How do Researchers Search for Scientific Resources: the Traditional (but still useful) Method? (2/3)

This research is still focused on search engines.

Names are useful, but identifiers that follow the nomenclature are often much more precise.
Names are useful, but identifiers that follow the nomenclature are often much more precise.
Suppose a researcher is interested in POU5F1. Searching for the name alone may return articles, products, laboratory services, educational pages, and even unrelated references. Searching for a stable identifier can reduce ambiguity.

The same principle applies to:

• genes
• proteins
• diseases
• chemical compounds
• datasets
• publications
• research software
• experimental resources

UniProt provides identifiers alongside protein names and synonyms, while scientific literature systems such as PubMed use old and new identifiers such as PMID. Crossref's metadata search can also retrieve scholarly records using information such as title, author, DOI and ORCID iD.
For a scientific website, exposing relevant identifiers is therefore more than a metadata exercise. It helps establish what the resource is actually about.
A page describing an antibody, for example, becomes considerably clearer when it identifies its target as: POU5F1 / OCT4 and connects that target to appropriate biological identifiers.

Scientific discovery rarely moves in a straight line.
A researcher may start with one article, follow a citation, discover a database, examine the database's documentation, find a related software package, and then locate a dataset associated with the software.

This creates a network of discovery:

Publication → Database → Dataset → Software → Documentation → Related publication

The scientific web is interconnected. A resource may be discovered not because the researcher searched for its exact name, but because another resource points toward it.
Europe PMC, for example, exposes links between publications, citations and external resources, helping users move between related research information.
This is one reason scientific websites should make their relationships visible. A database should identify the publication describing it. Software should identify relevant documentation and publications. Datasets should identify associated studies where appropriate. These connections create additional paths to discovery.

Researchers do not simply ask:
“Can I find the information?” They also ask: “Can I trust it?”

A search result is only a candidate. Before using the info, researchers may examine its documentation, authorship, institutional affiliation, citations, update history, version information and the methodology.
For example, a scientist comparing two bioinformatics tools may ask:

• Who publish the data?
• Is the methodology documented?
• Is the project maintained?
• Is there a publication describing the method?
• Are different piblications related to this one?
• Can the be trusted?

Researchers do not search for every type of resource in the same way.

• A scientist looking for a journal article may search by topic, author, DOI, journal, publication date or MeSH term.
• A researcher looking for a protein may use a gene symbol, protein name, accession number, organism, function or sequence.
• Someone seeking research software may search by method, analysis task, programming language, data type or biological application.
• A laboratory manager looking for a specialized service may search by technique, sample type, geographic location or analytical capability.
• A scientist looking for a dataset may search by disease, organism, experimental technique, accession number or publication.

The search vocabulary therefore changes depending on the object being searched for.

Understanding researcher behavior leads to several practical conclusions.

1. First, describe the resource in scientific language. Say what it does, what field it serves and what problem it solves.
2. Second, use terminology researchers actually use, including recognized synonyms and abbreviations where appropriate.
3. Third, identify important entities. When a page concerns a gene, protein, compound, organism, or publication, provide the relevant identifier whenever possible.
4. Fourth, make the purpose of the website immediately clear. A researcher should not have to spend several minutes discovering whether a website offers software, data, a service, publications or a biological data.
5. Fifth, connect related resources. Link software or product to documentation, studies, publications and molecules to their scientific targets.
6. Sixth, make the information easy to evaluate. Include authorship, institutional affiliation, references, version information, dates and documentation where it is relevant in a same way as a publication.

Finally, think beyond Search Engines. Scientific discovery happens across many information systems. PubMed, Europe PMC, UniProt, repositories, specialist databases and scientific directories each serve different purposes.

A link discoverability is therefore not determined by its position in a single search engine.