MarpolBase - RDF / MCP

MarpolBase RDF and MCP server

MarpolBase data on Marchantia polymorpha genes, structural annotation, expression (MBEX), co-expression, gene-linked literature and orthologs are available as linked data (RDF) with a public SPARQL endpoint, and through an MCP (Model Context Protocol) server that lets AI assistants such as Claude and ChatGPT query MarpolBase directly.

RDF overview SPARQL endpoint Linked data URIs Download MCP server Setup: Claude Setup: ChatGPT Tools Example questions

RDF overview

The RDF knowledge base is built from MarpolBase, MBEX, the MarpolBase ortholog analysis and the gene-linked literature curation. Gene IDs follow the current reference genome MpTak_v7.1r2 (e.g. http://marchantia.info/rdf/gene/Mp1g00710). The data are released under CC BY 4.0.

Graph	Contents	Triples
`rdf/graph/MpTak_v7.1r2/function`	18,010 genes with symbols (gene nomenclature), products, GO / Pfam / KOG / KEGG; 850 expression conditions from 2,447 RNA-seq datasets and highly expressed conditions per gene; 636,817 co-expression pairs (PCC, HRR, MR; HRR ≤ 200); 2,857 gene–paper links in 529 papers; orthologs, orthogroups and alleles across 13 genomes (OrthoFinder)	8,353,616
`rdf/graph/MpTak_v7.1r2/annotation`	Structural annotation of MpTak_v7.1r2: genes, transcripts, exons and CDS with coordinates (FALDO) and Sequence Ontology types	5,774,818
`rdf/graph/xref/uniprot`	Links from proteins to UniProt accessions	21,921
`rdf/graph/ontology/so`, `sio`, `faldo`; `rdf/graph/metadata`	Ontologies used by the data, and the dataset description (version, license, provenance)

Graph IRIs start with http://marchantia.info/. Counts are for the release of 2026-11-09.

Example: one gene in RDF (MpCEN1, Mp1g00710; click to expand)

The gene itself and one node of each kind linked to it: its location, a condition where it is highly expressed, a co-expression partner and a paper. Prefixes shorten the URIs; comments note what is left out.

@prefix gene:     <http://marchantia.info/rdf/gene/> .
@prefix mpo:      <http://marchantia.info/ontology/> .
@prefix cond:     <http://marchantia.info/rdf/condition/> .
@prefix enriched: <http://marchantia.info/rdf/enriched/> .
@prefix coexp:    <http://marchantia.info/rdf/coexp/> .
@prefix lit:      <http://marchantia.info/rdf/litassert/> .
@prefix region:   <http://marchantia.info/rdf/MpTak_v7.1r2/region/> .
@prefix obo:      <http://purl.obolibrary.org/obo/> .
@prefix faldo:    <http://biohackathon.org/resource/faldo#> .
@prefix skos:     <http://www.w3.org/2004/02/skos/core#> .
@prefix dct:      <http://purl.org/dc/terms/> .
@prefix foaf:     <http://xmlns.com/foaf/0.1/> .
@prefix xsd:      <http://www.w3.org/2001/XMLSchema#> .

# The gene MpCEN1 (Mp1g00710)
gene:Mp1g00710  a mpo:Gene , obo:SO_0001217 ;        # protein-coding gene
    dct:identifier      "Mp1g00710" ;
    skos:prefLabel      "MpCEN1"@en ;
    skos:altLabel       "Mapoly0103s0016"@en ;     # JGI 3.1 ID
    mpo:product         "centrin" ;
    mpo:functionSummary "Homolog of CENTRIN, essential for the motility apparatus; expression correlates with and depends on MpDUO1 activity." ;
    mpo:hasGOTerm       obo:GO_0005509 , obo:GO_0005813 ;   # 2 of 5 GO terms
    mpo:hasPfam         <http://identifiers.org/pfam/PF13499> ;
    mpo:inOrthoGroup    <http://marchantia.info/rdf/orthogroup/core13v71r3-20261017T014356Z/OG0000038> ;
    mpo:hasGenomeGene   <http://marchantia.info/rdf/gene/MPTAK1/Mp1g00710> ,   # Tak-1 / Tak-2 copies
                        <http://marchantia.info/rdf/gene/MPTAK2/Mp1g00710> ;   # (orthologs hang off these)
    obo:BFO_0000050     <http://marchantia.info/rdf/MpTak_v7.1r2/sequence/chr1> ;   # part of chr1
    faldo:location      region:chr1_635664_638217_minus ;
    foaf:page           <https://marchantia.info/MpTak_v7.1r2/Mp1g00710/> .

# Its location on the MpTak_v7.1r2 genome
region:chr1_635664_638217_minus  a faldo:Region ;
    faldo:reference  <http://marchantia.info/rdf/MpTak_v7.1r2/sequence/chr1> ;
    mpo:minPosition  635664 ;
    mpo:maxPosition  638217 .

