Skip to main content
ARS Home » Northeast Area » Ithaca, New York » Robert W. Holley Center for Agriculture & Health » Plant, Soil and Nutrition Research » Research » Publications at this Location » Publication #434197
Research Project: Enabling Mechanistic Allele Mining to Accelerate Genomic Selection for New Agro-Ecosystems

Location: Plant, Soil and Nutrition Research

Title: AI Agents for Plant Genomics Research via Model Context Protocol Integration with Gramene and Complementary Biological Databases

Author
item OLSON, ANDREW - Cold Spring Harbor Laboratory
item CHOUGULE, KAPEEL - Cold Spring Harbor Laboratory
item Gladman, Nicholas
item KUMARI, SUNITA - Cold Spring Harbor Laboratory
item LU, ZHENYUAN - Cold Spring Harbor Laboratory
item WEI, SHARON - Cold Spring Harbor Laboratory
item Ware, Doreen

Submitted to: Cold Spring Harbor Meeting
Publication Type: Abstract Only
Publication Acceptance Date: 5/26/2026
Publication Date: 5/26/2026
Citation: Olson, A., Chougule, K., Gladman, N.P., Kumari, S., Lu, Z., Wei, S., Ware, D. 2026. AI Agents for Plant Genomics Research via Model Context Protocol Integration with Gramene and Complementary Biological Databases. Cold Spring Harbor Meeting. 90th Cold Spring Harbor Laboratory Symposium on Quantitative Biology: AI in Biology.

Interpretive Summary:

Technical Abstract: The growing complexity of plant genomics data demands new approaches to knowledge integration. AI agents powered by large language models (LLMs) offer a promising path, but their utility depends on reliable, structured access to specialized databases. The Model Context Protocol (MCP), an open standard for connecting AI systems to external data sources and tools, enables this by exposing database APIs as composable, agent-callable tools through a uniform interface. We present the Gramene MCP server (https://data.gramene.org/mcp), a new resource that makes Gramene's comparative plant genomics infrastructure directly accessible to AI agents. The Gramene search API operates over two complementary backends; a Solr index supporting full-text and faceted gene search across all species, and MongoDB collections housing genes, gene trees, ontologies (GO, PO, TO), Plant Reactome pathways, expression experiments, QTLs, and germplasm records. The MCP server exposes these as a suite of tools that agents can chain together for multi-step reasoning. Example workflows include identifying genes within a yield QTL interval. This task can involve scoring candidates by trait ontology annotation and tissue expression, retrieving conserved orthologs across related crops, and synthesizing results into a biologically interpretable ranking. Previously, such tasks required programming expertise and a deep understanding of the Gramene search API. The Gramene MCP server is designed to compose with a broader ecosystem of open bioinformatics MCP servers covering complementary data domains, including Ensembl (genome annotation and variant effect prediction), Reactome and KEGG (pathway and compound databases), UniProt (protein function), the Protein Data Bank (3D structure), STRING (protein interaction networks), and AlphaGenome (deep learning-based regulatory and variant effect prediction). We demonstrate multi-server agent workflows spanning gene function discovery, cross-species pathway gap analysis, and pan-genome presence/absence variation analysis across major crop species including maize, sorghum, and rice. This MCP-based architecture lowers the barrier to sophisticated data integration, making expert-level plant genomics accessible through natural language to a broad community of plant biologists and breeders. All tools are open source and compatible with MCP-enabled AI clients.