Skip to main content
v2026.11,986 entries · CC-BY 4.0

PHASTER: Prophage Identification, API Workflows, and Genome Analysis

A practical guide to PHASTER for identifying prophage regions in bacterial genomes, running web and API submissions, and interpreting completeness scores.

Written and maintained by CASRAI Editorial Board

Published · Last updated

On this page

PHASTER (PHAge Search Tool – Enhanced Release) is a high-throughput web server and bioinformatic computing pipeline designed to identify, annotate, and visualize prophage sequences integrated within bacterial genomes and plasmids. By integrating rapid open reading frame (ORF) prediction, sequence similarity searches against curated viral databases, and spatial clustering algorithms, the platform detects integrated viral elements and scores their structural completeness in minutes.

What it is used for in research

Bacteriophages are viruses that infect bacteria, and they exist primarily through two life cycles: lytic and temperate. While lytic phages replicate immediately and destroy the host cell, temperate phages integrate their double-stranded DNA into the bacterial replicon, residing passively as prophages. In bacterial genomics, prophage sequences represent a major driver of intraspecies genetic variation, accounting for upwards of 10% to 20% of a bacterial genome’s total sequence content. These integrated elements frequently serve as vehicles for horizontal gene transfer, carrying functional accessory cargo such as antibiotic resistance genes, virulence determinants, exotoxins, and metabolic modifications that enhance host fitness in clinical and environmental niches.

In computational microbiology and genomic epidemiology, researchers face the challenge of rapidly distinguishing native bacterial sequence from integrated viral elements. Unannotated prophages confound genome assembly pipelines, skew phylogenetic trees, and obscure true bacterial core gene sets. PHASTER provides an automated, standardized pipeline that addresses these challenges by scanning whole-genome sequences, contig sets, and plasmids to locate prophage boundaries, classify predicted prophages by completeness, and identify the closest known viral relatives. When researchers analyze draft assemblies after evaluating assembly contiguity with QUAST or assembling metagenomic reads with tools like MEGAHIT, detecting prophages helps differentiate mobile viral cargo from chromosomal housekeeping genes.

Beyond baseline genomic characterization, PHASTER is employed in several specialized research settings:

  • Antimicrobial Resistance and Virulence Surveillance: Tracking whether critical resistance genes or Shiga toxins (such as stx1 and stx2 in Escherichia coli) reside within intact, mobilizable prophages or degraded pseudogenes.
  • Phage Therapy and Biotechnology: Identifying inducible, functional prophages within therapeutic host strains to prevent unintentional lysogen induction or horizontal spread during industrial fermentation. This bioinformatic screening is fundamentally distinct from experimental phage display libraries used for protein-protein interaction screening.
  • Comparative Plasmid and Chromosomal Genomics: Determining whether mobile genetic elements reside on extrachromosomal elements, resolving differences when comparing plasmid DNA vs genomic DNA in multidrug-resistant clinical isolates.
  • Evolutionary Genomics: Tracking the evolutionary decay of ancient viral integrations, distinguishing intact, replication-competent phages from degenerated phage remnants and cryptic prophage footprints.

How to install or access it

PHASTER operates primarily as a centralized high-performance computing web server and remote REST API hosted by the Wishart Node of the Metabolomics Innovation Centre at the University of Alberta. Because PHASTER relies on an extensive compute backend running on a dedicated 112-CPU cluster with multi-gigabyte reference databases, it does not distribute a standalone single-binary desktop installer. Instead, official access is provided through three supported channels:

  • Interactive Web Interface: Accessible through any modern web browser at the official domain (http://phaster.ca/ or the updated project gateway). The front-end interface, developed with Ruby on Rails, Angular, and D3 visualization frameworks, supports interactive file uploads, real-time job monitoring, and graphical genome inspection.
  • Remote URL REST API: Available at the programmatic endpoint (http://phaster.ca/phaster_api). The API accepts automated multipart file uploads and URL parameters using tools such as curl, wget, or Python scripts, allowing researchers to incorporate PHASTER directly into high-throughput bioinformatics workflows managed by environments such as Mamba vs Conda.
  • Next-Generation Server and Containerized Successor (PHASTEST): For laboratories requiring modern containerized local execution, the Wishart research group developed PHASTEST (http://phastest.ca/), which provides local Docker containers and an ultra-fast MMseqs2 alignment engine while maintaining backward compatibility with PHASTER query parameters.

