Persistent Identifiers in Bioinformatics: Selection Criteria and Implementation Workflow

Bioinformatics researchers generate datasets, software, pipelines, and analytical workflows that must remain findable and citable over long time horizons. Persistent identifiers (PIDs) provide stable references that continue to resolve even when files move between servers or institutions change their domain names. This article compares the main PID types used in bioinformatics, including DOIs, ORCIDs, and Handle-based identifiers, and presents a decision framework for matching identifier types to specific research outputs. An implementation workflow covers registration, metadata curation, quality checks, and ongoing maintenance.

The Role of Persistent Identifiers in Reproducible Bioinformatics

Reproducibility in bioinformatics depends on the ability to locate the exact data, code, and computational environment that produced a given result. A published analysis that references a dataset by URL alone becomes untraceable when that URL changes. Persistent identifiers solve this problem by separating the identifier from the physical location of the resource. The identifier resolves through a registry that can be updated to point to new locations as needed.

The FAIR Guiding Principles establish that research data should be findable, accessible, interoperable, and reusable [4]. Persistent identifiers are a core mechanism for achieving findability, since they provide globally unique and permanent references that can be indexed by search systems. The FAIR principles also emphasize that metadata should be registered or indexed in a searchable resource, which is what PID registries provide.

Large bioinformatics infrastructure providers have built their services around persistent identification. The European Bioinformatics Institute offers training and resources that emphasize the importance of stable identifiers for biological data management [1]. The National Center for Biotechnology Information maintains databases where records are assigned stable accession numbers that function as persistent identifiers for sequences, structures, and other biological entities [2].

Funding agencies increasingly require persistent identifiers as part of data sharing policies. The National Institutes of Health Genomic Data Sharing Policy specifies expectations for how genomic data should be shared and managed, including requirements that support the discoverability and accessibility of research outputs [3]. Researchers who plan to submit data under such policies need to understand which identifier systems meet the stated requirements.

Types of Persistent Identifiers Used in Bioinformatics

Digital Object Identifiers (DOIs)

DOIs are the most widely recognized persistent identifier type in scholarly communication. A DOI is a character string that uniquely identifies a digital object and resolves through the Handle System. DOIs are issued by registration agencies that operate under the International DOI Foundation.

DOIs are appropriate for datasets, software releases, published figures, and other research outputs that need formal citation. When a dataset is deposited in a repository such as Zenodo, Figshare, or a domain-specific repository, the repository typically assigns a DOI at the time of deposit. The DOI remains stable even if the underlying files are updated, provided the repository maintains the resolution link.

The Distributed System of Scientific Collections (DiSSCo) evaluated persistent identifier schemes for digital representations of physical specimens and selected a DOI-based approach customized with natural science community characteristics [20]. The evaluation weighed scalability, community trust, persistence, governance, and suitability for global adoption. This example demonstrates that DOI adoption extends beyond traditional publication contexts into specialized research infrastructure.

ORCID Identifiers

ORCID provides persistent identifiers for individual researchers. An ORCID iD distinguishes a researcher from others with similar names and links that researcher to their publications, datasets, and other outputs across different systems.

ORCID identifiers are not substitutes for DOIs. They identify people instead of objects. In a bioinformatics workflow, a researcher registers an ORCID iD and then associates that iD with datasets, software, and publications that carry their own DOIs. The ORCID record functions as a profile that aggregates a researcher's outputs.

The INGV Data Registry, a curated metadata infrastructure for Earth Science data, integrated ORCIDs alongside other persistent identifiers to ensure high-quality, interoperable metadata [19]. The registry assigned DOIs to hundreds of records and mapped them to international standards, with ORCIDs providing researcher identification within the metadata structure.

Handle System Identifiers

The Handle System is the underlying technology that supports DOI resolution. Handles can also be issued independently of the DOI system for specialized use cases. Handles provide persistent resolution through a distributed registry.

