Search Strategy Development

Published

Jun 2026

Why Search Strategy Matters

A systematic dataset discovery workflow requires a clear search strategy.

The search strategy defines how candidate public studies will be identified across literature databases, project repositories, sequence archives, and linked records.

Without a documented search strategy, dataset discovery can become inconsistent and difficult to reproduce. Different analysts may search different sources, use different keywords, apply different assumptions, and produce different candidate study lists.

A structured search strategy helps make the discovery process transparent.

It records:

what was searched
where it was searched
which terms were used
when the search was performed
what types of records were expected
how candidate studies were captured
how records were prepared for screening

The search strategy does not decide which studies are eligible. That happens later during eligibility assessment and screening.

The purpose of the search strategy is to identify candidate records in a reproducible way.

From Research Question to Search Concepts

A search strategy begins by translating the research question into searchable concepts.

For the case study in this guide, the discovery question is:

Which public omics studies contain healthy human gut microbiome sequencing data suitable for reference dataset assembly?

This question can be broken into major concepts:

Concept	Meaning
Human	Studies involving human participants or samples
Gut microbiome	Studies involving gut, stool, fecal, or intestinal microbiome samples
Healthy	Studies involving healthy, non-diseased, control, or baseline participants
Sequencing data	Studies with public omics sequencing accessions
Reference dataset assembly	Studies suitable for reuse and downstream CDI-DAS acquisition

Each concept can then be expanded into search terms.

Search Term Expansion

Public records may use different words for similar ideas.

For example, a study may refer to “gut microbiome,” “intestinal microbiota,” “fecal microbiome,” or “stool microbial communities.” A strict search for only one phrase may miss relevant studies.

Search term expansion helps improve recall.

For the healthy human gut microbiome case study, possible terms include:

Concept	Example Search Terms
Human	human, Homo sapiens, adult, participant, subject
Gut	gut, intestinal, gastrointestinal, stool, fecal, faecal
Microbiome	microbiome, microbiota, microbial community, metagenome
Healthy	healthy, control, non-diseased, baseline, normal
Sequencing	sequencing, 16S, metagenomic, amplicon, Illumina, MiSeq
Public data	BioProject, SRA, ENA, accession, public dataset

Search terms should be broad enough to find relevant studies but focused enough to avoid overwhelming irrelevant results.

Combining Search Concepts

Search concepts can be combined using Boolean logic.

Common Boolean operators include:

Operator	Use
AND	Requires both concepts to be present
OR	Allows alternative terms for the same concept
NOT	Excludes unwanted concepts

For example:

human AND gut AND microbiome

is narrower than:

human AND (gut OR stool OR fecal) AND (microbiome OR microbiota)

A more structured search expression may look like:

(human OR "Homo sapiens")
AND
(gut OR intestinal OR stool OR fecal OR faecal)
AND
(microbiome OR microbiota OR metagenome)
AND
(healthy OR control OR baseline)

The exact syntax may differ by source. PubMed, NCBI BioProject, SRA, and ENA may support different search fields and search behavior.

Source-Specific Search Strategy

A single search string may not work equally well across all sources.

Each source should be searched according to its role.

Source	Search Strategy Role
PubMed	Find candidate publications and study context
NCBI BioProject	Find project-level accessions
NCBI SRA	Validate sequencing runs and technical metadata
NCBI BioSample	Check sample-level metadata
ENA	Validate FASTQ links, layout, and checksums
Supplementary materials	Confirm sample annotations and accession mapping

This means the search strategy should include both broad discovery searches and targeted validation searches.

PubMed Search Strategy

PubMed is useful for identifying candidate publications.

A PubMed search may combine biological context, sample type, and sequencing terms.

Example:

(human[Title/Abstract] OR "Homo sapiens"[Title/Abstract])
AND
(gut[Title/Abstract] OR intestinal[Title/Abstract] OR stool[Title/Abstract] OR fecal[Title/Abstract] OR faecal[Title/Abstract])
AND
(microbiome[Title/Abstract] OR microbiota[Title/Abstract] OR metagenome[Title/Abstract])
AND
(healthy[Title/Abstract] OR control[Title/Abstract] OR baseline[Title/Abstract])

This search is used to identify publications that may describe eligible public datasets.

Publication records should then be checked for repository accessions, data availability statements, and supplementary materials.

NCBI BioProject Search Strategy

NCBI BioProject is useful for finding project-level public dataset records.

A BioProject search may use a simpler query because repository records may contain shorter descriptions than publications.

Example:

human gut microbiome healthy

or:

human stool microbiome

BioProject search results should be reviewed for relevance and accession information.

For the case study, the prioritized BioProject is:

PRJNA802976

This BioProject becomes the main case-study accession for downstream validation and CDI-DAS handoff.

SRA Search Strategy

SRA is useful for validating whether a candidate study contains sequencing runs suitable for acquisition.

For a known BioProject, SRA can be queried using the BioProject accession.

