Ai-enhanced knowledge graphs for plant digital twin solutions

Overview

KnetMiner is a software platform designed to accelerate gene and trait discovery in plant science through AI-enhanced knowledge graphs. The platform integrates multi-omics data across multiple crop species into FAIR (Findable, Accessible, Interoperable, Reusable) knowledge graphs, enabling researchers and breeders to rapidly identify candidate genes, gene networks, and trait relationships critical for crop improvement. By combining knowledge graph technology with large language models and retrieval augmented generation (RAG), KnetMiner supports the development of Digital Twins for plant science, facilitating swift queries on gene orthologues, genetic interactions, and expression patterns across cultivars and environmental conditions.

The platform addresses the challenge of navigating vast amounts of biological literature and genomic data by automating complex search tasks and delivering structured insights. Applications include accelerating crop resilience research against biotic and abiotic stresses, supporting digital twin development, and enabling data-driven decision-making in plant breeding programs. The technology is relevant to organizations seeking to streamline gene discovery processes for sustainable agriculture and food security goals.

Technical specifications

Core capabilities:

• Multi-species knowledge graph construction integrating genome, multi-omics, and scientific literature data
• GraphRAG search engine combining knowledge graphs with large language models for AI-powered reasoning
• Semantic search and query interface that mines genome-scale knowledge graphs to reveal gene-trait relationships
• Subgraph visualization displaying gene interactions, networks, and pathways through an intuitive graphical user interface
• Neo4j database backend with APIs for data access and manipulation
• ETL pipeline for processing diverse, large-scale datasets including private genomes and homologies
• Adherence to community data standards including BioLink, BioCypher, BioSchemas, and AgroSchemas
• Tiered SaaS service levels including premium resources for advanced research

Key development areas for collaboration:

• Enhancing the ETL pipeline for efficient handling of diverse, large-scale datasets
• Optimizing data schemas for complex biological data representation
• Expanding Neo4j database capabilities and upgrading APIs
• Implementing large language models for augmented semantic search and retrieval augmented generation

Technology readiness level

KnetMiner is at TRL Level 5, having been validated through numerous FAIR knowledge graphs created from open and private data on crop traits and multi-omics resources. The platform has been deployed as a national data resource within ELIXIR-UK and serves as a partner in the EU's AgroServ.eu network. Originally developed as an open-source project at Rothamsted Research, the technology transitioned to a commercial SaaS model following the formation of KnetMiner Limited in late 2023. The platform is supported by both public and commercial funding and has been utilized by academic and commercial crop science research organizations. Future development will focus on enhancing the platform for advanced plant biology research and digital twin applications through collaborative partnerships.