BIOETHICS

The Bioethics Imperative XV

Ethics and the Literature: Citations III

“Mokita”: the truth we all know and agree not to talk about.

Scenario: Frank Lee Nayeff has just cloned his first gene. He eagerly searches GenBank for homologies and finds hundreds of partial matches. He is bewildered by the various functions of the genes with homology and goes home late that night discouraged. Fortunately, the next morning, his adviser shows him that if he strips off the vector sequences, the confusion is resolved. Frank trots happily to the library to search for older references on the physiology of the protein that he has cloned and, a month later, submits a manuscript on his clone and the 20-year history of the physiology of this protein. He and his adviser are both humbled by reviewers’ comments that the literature on this important protein actually goes back 50 years, a body of literature that Frank missed because of changes over time in the terms used to describe the physiology involved.

How many databases do you use? Do you know if they are curated—that humans have processed each incoming file for accuracy? How is each database organized and updated? When you search several databases for the same information, do you adapt your terms and strategy to match each database’s requirements; does a small or zero retrieval mean that you missed something, or was there actually little or nothing to be found?

In molecular biology, a curated database makes a difference in the quality of the information retrieved because bad data can mislead, distort, or lead to serious errors. SwissProt (http://www.ebi.ac.uk/swissprot/index.html) is curated, but GenBank (http://www.ncbi.nlm.nih.gov/Genbank/index.html) is not. In a research apprenticeship course, we had students search GenBank for a common vector sequence using Sequencer. In minutes, we found hundreds of examples of vector sequences. Obviously, the genes had not been stripped of the vectors used to clone them prior to submission, and we had a chuckle over some of the famous people who had entered vectors—some without inserts—in GenBank.

Another data retrieval concern is that a search term might not exist prior to a particular date. For example, until 1983, the term AIDS did not exist in MEDLINE. Informally, the disease was called “gay-related immunodeficiency syndrome,” or the “gay cancer.” The disease was going strong, but from 1979 to 1982, articles about it were indexed under “immunologic deficiency syndromes.” Articles were reindexed under “acquired immunodeficiency syndrome” in 1983; then, if the term AIDS was entered, MEDLINE automatically mapped the acronym to the new, correct search term. Without checking on the indexing history, searching with the term AIDS might well make one conclude that the disease did not exist until 1983, and missing all the pre-1983 literature might look a bit foolish. How can you deal with this issue? Ask a librarian to show you how to read “scope notes” in the database thesaurus and how to read the chronology of term changes, additions, and deletions for a database of interest.

Like the science they report, reproducibility and consensus are criteria used to update databases. Until something is reproducible and, therefore, credible, a new topic is often subsumed under a broader heading. Most database producers will add a new search term or concept, drug name, and so forth only after a significant number of articles have been written on the topic. For a database such as MEDLINE (PubMed), the National Library of Medicine also uses the input of librarians, physicians, and researchers when deciding to add, change, or delete search terms. This method for updating search terms is probably true for Agricola (http://www.nal.usda.gov/ag98/ag98.html) as well.

It is good practice to read each database’s published description to understand seven critical factors: (1) the journals covered; (2) the span of coverage in years; (3) whether a controlled vocabulary (specific index terms that must be used), free text, or both are used; (4) whether indexing is done by machines or people, and the first language of the indexers (if done by people); (5) the indexing priority (e.g., are articles indexed from the top-10 plant biology journals before the top-10 insect journals are indexed, implying that citations for plant biology will be more current than those for insect journals?); (6) the time lag between journal publication date and the date that articles appear online; and (7) the mechanism or procedure by which the database producer collects suggestions for new terms, new capabilities, and new journals.

In biotech pharmaceutical work, even the computers used in research have to be validated for accurate performance. Furthermore, when biotech scientists include a literature search in a U.S. Food and Drug Administration application, they must explain why certain databases were chosen and describe them using factors such as those listed above. Early on, students should be required to get in the habit of seeking out the capabilities and the limitations so they can quickly ascertain which databases match their search requirements and what search techniques are specific to each database. Again, a librarian can help set you on the fast track.

Next: A summary of citations guidelines.

Tamara Turner
Librarian and editor, Seattle

Dina Mandoli
mandoli@u.washington.edu