# Highly expressed in sperm cells (1 of 12 such conditions)
enriched:cPRJNA685181_G2.Mp1g00710  a mpo:ConditionEnrichedGene ;
    mpo:refGene       gene:Mp1g00710 ;
    mpo:refCondition  cond:PRJNA685181_G2 ;
    mpo:meanTPM       "8358.7481"^^xsd:double ;
    mpo:foldVsMedian  "82.9465"^^xsd:double .

cond:PRJNA685181_G2
    mpo:tissue          <http://marchantia.info/rdf/vocab/tissue/sperm_cell> ;
    mpo:strain          <http://marchantia.info/rdf/vocab/strain/Tak-1> ;
    mpo:sexLabel        "male" ;
    mpo:runCount        2 ;
    mpo:referenceLabel  "Walker et al., 2025" .

# Co-expressed with Mp1g26930 (1 of 47 partners with HRR <= 200)
coexp:Mp1g00710.Mp1g26930  a mpo:CoexpressionEdge ;
    mpo:gene1  gene:Mp1g00710 ;
    mpo:gene2  gene:Mp1g26930 ;
    mpo:pcc    "0.791925"^^xsd:double ;
    mpo:hrr    2 ;
    mpo:mr     "1.4142"^^xsd:double .

# Described in a paper (1 of 3)
lit:Mp1g00710.pmid30538242  a mpo:GeneLiteratureAssertion ;
    mpo:aboutGene        gene:Mp1g00710 ;
    mpo:fromPaper        <http://marchantia.info/rdf/paper/pmid/30538242> ;
    mpo:evidenceType     mpo:Experimental ;
    mpo:geneRole         mpo:Subject ;
    mpo:isCoreReference  true .

The full description of the gene:

curl -H 'Accept: text/turtle' https://marchantia.info/rdf/gene/Mp1g00710

SPARQL endpoint

https://marchantia.info/sparql

Queries run over all graphs above by default. Each query may run up to 30 seconds and return up to 1,000,000 rows. Requests are rate-limited (about 5 per second per client); please use the download files for bulk analysis.

Example: genes whose symbol starts with MpCEN

PREFIX mpo:  <http://marchantia.info/ontology/>
PREFIX skos: <http://www.w3.org/2004/02/skos/core#>
SELECT DISTINCT ?gene ?symbol ?product WHERE {
  ?gene a mpo:Gene ;
        skos:prefLabel ?symbol .
  OPTIONAL { ?gene mpo:product ?product }
  FILTER(STRSTARTS(STR(?symbol), "MpCEN"))
}
ORDER BY ?symbol
LIMIT 10

Example: number of orthologs of MpCEN1 (Mp1g00710) per species

PREFIX mpo:  <http://marchantia.info/ontology/>
PREFIX dct:  <http://purl.org/dc/terms/>
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
SELECT ?species (COUNT(DISTINCT ?other) AS ?orthologs) WHERE {
  <http://marchantia.info/rdf/gene/Mp1g00710> mpo:hasGenomeGene ?copy .
  ?copy ?relationship ?other .
  ?relationship rdfs:subPropertyOf mpo:hasOrtholog .
  ?other mpo:genome/dct:identifier ?species .
}
GROUP BY ?species
ORDER BY ?species

From the command line:

curl -G https://marchantia.info/sparql \
     -H 'Accept: text/csv' \
     --data-urlencode 'query=SELECT * WHERE { ?s ?p ?o } LIMIT 5'

Linked data URIs

Gene URIs are dereferenceable. A web browser is redirected to the MarpolBase gene page; RDF clients get a description of the gene by content negotiation (Turtle, RDF/XML, JSON-LD or N-Triples).

curl -H 'Accept: text/turtle' https://marchantia.info/rdf/gene/Mp1g00710

Download

marpolbase.nt.gz (function graph, N-Triples) · sha256
marpolbase.ttl.gz (function graph, Turtle)
mpv71.ttl.gz (structural annotation of MpTak_v7.1r2, Turtle) · sha256

MCP server

The MarpolBase MCP server gives AI assistants a set of tools to look up genes, expression, co-expression, literature, orthologs, sequences and gene ID histories, and to run SPARQL queries. It needs no account or API key. Every answer comes with its provenance (data version, conditions, thresholds and warnings).

MCP endpoint: https://marchantia.info/mcp

Transport: Streamable HTTP. Authentication: none.

Claude (web and desktop): custom connector (recommended)

Open Claude and go to Settings → Customize → Connectors.
Click + and choose Add custom connector.
Enter a name (e.g. MarpolBase) and the server URL https://marchantia.info/mcp, then click Add.
In a chat, click +, open Connectors and turn on MarpolBase for that conversation.

Custom connectors are available on all Claude plans (Free accounts can add one). On Team and Enterprise plans, an organization owner must first add the connector under Organization settings → Connectors; members then connect it under Customize → Connectors. If a connector was added before an update of the tool list, disconnect and connect it again to load the new tools.

