AI/ML
AI-Aided Biodesign
Biochemistry-informed hybrid models, graph learning, and foundation models for more predictable biological design.
- AI/ML
- biodesign
- hybrid models
AI-aided biodesign uses standardized biological data to improve how we predict, design, and reason about genetic and genomic networks. DRAGGON Lab will combine mechanistic models with machine learning so that design tools can learn from experimental context while remaining interpretable to engineers and biologists.
Model families
We will develop biochemistry-informed hybrid models for gene expression and regulation, graph neural networks that treat biological designs as structured objects, and transformer or multimodal foundation models trained or fine-tuned on standardized genetic, genomic, metadata, and experimental data.
Data strategy
The lab will streamline data collection through SBOL, SynBioHub, Flapjack, and related workflow tools. SeqTrainer will help convert standardized data into tabular, sequence, graph, and benchmark-ready forms so researchers can compare models across biological inference tasks.
Long-term direction
As laboratory automation improves, agentic software systems could help orchestrate parts of the DBTL cycle: proposing designs, executing explainable code representations, requesting data, and using new measurements to improve models. The ambition is not automation for its own sake, but better biological understanding and more reliable engineering.