PlantScience.si · the vision

Towards plant science superintelligence

An intelligence system that draws on all of plant science at once, from the literature and its images to plant databases, and answers with the evidence behind each claim.

Concept: a plant growing from a knowledge graph Roots drawn as a knowledge graph feed a stem with four leaves, labelled visual evidence, database integration, education and evolutionary insight, rising to a glowing node labelled plant science superintelligence. KNOWLEDGE GRAPH Visual evidence Database integration Education Evolutionary insight PLANT SCIENCE SUPERINTELLIGENCE PlantScience.ai today

PlantScience.ai today · the foundation PlantScience.si builds on

480,253
open-access articles read into the PlantScience.ai knowledge graph
9.66M
entities, each linked to its source publication
47.2M
relationships between genes, traits, species and more

With support from colleagues at

Scientists, educators and resource leaders across UK plant science are helping to shape PlantScience.si.

John Innes Centre The Sainsbury Laboratory Sainsbury Laboratory Cambridge University Crop Science Centre Cambridge University Botanic Garden Department of Plant Sciences, University of Cambridge Gatsby Plant Science Education Programme EMBL-EBI The James Hutton Institute Department of Biology, University of Oxford University of Cambridge

The vision

Plant science knows more than any one of us can read.

Its knowledge is scattered across millions of papers, images and specialised databases, and it grows every year. A researcher asking what is known about a gene, a breeder narrowing a trait interval and a student learning to weigh evidence still assemble that picture by hand. PlantScience.si brings it together in one place.

PlantScience.si overview Open-access papers with figures, plant databases and a gene orthology network flow into the plant science knowledge graph. An AI agent reasons over the graph and answers plant scientists, breeders, educators and students. SOURCES KNOWLEDGE REASONING PEOPLE Papers text, figures andlegends Plant databases genomes, proteins,expression, RNA Gene orthology shared ancestryacross genomes Plant Science Knowledge Graph every edge traceable to its source AI agent plans, retrieves, cites Scientistsweigh evidence Breederstrait to gene Educatorsand students SOURCES Paperstext and figures Databaseslive records Orthologyshared ancestry Plant Science Knowledge Graph every edge traceable to its source AI agent plans, retrieves, cites Scientistsweigh evidence Breederstrait to gene Educatorsand students
Concept 1 · One intelligence, all of plant science. Papers with their figures, plant databases and a gene orthology network build one knowledge graph. An AI agent reasons across it and answers each person with the evidence behind every claim.

Four new capabilities

From answers in text to evidence you can see, query, teach with and trace through evolution.

Gene Tissue Finding
Visual evidence

See the image behind the finding

Much of plant science's primary evidence is visual. Phenotype photographs, micrographs and plots from the open literature are linked to the genes, tissues and findings they show, so an answer can display its evidence, with author, journal and licence.

gene name → identifier GenomesExpressionProteins ID · release · linkID · release · linkID · release · link
Database integration

Live records, never invented identifiers

Language models invent identifiers and numbers. The agent instead calls skills that query established plant bioinformatics databases directly, and every statement is labelled as literature-supported or database-supported, with identifier, release and link.

1The AI asks a question 2The student findsthe evidence 3The studentconcludes
Education

A teaching mode that asks, not answers for you

Developed with educators and students. Instead of handing over an answer, the agent poses a question at the student’s level. The student searches the published findings, figures and data, then reaches their own conclusion, with safeguards against outsourcing assessed work.

ArabidopsisRiceWheatBarley conserveddivergednot studied
Evolutionary insight

From Arabidopsis to crops, through shared ancestry

Most functional evidence comes from model species, while decisions are made in crops. A gene orthology network links each gene family across plant genomes, showing which functions are conserved, which have diverged and where evidence is missing.

In use

One question. Four kinds of evidence.

A plant scientist and a wheat breeder ask about the same gene, and each can follow every statement back to its source. In Education mode the AI asks the question, and the student finds the evidence.

PlantScience.siIllustrative mock-up

How it earns trust

It synthesises what is known. Scientists stay responsible for judgement.

Traceable

Every statement links to the publication or database record it came from.

Labelled

Literature-supported and database-supported claims are marked as such.

Honest about gaps

Inferences across species are hypotheses with confidence levels. Missing evidence is shown as missing.

Open

Versioned, FAIR releases, an open benchmark, and an API any AI agent can use.

Plant Science AI Initiative

A resource built and steered by its community.

Plant science researchers, breeders, educators and database providers unite around one open intelligence resource.

  • Contribute skills. Database providers add their resources through an open specification.
  • Set the standard. Members submit and review benchmark questions, published every year.
  • Guide priorities. A public roadmap, ranked by the people who use it.
  • Build on it. Open API and Model Context Protocol endpoints, with versioned knowledge-graph releases.
Community Literature update Benchmarks Release gate Versionedrelease

Team

Project leads

Portrait of Yiliang Ding
Project lead

Yiliang Ding

Professor and Group Leader
John Innes Centre

Profile ↗
Portrait of Haopeng Yu
Project co-lead

Haopeng Yu

Principal Investigator
Sainsbury Laboratory, University of Cambridge (SLCU)

Profile ↗