No account registration or commercial licensing fee is required. The web server and REST API are freely available for academic and non-commercial research use, funded by the Canadian Institutes of Health Research (CIHR) and Genome Alberta. To preserve shared server stability, the maintainers implement request throttling and ask that automated pipelines submit one job at a time rather than launching massive concurrent batches.

Step-by-step instructions

Submitting genomic sequences to PHASTER can be accomplished either through the official web browser graphical interface or programmatically via the command-line REST API. Both workflows require correctly formatted genomic files and appropriate parameter selection.

Task A: Interactive Web Submission

  1. Prepare Genomic Input: Ensure your sequence file is formatted as either FASTA format (raw nucleotide sequence) or pre-annotated GenBank format (.gbk or .gb). While raw FASTA sequences trigger automated de novo gene prediction, supplying an annotated GenBank file bypasses the initial gene-calling step, reducing server processing time by approximately 50%. The maximum recommended single-file size for web submission is 250 MB.
  2. Navigate to Submission Form: Open http://phaster.ca/ in your browser and locate the main submission interface panel.
  3. Select Input Mode: Choose one of the three input tabs:
    • Upload File: Click Choose File or Browse to select your local FASTA or GenBank file.
    • Enter Accession: Enter an official NCBI GenBank accession number (such as NC_002655) or GI number to pull the sequence directly from public repositories like Ensembl or NCBI.
    • Paste Sequence: Paste plain nucleotide sequence text directly into the text area.
  4. Configure Multi-Contig Assemblies: If submitting a draft genome consisting of multiple scaffolds or contigs (such as an unclosed de novo assembly or metagenomic bin), select the Metagenomic data or multi-contig checkbox. If this option is omitted on a multi-FASTA file, the server will process only the first 10 contigs individually rather than evaluating the complete contig collection.
  5. Enter Email Notification (Optional): Enter an email address to receive an alert when analysis finishes, or leave blank to rely on web browser session tracking.
  6. Submit Job: Click the Submit button. The browser will navigate to a unique tracking URL containing your job identifier (e.g., http://phaster.ca/jobs/ZZ_...). Bookmark this URL to retrieve your results.

Task B: Programmatic Command-Line API Submission

For automated pipelines, the PHASTER REST API provides programmatic submission and status polling using command-line HTTP clients.

  1. Submit Sequence via curl: Post your FASTA file to the API endpoint. For complete circular genomes, execute:
    curl -s -X POST -F "[email protected]" http://phaster.ca/phaster_api > submission.json

    For draft assemblies containing multiple contigs, append the contigs=1 parameter to ensure all scaffolds are evaluated:

    curl -s -X POST -F "[email protected]" -F "contigs=1" http://phaster.ca/phaster_api > submission.json

    The server returns a JSON response containing the assigned job identifier:

    {"job_id":"ZZ_1234567890","status":"Waiting"}
  2. Poll Job Status: Query the API periodically using the assigned job identifier until the status changes from Waiting or Running to Complete:
    curl -s "http://phaster.ca/phaster_api?id=ZZ_1234567890" > status.json

    While running, the API returns:

    {"job_id":"ZZ_1234567890","status":"Running"}
  3. Download Results: When the status report indicates Complete, the response includes download links for the complete output zip bundle and summary text files. Retrieve the zipped archive:
    curl -s -O "http://phaster.ca/phaster_api?id=ZZ_1234567890andformat=zip"

    Unpack the archive to inspect text and tabular output files.

Output and how to read it

PHASTER generates a comprehensive suite of textual, tabular, and graphical outputs designed to facilitate both rapid triage and granular prophage annotation. Understanding how the underlying algorithm scores prophage candidates is essential for correctly interpreting these reports.