Example:

PRJNA802976[BioProject]

The expected output is a set of run accessions and technical metadata.

For the prioritized case study, the primary test subset is:

SRR17868090
SRR17868091
SRR17868092

These accessions provide a small test set for validating the downstream acquisition workflow before scaling to the full BioProject.

ENA Search Strategy

ENA is useful for validating file-level availability.

For a candidate BioProject or run accession list, ENA can be used to check:

FASTQ file links
file sizes
checksums
library layout
sequencing platform
study and sample accessions

The ENA validation step helps determine whether a candidate study can be acquired efficiently through direct FASTQ download.

For CDI-DAS, ENA is especially useful when FASTQ links and checksums are available.

Recording and Running the Search Strategy

Every search strategy should be documented before candidate studies are screened.

At minimum, the workflow should record:

source searched
search date
search query
search purpose
expected output
notes
number of records found, when available

A search log allows the discovery workflow to be reviewed and repeated.

The main search record for this chapter is:

outputs/search-strategy.tsv

This file becomes part of the discovery audit trail.

The search workflow is implemented through three scripts:

bash scripts/bash/04a-search-bioprojects.sh
bash scripts/bash/04b-search-literature.sh
bash scripts/bash/04c-create-candidate-studies-template.sh

The expected outputs are:

outputs/search-strategy.tsv
outputs/bioproject-search-results.tsv
outputs/literature-search-results.tsv
outputs/candidate-studies.tsv

The BioProject search identifies candidate project-level accessions.

The literature search identifies candidate publications, study context, data availability statements, and possible accession-linked records.

The candidate study template provides a consistent table structure for recording and screening candidate studies.

Together, these outputs provide the starting evidence base for dataset screening.

Search Strategy
        ↓
BioProject Search
        ↓
Literature Search
        ↓
Candidate Study Template
        ↓
Candidate Study Evidence
        ↓
Dataset Screening

Candidate Study Capture

Search results should be captured into a candidate study table.

A candidate study table may include:

candidate ID
source
title or record name
accession
publication identifier
organism
sample context
data type
repository link or accession
notes
screening status

The candidate study table is not yet the included study list.

It is the working list of records that will be screened in later chapters.

The expected output is:

outputs/candidate-studies.tsv

The candidate study template is generated as part of the Chapter 04 script sequence:

bash scripts/bash/04c-create-candidate-studies-template.sh

This template provides a consistent structure for recording candidate studies identified during the BioProject and literature searches.

Search Strategy for the Case Study

For the healthy human gut microbiome case study, the search strategy focuses on identifying public studies that contain human gut or stool microbiome sequencing data with enough metadata and accession clarity for reuse.

The prioritized case-study accession is:

PRJNA802976

The primary test subset is:

SRR17868090
SRR17868091
SRR17868092

A secondary comparison BioProject may be retained for technical contrast:

PRJNA322554

This secondary accession can help demonstrate why sequencing layout, platform, metadata completeness, and acquisition readiness matter during prioritization.

Search Strategy Boundaries

The search strategy should also define boundaries.

For this guide, the initial search focuses on:

human samples
gut, stool, fecal, or intestinal microbiome context
public sequence data
accession-linked studies
records that can support CDI-DAS handoff

The search does not aim to capture every possible microbiome publication. It focuses on studies that can move from discovery into acquisition and validation.

Avoiding Overly Narrow Searches

A search that is too narrow may miss relevant datasets.

For example, searching only for:

healthy human gut microbiome

may miss records described as:

control stool microbiota

or:

intestinal microbial communities in adults

This is why search term expansion is important.

A broad initial search can be followed by structured screening.

Avoiding Overly Broad Searches

A search that is too broad can produce too many irrelevant records.

For example:

microbiome

may return studies from soil, plants, animals, oral samples, disease cohorts, and environmental samples.

The search strategy should balance sensitivity and specificity.

For systematic discovery, the goal is not simply to maximize search results. The goal is to identify a manageable and relevant candidate set that can be screened transparently.

Updating the Search Strategy

Search strategies may be refined during early exploration.

For example, the analyst may discover that relevant records use “fecal microbiota” more often than “gut microbiome,” or that a repository uses different metadata fields than expected.

When changes are made, they should be documented.

A reproducible workflow should preserve:

original search terms
revised search terms
reason for revision
date of revision
effect on candidate records, when known

This prevents silent changes from weakening the audit trail.

Summary

Search strategy development translates the research question into a reproducible discovery plan.

It defines the concepts, search terms, sources, queries, and candidate capture process used to identify public omics studies.

For the healthy human gut microbiome case study, the search strategy prioritizes accession-linked public studies that can be screened, evaluated, and handed off to the CDI Data Acquisition System.

The prioritized case-study BioProject is PRJNA802976, with SRR17868090, SRR17868091, and SRR17868092 used as the primary test subset.

Looking Ahead

In the next chapter, we define eligibility criteria for deciding which candidate studies should be included, excluded, or set aside for later review.