Claude Desktop: configuration file (alternative)

If custom connectors are not available to you, add the server to claude_desktop_config.json (Settings → Developer → Edit Config) and restart Claude Desktop. This uses the community bridge mcp-remote and requires Node.js.

{
  "mcpServers": {
    "marpolbase": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://marchantia.info/mcp"]
    }
  }
}

Claude Code

claude mcp add --transport http marpolbase https://marchantia.info/mcp

ChatGPT (web)

How to add a custom MCP server depends on the ChatGPT plan and interface version. Use ChatGPT on the web; the mobile apps cannot add one.

Plus: add as a custom MCP server (confirmed in November 2026, without Developer mode)

Open Plugins in the left sidebar.
Click + and choose Add custom MCP server (shown as “カスタム MCP サーバーを追加” in the Japanese interface).
Enter the name MarpolBase and the server URL https://marchantia.info/mcp, with no authentication.

Pro, Business, Enterprise and Edu use Developer mode as below (Pro allows read-only tools, which is all MarpolBase needs).

Pro: add as a plugin

Turn on Developer mode under Settings → Apps → Advanced settings (in newer versions, Settings → Security and login).
Open Plugins in the left sidebar and click +.
Fill in the form: Name MarpolBase; Connection: Server URL https://marchantia.info/mcp; Authentication: No Auth.
Acknowledge the custom-server warning, click Create, then Try in Chat.

Business, Enterprise and Edu: add as an app

Enable Developer mode under Settings → Apps → Advanced settings (on Enterprise and Edu, an admin first grants access under Workspace settings → Permissions & roles → Connected data).
Go to Apps → Create, enter the name MarpolBase, the MCP server URL https://marchantia.info/mcp and a description.
Click Scan Tools, wait for the scan, then click Create.
The app is created as a draft; an admin publishes it from Workspace settings → Apps → Drafts.

In a chat, select MarpolBase from the tools menu or mention it in your prompt; the selection applies to that message only. ChatGPT keeps the tool list from the time the app was created. When MarpolBase adds tools, recreate the plugin (Pro) or app (Business), or refresh it under Workspace settings → Apps → … → Action control (Enterprise and Edu). Menu names follow the ChatGPT interface as of November 2026 and may change.

Tools

Tool	What it does
Genes
`search_genes`	Find genes by symbol, gene ID, v3 (Mapoly) ID, product, GO or Pfam
`get_gene`	Annotation of one gene: symbol, product, GO / Pfam / KOG / KEGG, orthogroup, function summary, literature count, highly expressed conditions, genomic location
`convert_gene_ids`	Map gene IDs between genome versions (JGI 3.1 to MpTak_v7.1r2) with the history of each change; flags IDs that name different genes in different versions
`list_genes_in_region`	Genes overlapping a genomic region (MpTak_v7.1r2 coordinates)
`get_orthologs`	Orthologs in 13 genomes with relationship type (one-to-one, one-to-many, ...), orthogroup and alleles
`get_gene_sequence`	Protein, CDS, mRNA or genomic sequence of a gene
`get_genomic_region`	Nucleotide sequence of a genomic region
Expression (MBEX)
`list_conditions`	Expression conditions filtered by tissue, strain, stage, sex, mutation, treatment or infection
`resolve_vocabulary`	Map free-text terms to the controlled vocabulary used for conditions
`get_gene_expression_profile`	Mean TPM of a gene in every condition
`find_genes_by_condition`	Genes highly expressed in given conditions
`compare_conditions`	Differential expression between two conditions (DESeq2)
`get_coexpressed_genes`	Co-expressed genes with PCC, HRR and MR
`get_coexpression_subnetwork`	Co-expression network among a set of genes
Literature
`search_papers`	Find papers by title words, journal, year, DOI or PMID
`get_gene_literature`	Papers about given genes, with evidence type and the gene's role in each paper
`get_paper_genes`	Genes discussed in a paper
Schema and SPARQL
`describe_schema`	Classes, predicates and example SPARQL queries of the knowledge base
`run_sparql`	Run any SPARQL SELECT / ASK query

Example questions

In which tissues is MpCEN1 expressed most strongly, and which genes are co-expressed with it?
List the genes on chr1 between 7,630,000 and 7,660,000 and summarize what is known about each.
A paper from 2019 mentions Mp1g07140. Which gene does that ID refer to in MpTak_v7.1r2?
What are the orthologs of MpARF1 in Physcomitrium patens and Arabidopsis thaliana?
Which genes are upregulated in the antheridiophore compared with the thallus?
Find papers with experimental evidence for MpKAI2A and summarize them.

Questions may be written in any language. Vocabulary for conditions is in English, and the assistant translates terms as needed.

Citation

Please cite the MarpolBase paper when using MarpolBase data: MarpolBase: Genome database for Marchantia polymorpha featuring high quality reference genome sequences. Plant & Cell Physiology (2025).