The Algorithmic Workflow and Scoring Engine

When an unannotated FASTA sequence enters the pipeline, PHASTER executes a multi-stage bioinformatic workflow:

  1. Gene and Feature Calling: Open reading frames are called using GLIMMER 3.02. Concurrently, host tRNA genes and tmRNA elements are mapped using tRNAscan-SE and ARAGORN, establishing reference landmarks for common phage integration hotspots.
  2. Homology Search: Translated protein coding sequences (CDS) are searched against the custom PHASTER viral protein database (containing over 187,000 phage proteins) and non-redundant bacterial protein databases using BLAST+ 2.3.0+.
  3. DBSCAN Spatial Clustering: To identify prophage candidate regions, PHASTER groups proximal viral-like genes using the density-based spatial clustering of applications with noise (DBSCAN) algorithm. The algorithm enforces two primary empirical thresholds:
    • Minimum cluster size (n): Fixed at 6 genes, requiring at least 6 phage-like genes to initiate a cluster.
    • Maximum intergenic spatial distance (e): Set to 3,000 base pairs, defining the maximum nucleotide distance permitted between adjacent viral genes within a cluster.

    In pre-annotated GenBank files, an additional moving-window scan searches for clusters containing 6 or more phage-specific keywords (such as integrase, capsid, terminase, or tail fiber) across a window of 60 proteins.

  4. Attachment Site (att) Identification: If an integrase gene is identified within the cluster, the flanking genomic regions are scanned for short direct nucleotide repeats (12 to 80 base pairs) to delineate precise attL (left) and attR (right) insertion boundaries.

Completeness Scoring and Classification

Each identified prophage region is evaluated according to a standardized scoring algorithm that classifies the prophage into one of three confidence tiers:

  • Intact (Score > 90): High confidence that the region represents a complete, potentially functional or viable prophage. The region typically possesses a full complement of structural and functional modules, including head/capsid genes, tail proteins, DNA packaging machinery (terminase), replication regulation, and a complete lysis cassette (holin and endolysin).
  • Questionable (Score 70 to 90): Moderate confidence. The region contains several recognizable phage-like genes or homology matches but lacks key structural modules, exhibits significant gene loss, or is shorter than typical temperate phages (under 25–30 kb). These regions often represent defective prophages, evolutionary remnants undergoing mutational decay, or novel phages with distant homology.
  • Incomplete (Score < 70): Low confidence. The cluster consists of isolated phage fragments, individual transposases, or cryptic remnants containing only a few viral genes. They are incapable of independent excision or virion formation.

The mathematical score is calculated under three distinct operational scenarios:

  • Scenario A: The region contains all genes corresponding to a known reference phage. The region is automatically assigned a maximum completeness score of 150.
  • Scenario B: Greater than 50% of the proteins in the region match a specific known phage. The score is calculated as the sum of two normalized ratios: (number of bases in region / number of bases in related phage) × 100, plus (number of genes in region / number of genes in related phage) × 100.
  • Scenario C: Fewer than 50% of the proteins in the region match a known phage. The score begins at 0 and adds points incrementally:
    • +10 points if the total nucleotide length exceeds 30 kb.
    • +10 points if the region contains more than 40 predicted genes.
    • +10 points for each identified ‘cornerstone’ phage gene (including structural genes: capsid, head, plate, tail, coat, portal, holin; DNA regulation genes: integrase, transposase, terminase; and functional lysis genes: lysin, bacteriocin).
    • +10 points if phage-like proteins constitute 70% or more of the total genes in the region.

Key Output Files

The downloadable results bundle contains several primary files:

  • summary.txt: A tab-delimited overview listing each detected prophage region. Columns include:
    • REGION: Ordinal index of the detected prophage region (e.g., Region 1, Region 2).
    • REGION_LENGTH: Total span of the prophage in base pairs (bp).
    • COMPLETENESS: Quality classification (intact, questionable, or incomplete).
    • SCORE: Numeric completeness score calculated by the scoring engine.
    • SPECIFIC_KEYWORD: Phage-related keywords detected within the prophage boundary.
    • REGION_POSITION: Nucleotide start and end coordinates on the host chromosome or contig.
    • MOST_COMMON_PHAGE_NAME: Reference bacteriophage name displaying the highest sequence similarity across the region.
    • GC_PERCENTAGE: Overall GC content percentage of the prophage region, useful for identifying compositional divergence from host genomic GC baselines.
  • detail.txt: A comprehensive gene-by-gene tabular inventory. For every predicted CDS within the region, this file records protein start and stop coordinates, coding direction (strand +/-), functional category, closest BLAST hit accession, alignment E-value, and percentage identity.
  • predicted_att.txt: Direct nucleotide sequences and genomic coordinates for predicted attL and attR attachment sites and flanking tRNA integration markers.
  • prophage_sequences.ffn and .faa: Multi-FASTA files containing the nucleotide coding sequences and translated amino acid sequences of all predicted prophage proteins.
  • Graphical Genome Viewers: High-resolution circular and linear genome representations rendered in the web interface. Regions are color-coded by completeness: green for intact, blue for questionable, and red for incomplete. Clicking any region expands an interactive linear track showing gene orientations and functional annotations.

