web experience will help expose these connections and promote discovery of the most biologically significant records and relationships in the NCBI databases.