Some research communities choose Handle-based identifiers when they need control over the identifier namespace or when their use case requires characteristics that DOI registration agencies do not provide. The DiSSCo technical experts initially selected the Handle System to meet core PID requirements before later determining that a DOI-based scheme would better serve their community [20].

Accession Numbers and Database-Specific Identifiers

Biological databases maintain their own identifier systems. NCBI accession numbers identify sequences, genes, proteins, and other entities within NCBI databases [2]. These identifiers are stable within the database and are widely cited in the literature.

Database-specific identifiers differ from DOIs in important ways. They are managed by the database instead of by a DOI registration agency. They may not resolve through the Handle System. They are appropriate for entities that live within a specific database, but they do not provide a general-purpose citation mechanism for arbitrary research outputs.

Compact Identifiers

Compact identifiers provide a uniform resolution mechanism for biomedical data by combining a namespace prefix with a local identifier. The namespace identifies the database or resource, and the local identifier identifies the specific record within that resource. Resolution services can then direct users to the appropriate database.

Uniform resolution of compact identifiers for biomedical data has been addressed as a mechanism for improving interoperability across databases [27]. Compact identifiers are useful when a workflow needs to reference records across multiple databases without requiring each database to adopt a common identifier scheme.

Decision Framework for Selecting Persistent Identifiers

The choice of persistent identifier depends on the type of research output, the expected lifespan of the resource, the requirements of funders and journals, and the infrastructure available to the research group.

Identifier Selection by Output Type

Criteria for Identifier Choice

The first criterion is the expected persistence horizon. Research outputs that must remain citable for decades, such as datasets supporting published findings, warrant DOIs because the DOI system has established governance and funding models. Short-lived intermediate files may not need formal persistent identifiers.

The second criterion is the citation requirement. Journals and funders increasingly require DOIs for datasets and software cited in publications. If a manuscript will cite a dataset, that dataset needs a DOI. If a dataset is internal to a project and will never be cited, a DOI may be unnecessary overhead.

The third criterion is the community standard. Some research communities have established identifier conventions. Genomics researchers commonly cite NCBI accession numbers [2]. Structural biologists cite PDB identifiers. Researchers should follow the conventions of their primary community while adding DOIs where citation is required.

The fourth criterion is infrastructure capacity. Registering DOIs requires access to a repository or registration agency. Researchers without institutional repository access can use general-purpose repositories that assign DOIs at deposit. ORCID registration is free and open to all researchers.

Implementation Workflow for Persistent Identifiers

Step 1: Inventory Research Outputs

Begin by listing the research outputs that need persistent identifiers. Include datasets, software, pipelines, analysis scripts, container images, and published articles. For each output, record the current storage location, the expected lifespan, and the citation requirements.

A bioinformatics project might produce raw sequencing data, processed count matrices, differential expression results, analysis scripts, and a software package. Each of these outputs has different identifier needs. Raw sequencing data may be deposited in NCBI with accession numbers [2]. Processed data may receive DOIs through a repository. The software package may receive a DOI for each versioned release.

Step 2: Select Identifier Types

Apply the decision framework to each output. Assign DOIs to outputs that will be cited in publications or shared with external collaborators. Assign ORCID iDs to all researchers who will be associated with the outputs. Use database accession numbers for records deposited in NCBI or other biological databases.

Step 3: Register Identifiers

Register ORCID iDs first, since these will be associated with all other identifiers. ORCID registration is completed through the ORCID website and requires basic researcher information.

For DOIs, deposit the output in a repository that assigns DOIs. The repository will require metadata including title, creators, description, and license information. Prepare this metadata before deposit to avoid delays.

For database accession numbers, submit the data to the appropriate database following its submission procedures. NCBI provides submission portals for sequences, genomes, and other data types [2].

Step 4: Link Identifiers in Metadata

The value of persistent identifiers increases when they are linked. Associate ORCID iDs with dataset DOIs in repository metadata. Include accession numbers in dataset descriptions. Reference dataset DOIs in software documentation.