Troubleshooting

When running prophage prediction pipelines, bioinformaticians frequently encounter biological edge cases and technical constraints. The following troubleshooting protocols resolve common issues observed with PHASTER.

1. Multi-Contig Submissions Process Only 10 Sequences

Symptom: When submitting a multi-FASTA file from an unclosed draft genome containing 45 contigs, the output report lists results for only contigs 1 through 10, omitting the remainder of the genome.

Cause: By default, the PHASTER server assumes that multi-FASTA files represent separate standalone replicons and caps processing at 10 sequences to prevent accidental server overloading. When handling draft assemblies or metagenomic contigs, the server must be explicitly instructed to treat the file as a single fragmented assembly.

Resolution: On the web interface, ensure the Metagenomic data or multi-contig checkbox is checked prior to clicking Submit. In the REST API, append contigs=1 (or contig=1) as a form field or URL query parameter. The server will then process all contigs exceeding 2,000 bp as parts of a unified genomic dataset.

2. Low-Confidence or Questionable Prophage Regions

Symptom: PHASTER identifies a prophage region with a completeness score between 70 and 85, flagging it as questionable.

Cause: A questionable score can indicate either a decayed phage remnant (which has lost structural genes over evolutionary time) or a novel, uncharacterized bacteriophage whose proteins share weak homology with existing reference databases.

Resolution: Open detail.txt and inspect the functional gene annotations. If the region contains an intact integrase, a terminase, and structural genes (capsid and tail proteins) that were assigned low scores solely due to divergent amino acid identity, the region likely represents a functional, novel prophage. Cross-validate the coordinates using complementary compositional tools like PhiSpy or machine-learning classifiers such as VirSorter2 or geNomad. Conversely, if the region consists of a transposase accompanied by hypothetical bacterial proteins, classify it as a degraded genomic island.

3. Missing attL and attR Attachment Sites

Symptom: An intact prophage region (score > 100) displays exact gene annotations, but the predicted_att.txt report lists no attachment sites.

Cause: Attachment site detection depends on locating short direct repeats (12–80 bp) flanking an integrase. If the prophage has experienced genomic rearrangements, secondary deletions, or integrated without a canonical tyrosine/serine integrase mechanism, sequence repeat matching fails. Furthermore, if the prophage spans across contig boundaries in a draft assembly, the terminal repeats reside on separate contigs.

Resolution: Manually inspect the flanking genomic boundaries in a genome browser. Check whether the prophage borders a tRNA gene (e.g., tRNA-Arg, tRNA-Leu), which serve as primary integration targets for temperate phages. Extract 500 bp upstream and downstream of the prophage boundary and perform local alignment against host tRNA sequences to locate degenerate att cores.

4. REST API Timeouts and Cloudflare 524 Responses

Symptom: Automated HTTP client requests hang or terminate with HTTP 504/524 gateway timeout errors during submission.

Cause: Submitting massive sequence files over slow network connections or sending immediate synchronous queries during periods of peak compute cluster utilization triggers web proxy timeouts.

Resolution: Never execute synchronous waiting on the submission POST call. Ensure your client receives the initial HTTP 200 containing the job_id, then implement an asynchronous polling loop with exponential backoff (checking status every 30 to 60 seconds). Set client socket read timeouts to at least 120 seconds for the initial multipart upload.

