Graph-based digital mechanistic twin for crop growth prediction and molecular breeding optimization

Overview

This solution offers a graph-based digital mechanistic twin of crop growth that integrates biological models from gene expression networks through protein expression, metabolic networks, leaf physiology, and whole-plant growth and development. By unifying these layers within a graph-based building and querying context, the platform enables rapid data integration, AI-driven analysis, and improved prediction fidelity for crop performance across multiple genotypes and locations. The technology supports molecular breeding, gene editing, and transgenic strategy optimization by identifying gene targets for knock-out, knock-down, or up-regulation to enhance yield. Soybean serves as the initial test-bed, supported by 300 fully sequenced lines from a DOE-led project and extensive yield and atmospheric datasets from the SoyFACE facility.

Technical specifications

Key capabilities:

• Multi-layer integration of gene expression, protein, metabolic, leaf physiological, and plant growth models within a unified graph-based framework
• NoSQL-based architecture (under review) enabling integration of large volumes and diverse sources of biological and environmental data
• AI-driven querying of network nodes across each organizational layer
• Optimization routines that identify candidate genes for manipulation to improve crop yield
• Prediction of spatial and temporal yield distribution across locations and growing conditions
• Compatibility with Free-Air CO2 enrichment (FACE) technology data for elevated CO2 scenario modeling
• Built on advanced computing infrastructure provided by the National Center for Supercomputing Applications

Validated outcomes:

• Successful prediction of biomass crop yields and soybean seed yield
• Identification of specific genes whose up-regulation increases leaf photosynthesis, validated in field trials showing rising benefit with elevated CO2
• Accurate prediction of spatial yield variation for two biomass crops

Technology readiness level

The digital twin components have been individually validated against extant multi-location field trial data, with confirmed predictions of gene targets, photosynthetic improvements, and spatial yield patterns. The graph-based integration framework represents the next development phase, building on existing validated models and leveraging ongoing sequencing efforts (300 soybean lines) and established atmospheric and climate datasets. The platform is positioned for further validation within the soybean test-bed, with potential extension to additional crops and germplasm contexts.