The INGV Data Registry demonstrates this approach by integrating ORCIDs and RORs within a metadata infrastructure that assigns DOIs to records [19]. The registry decouples data description from physical storage, allowing metadata to be curated independently of the underlying data files.

Step 5: Verify Resolution

After registration, test that each identifier resolves correctly. Enter the identifier into a resolution service or browser and confirm that it reaches the intended resource. Check that the landing page displays the correct metadata and provides access to the underlying data or software.

Resolution failures can occur when metadata is incomplete, when the repository has technical issues, or when the identifier was registered incorrectly. Document any failures and correct them before sharing the identifiers in publications or reports.

Step 6: Maintain the Identifier Registry

Persistent identifiers require ongoing maintenance. When files move to new storage locations, update the resolution links in the repository. When researchers change institutions, update their ORCID records. When datasets are revised, deposit new versions and link them to the original identifiers.

The INGV Data Registry operates as a dynamic ecosystem that manages hundreds of records, each assigned a DOI and mapped to international standards [19]. This model demonstrates that identifier maintenance is an ongoing activity instead of a one-time registration event.

Metadata Requirements for Persistent Identifiers

Core Metadata Fields

Persistent identifiers are only useful when accompanied by sufficient metadata. The minimum metadata for a dataset DOI includes title, creators, publication date, publisher, and resource type. Additional fields such as description, license, and related identifiers improve discoverability and reusability.

The FAIR Guiding Principles specify that metadata should be richly described with a plurality of accurate attributes [4]. This means that minimal metadata is insufficient for full FAIR compliance. Researchers should provide as much contextual information as possible when registering identifiers.

Community-Specific Metadata

Bioinformatics datasets often require domain-specific metadata. Genomic datasets need information about the organism, sequencing platform, assembly version, and annotation. Proteomic datasets need information about the mass spectrometry platform, search parameters, and protein database version.

The INGV Data Registry maps records to international standards to ensure interoperability [19]. Bioinformatics researchers should similarly map their metadata to community standards such as those maintained by the Genomics Standards Consortium or the Proteomics Standards Initiative.

Metadata Quality Control

Metadata errors undermine the value of persistent identifiers. A dataset with an incorrect creator name or a missing license may be difficult to discover or reuse. Implement quality checks before registering identifiers.

The INGV Data Registry uses a three-tiered validation workflow to ensure high-quality, interoperable metadata [19]. This workflow includes automated validation against schema requirements, manual review by curators, and verification of cross-references between records.

Common Failure Patterns in Persistent Identifier Implementation

Identifier Proliferation Without Coordination

Research groups often register identifiers in multiple systems without linking them. A dataset may have a DOI in one repository, an accession number in a database, and a mention in a publication, with no connections between these references. This fragmentation makes it difficult for readers to navigate from one identifier to related resources.

The solution is to include related identifiers in metadata. When depositing a dataset, reference the associated accession numbers and publication DOIs. When publishing an article, include dataset DOIs and accession numbers in the methods section.

Resolution Failures After Repository Migration

Repositories sometimes change platforms or merge with other services. If the repository does not maintain its DOI resolution links, previously registered identifiers may stop resolving. This failure mode is particularly common when institutions migrate their repositories without preserving the DOI prefix mappings.

Researchers can mitigate this risk by choosing repositories with established track records and by testing identifier resolution periodically. The DiSSCo evaluation noted that persistent identifiers should transcend changes in underlying technologies [20]. This requirement applies to the repository infrastructure as much as to the identifier scheme.

Metadata Drift

Metadata can become outdated as projects evolve. A dataset deposited with one set of creators may later involve additional researchers. A software package may change its license. If the metadata is not updated, the persistent identifier points to inaccurate information.

Establish a schedule for reviewing and updating metadata. Assign responsibility for metadata maintenance to a specific team member. Document any changes to the metadata so that the history of the identifier is transparent.

Version Confusion

Datasets and software are often updated after initial release. If versioning is not handled carefully, readers may not know which version corresponds to a cited result. Some repositories assign new DOIs for each version, while others maintain a single DOI with version metadata.

The choice between these approaches depends on the citation requirements. If a publication cites a specific version, that version needs its own DOI. If the citation refers to the dataset as a whole, a single DOI with version information may suffice.

Observations and Measurements for Identifier Management

Tracking Identifier Coverage

Maintain a registry of all research outputs and their identifier status. For each output, record whether it has a DOI, an accession number, an ORCID association, and a verified resolution link. Review this registry quarterly to identify gaps.

A spreadsheet or database table with columns for output name, output type, storage location, DOI, accession number, ORCID iD, resolution status, and last verification date provides a practical tracking mechanism.

Measuring Resolution Success

Periodically test a sample of registered identifiers to confirm that they resolve correctly. Record the resolution success rate and investigate any failures. A resolution success rate below 95 percent indicates a maintenance problem that requires attention.

Resolution testing can be automated using scripts that query the resolution services and check the HTTP status codes. Manual testing is appropriate for small numbers of identifiers or when automated testing is not feasible.

Documenting Identifier Usage

Track how identifiers are used in publications, reports, and data sharing activities. This documentation demonstrates the value of the identifier infrastructure and supports reporting to funders who require evidence of data sharing.

The National Institutes of Health Genomic Data Sharing Policy expects researchers to document how genomic data are shared and accessed [3]. Identifier usage records provide this documentation.

Quality Controls for Persistent Identifier Implementation

Pre-Registration Checklist

Before registering an identifier, verify that the output is complete, the metadata is accurate, and the storage location is stable. Check that all files are included and that the data can be accessed without errors.

For datasets, verify that the data files open correctly and that the file formats are documented. For software, verify that the code runs and that the dependencies are specified. For databases, verify that the submission meets the database requirements.

Post-Registration Verification

After registration, confirm that the identifier resolves and that the landing page displays the correct metadata. Test the access links to ensure that the underlying files are retrievable.

The INGV Data Registry's three-tiered validation workflow provides a model for post-registration verification [19]. The workflow ensures that records meet schema requirements, that metadata is complete, and that cross-references are valid.

Ongoing Monitoring

Schedule regular checks of identifier resolution and metadata accuracy. Update metadata when project details change. Re-verify identifiers after any repository migration or infrastructure change.

The DiSSCo evaluation emphasized that persistent identifiers must remain valid over very long timescales and transcend changes in underlying technologies [20]. Ongoing monitoring is the mechanism for ensuring this persistence.

Limitations of Persistent Identifiers

Identifiers Do Not Guarantee Access

A persistent identifier ensures that a resource can be found, but it does not guarantee that the resource remains accessible. If the underlying data files are deleted or the repository goes out of business, the identifier may resolve to an error page or a tombstone record.

Researchers should choose repositories with preservation commitments and should maintain local copies of important data. The identifier provides a stable reference, but the data stewardship responsibilities remain with the research team.

Identifier Systems Have Governance Requirements

DOI registration requires adherence to the policies of the registration agency. These policies may include metadata requirements, fee structures, and usage restrictions. Researchers who need to register DOIs must work within these governance frameworks.

The DiSSCo evaluation considered governance as one of the assessment dimensions for PID schemes [20]. The choice of identifier system should account for the governance burden and the long-term sustainability of the system.

Database Accession Numbers Are Database-Specific

Accession numbers from NCBI and other databases are stable within their databases, but they do not resolve through general-purpose identifier systems. A reader who encounters an accession number must know which database to query.

This limitation can be addressed by including database names alongside accession numbers and by using compact identifiers that encode the database namespace [27].

Identifier Adoption Requires Community Participation

Persistent identifiers provide value when they are used consistently across the research community. If some researchers do not register identifiers for their outputs, the overall discoverability of research is reduced.

Individual researchers can encourage adoption by requiring identifiers for outputs they cite and by promoting identifier use within their collaborations.

Safety and Regulatory Context

Data Sharing Policy Compliance

Funding agencies and journals increasingly require persistent identifiers as part of data sharing policies. The National Institutes of Health Genomic Data Sharing Policy establishes expectations for genomic data sharing that include mechanisms for data discoverability and access [3].

Researchers who receive funding under such policies should verify that their identifier implementation meets the policy requirements. This verification should occur before data generation begins, since retrofitting identifiers to existing data is more difficult than planning for them in advance.

Sensitive Data Considerations

Persistent identifiers make data more discoverable, which may conflict with privacy or security requirements for sensitive data. Researchers working with human genomic data must balance the benefits of persistent identification against the risks of unauthorized access.

The Genomic Data Sharing Policy addresses this balance by specifying conditions for data access [3]. Researchers should consult their institutional review boards and data access committees when planning identifier implementation for sensitive data.

Professional Escalation Criteria

Researchers should escalate identifier issues to institutional data managers or librarians when they encounter problems that exceed their technical expertise. Escalation is appropriate when a repository fails to assign identifiers, when resolution failures persist after troubleshooting, or when funder requirements are unclear.

Institutional data services can provide guidance on repository selection, metadata standards, and compliance requirements. Bioinformatics core facilities can assist with technical implementation and integration with existing workflows.

Frequently Asked Questions

What is the difference between a DOI and an ORCID iD?

A DOI identifies a digital object such as a dataset, software release, or publication. An ORCID iD identifies a person. A researcher registers an ORCID iD and then associates that iD with the DOIs of their outputs. The two identifier types serve complementary purposes in the research ecosystem.

When should I register a DOI for a dataset?

Register a DOI when the dataset will be cited in a publication, shared with external collaborators, or required to be accessible by a funder or journal. Datasets that are internal to a project and will never be cited may not need DOIs, but registering them is often good practice for transparency and future reuse.

How do I choose between a DOI and a database accession number?

Use a database accession number when the data lives in a specific biological database such as NCBI [2]. Use a DOI when the data needs a general-purpose citation mechanism or when the data is not appropriate for a domain-specific database. Many datasets have both an accession number and a DOI, with the DOI providing the citation link and the accession number providing the database record.

What metadata should I include when registering a persistent identifier?

Include at minimum the title, creators, publication date, publisher, and resource type. Add a description, license, and related identifiers to improve discoverability and reusability. The FAIR Guiding Principles specify that metadata should be richly described with a plurality of accurate attributes [4].

How often should I verify that my persistent identifiers resolve?

Verify resolution immediately after registration and then on a regular schedule, such as quarterly or annually. Re-verify after any repository migration or infrastructure change. Automated resolution testing can reduce the burden of manual verification.

Can I change the metadata associated with a persistent identifier after registration?

Most identifier systems allow metadata updates after registration. The identifier itself remains stable, but the associated metadata can be corrected or enhanced. Document any metadata changes to maintain transparency about the identifier's history.

What happens if a repository that assigned a DOI goes out of business?

The DOI should continue to resolve if the registration agency maintains the resolution link. However, the underlying data may become inaccessible if the repository does not transfer its holdings to another institution. Choose repositories with preservation commitments and maintain local copies of important data.

Do persistent identifiers work for software and pipelines?

Yes. Software releases can receive DOIs through repositories such as Zenodo or Figshare. Versioned releases should receive separate DOIs so that publications can cite the specific version used. Container images can also receive identifiers through container registries that support persistent identification.

Related Bioinformatics Guides

• Data Sharing and Privacy in Genomic Research
• Orchestrating Bioinformatics at Scale: Workflows, Containers, and Cloud Infrastructures
• Workflow Management: Snakemake vs. Nextflow: Architectural Comparisons and Workflow Design Rules
• The Protein Data Bank (PDB): Archival Standards, Structural Validation Metrics, and Bioinformatics Integration Protocols
• The History and Evolution of Bioinformatics

References and Further Reading

• EMBL-EBI Training. European Bioinformatics Institute.
• NCBI Data Resources. National Center for Biotechnology Information.
• Genomic Data Sharing Policy. National Institutes of Health.
• The FAIR Guiding Principles. Scientific Data.
• Genomic findings in schizophrenia and their implications.. Molecular psychiatry, 2023.
• Decoding per- and polyfluoroalkyl substances (PFAS) in hepatocellular carcinoma: a multi-omics and computational toxicology approach.. Journal of translational medicine, 2025.
• Identifying proteomic risk factors for cancer using prospective and exome analyses of 1463 circulating proteins and risk of 19 cancers in the UK Biobank.. Nature communications, 2024.
• Hepatic stellate cells suppress NK cell-sustained breast cancer dormancy.. Nature, 2021.
• Organoid single-cell genomic atlas uncovers human-specific features of brain development.. Nature, 2019.
• KDELR3 and YOD1 proteins as critical endoplasmic reticulum stress mediators and potential therapeutic targets in diabetic foot ulcers: An integrated bioinformatics analysis.. International journal of biological macromolecules, 2025.
• Multivariate analysis of 1.5 million people identifies genetic associations with traits related to self-regulation and addiction.. Nature neuroscience, 2021.
• Conservation Genomics-Making a Difference.. Molecular ecology, 2025.
• Screening for hypertension in a public dental clinic: A single-centre cross-sectional study in Australia.. 2026.
• Profiles of social-emotional competence in elementary students: A latent profile analysis.. 2026.
• Soluble ST2 as a biomarker for predicting severe adverse events among pediatric patients with Mycoplasma pneumoniae pneumonia.. 2026.
• Topological data analysis for predicting disease outbreaks in humanitarian settings: A machine learning approach.. 2026.
• Risk factors associated with diabetic foot amputation: A retrospective study from a tertiary hospital in Central Malaysia.. 2026.
• Dressing-induced hemodynamic instability in patients with heart failure: Implications for nursing care.. 2026.
• The INGV data registry as a curated metadata infrastructure for Earth Science data stewardship. Scientific Data, 2026.
• A choice of persistent identifier schemes for the Distributed System of Scientific Collections (DiSSCo). Research Ideas and Outcomes, 2021.
• Exploring the implementation of a novel community pharmacist-led support service for people living with severe and persistent mental illness.. Research in Social and Administrative Pharmacy, 2024.
• Chipless RFID Tag Implementation and Machine-Learning Workflow for Robust Identification. IEEE transactions on microwave theory and techniques, 2023.
• Enhancing Safety in AI-Driven Cone Beam CT-based Online Adaptive Radiation Therapy: Development and Implementation of an Interdisciplinary Workflow. Advances in Radiation Oncology, 2023.
• Persistent AI Agents in Academic Research: A Single-Investigator Implementation Case Study. arXiv.org, 2026.
• Implementation and workflow strategies for integrating digital therapeutics for alcohol use disorders into primary care: a qualitative study. Addiction Science & Clinical Practice, 2023.
• Implementation of single IRB review for multisite human subjects research: Persistent challenges and possible solutions. Journal of Clinical and Translational Science, 2023.
• Uniform resolution of compact identifiers for biomedical data. Scientific Data, 2018.
• Designing PIDs for Reproducible Science Using Time-Series Data. Communications in Computer and Information Science, 2023.
• The e!DAL JAVA-API: Store, share and cite primary data in life sciences. Proceedings 2012 IEEE International Conference on Bioinformatics and Biomedicine Bibm 2012, 2012.
• CellMiner: A relational database and query tool for the NCI-60 cancer cell lines. BMC Genomics, 2009.
• An integrated web-based interactive data platform for molecular dynamics simulations. Scalable Computing, 2018.

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.