Free alternatives

While PHASTER remains an established standard for quick, homology-driven prophage identification, researchers frequently deploy alternative open-source bioinformatic tools. Different tools utilize distinct detection paradigms, ranging from nucleotide composition profiling to deep learning classifiers. The data table below compares leading free prophage and viral identification tools:

Tool Name Primary Detection Paradigm Execution Interface Deployment / Container Key Strengths Primary Limitations
PHASTER BLAST+ homology matching and DBSCAN spatial clustering Web Server, REST API Hosted cloud cluster (112 CPU cores) Zero local setup; automated completeness scoring; interactive D3/Angular visualizers Constrained local scaling; limited sensitivity for novel phages lacking database matches
PHASTEST MMseqs2 sequence translation, homology search, and clustering Web Server, REST API, Docker Hosted cloud server or local Docker container 45–80× faster than PHASTER; updated viral databases; official local container support Local Docker container requires substantial RAM and reference database downloads
PhiSpy Genomic composition (AT/GC skew, k-mers, strand bias) and Random Forest Command-line interface (CLI) Python package, Bioconda Excellent at detecting novel or divergent prophages; lightweight local execution Requires pre-annotated GenBank files; susceptible to false positives on genomic islands
VirSorter2 Multi-classifier machine learning (Random Forest, neural nets) and hallmark genes Command-line interface (CLI) Conda, Snakemake, Docker Detects diverse DNA and RNA viruses; modular architecture; high boundary accuracy Computationally intensive; requires complex multi-step database configuration
VIBRANT HMM protein profile matching (Pfam, VOG, KEGG) and neural networks Command-line interface (CLI) Conda, Python script Recovers proviruses and characterizes metabolic auxiliary genes (AMGs); handles mixed assemblies Higher false-negative rate on heavily degraded or short (<1,000 bp) prophage remnants
geNomad Neural network classification and marker protein enrichment (MMseqs2) Command-line interface (CLI) Conda, Docker, Singularity Ultra-fast high-throughput scalability; separates plasmids from viruses; low false-positive rates Requires high-memory local workstations for large reference database indexing

When selecting an alternative, consider your project scale and infrastructure constraints. For analyzing isolated bacterial genomes without local command-line configuration, PHASTER and PHASTEST offer immediate web-based utility. For high-throughput metagenomic studies or discovery of highly divergent phages, local command-line tools like geNomad, VirSorter2, and PhiSpy are recommended.

Frequently asked questions

What is the difference between PHAST, PHASTER, and PHASTEST?

PHAST (2011) was the original phage search tool developed by the Wishart research group, utilizing legacy BLAST and running on a 32-core cluster. PHASTER (2016) introduced a complete software and hardware overhaul, upgrading to BLAST+ 2.3.0+, expanding the compute cluster to 112 CPU cores, quadrupling the reference viral database, and reducing average runtimes by 4.3-fold. PHASTEST (2023) is the latest iteration, replacing BLAST+ with MMseqs2 to achieve an additional 45-fold to 80-fold speed acceleration while introducing standalone Docker container distribution.

Can PHASTER analyze fragmented draft genomes with hundreds of contigs?

Yes. However, when submitting multi-FASTA files containing fragmented draft genomes or metagenomic assemblies, you must enable the multi-contig parameter (checking the metagenomic data box on the web interface or supplying contigs=1 to the REST API). Without this parameter, PHASTER defaults to processing only the first 10 contigs. Contigs shorter than 2,000 base pairs are automatically excluded from analysis to prevent false-positive boundary artifacts.

What does an ‘intact’ prophage classification mean biologically?

An intact classification (completeness score greater than 90) indicates that the detected region possesses sufficient genomic length, viral protein density, and essential cornerstone genes (capsid, tail, portal, terminase, integrase, and lysis enzymes) to potentially form a functional, replication-competent bacteriophage capable of lysogenic induction and virion assembly.

How does PHASTER identify prophage integration boundaries?

PHASTER determines prophage boundaries through two complementary methods: first, by identifying spatial clusters of phage-like genes using the DBSCAN algorithm (with a maximum intergenic distance of 3,000 bp), and second, by scanning the flanking regions of detected integrase genes for short direct nucleotide repeats (12 to 80 bp) that define the attL and attR attachment sites.

Why are my submitted jobs returning an error or failing to process?

Common causes of job failures include uploading non-standard FASTA headers containing special characters or punctuation, submitting sequence files exceeding 250 MB, or uploading raw read archives (such as FASTQ files) instead of assembled contigs. Ensuring files use valid IUPAC nucleotide codes and standard FASTA or GenBank formatting resolves most submission errors.

Can PHASTER run offline on a local high-performance computing cluster?

PHASTER itself is maintained strictly as a centralized web server and REST API due to its backend cluster architecture and proprietary database indexing. Researchers requiring standalone local execution should deploy its official successor, PHASTEST, which is distributed as a containerized Docker image with local database support.

Sources

  • Arndt, D., Grant, J. R., Marcu, A., Sajed, T., Pon, A., Liang, Y., and Wishart, D. S. (2016). PHASTER: a better, faster version of the PHAST phage search tool. Nucleic Acids Research, 44(W1), W16–W21. https://doi.org/10.1093/nar/gkw387 (Accessed 2026-11-14).
  • Zhou, Y., Liang, Y., Lynch, K. H., Dennis, J. J., and Wishart, D. S. (2011). PHAST: A Fast Phage Search Tool. Nucleic Acids Research, 39(suppl_2), W347–W352. https://doi.org/10.1093/nar/gkr485 (Accessed 2026-11-14).
  • Wishart, D. S., et al. (2023). PHASTEST: Faster and more accurate identification of prophages in bacterial genomes. Nucleic Acids Research, 51(W1), W208–W214. https://doi.org/10.1093/nar/gkad382 (Accessed 2026-11-14).
  • Akhter, S., Aziz, R. K., and Edwards, R. A. (2012). PhiSpy: a novel algorithm for finding prophages in bacterial genomes that combines similarity- and composition-based strategies. Nucleic Acids Research, 40(16), e126. https://doi.org/10.1093/nar/gks406 (Accessed 2026-11-14).
  • Guo, J., Bolduc, B., Zayed, A. A., et al. (2021). VirSorter2: a multi-classifier, expert-guided approach to detect diverse DNA and RNA viruses. Microbiome, 9, 37. https://doi.org/10.1186/s40168-020-00990-1 (Accessed 2026-11-14).
  • Kieft, K., Zhou, Z., and Anantharaman, K. (2020). VIBRANT: automated recovery, annotation and curation of microbial viruses, and evaluation of viral community function from genomic sequences. Microbiome, 8, 90. https://doi.org/10.1186/s40168-020-00867-0 (Accessed 2026-11-14).
  • PHASTER Web Server and REST API Instructions. University of Alberta / Wishart Research Group. http://phaster.ca/instructions (Accessed 2026-11-14).

Data: Prophage and Viral Identification Bioinformatic Tools Specification Matrix

Comparative specifications, computational algorithms, minimum input thresholds, and deployment modes for prominent prophage and viral identification tools, compiled from primary peer-reviewed literature and official software documentation.

Period covered: 2011/2026. Coverage: Global.

Prophage and Viral Identification Bioinformatic Tools Specification Matrix
Tool NameRelease Year (year)Detection EngineMinimum Contig Length (bp)Interface TypeCompute EnvironmentAttachment Site Search
PHASTER2016BLAST+ 2.3.0+ homology search and DBSCAN spatial clustering2000Web Server and REST APIHosted 112-core Linux compute clusterYes (12-80 bp repeats)
PHAST2011Legacy BLAST homology search and DBSCAN spatial clustering2000Web Server and CGI scriptHosted 32-core Linux compute clusterYes (12-80 bp repeats)
PHASTEST2023MMseqs2 sequence translation and rapid homology clustering2000Web Server, REST API, DockerHosted cloud server or local containerYes (enhanced repeat search)
PhiSpy2012Genomic composition profiling and Random Forest classification2000Command-Line Interface (CLI)Local workstation or HPC cluster (Python)Yes (attL/attR scanning)
VirSorter22021Multi-classifier machine learning and hallmark viral gene models1000Command-Line Interface (CLI)Local workstation or HPC cluster (Conda/Snakemake)No (requires external tools)
VIBRANT2020HMM profile matching (Pfam, VOG, KEGG) and neural networks1000Command-Line Interface (CLI)Local Linux workstation (Conda environment)No (extracts provirus windows)
geNomad2023Neural network classification and marker protein enrichment (MMseqs2)1000Command-Line Interface (CLI)Local workstation or HPC cluster (Conda/Docker)Yes (integrated provirus excision)

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about PHASTER: Prophage Identification, API Workflows, and Genome Analysis

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →