[Paper Review] Genesis: Towards the Automation of Systems Biology Research

Ievgeniia Tiukova, Daniel Brunnsåker+9 more|arXiv (Cornell University)|Aug 20, 2024

Gene Regulatory Network AnalysisBiochemistry, Genetics and Molecular Biology3 citations

TL;DR

Genesis proposes a fully automated, hypothesis-driven robot scientist for systems biology, integrating 1,000 automated μ-bioreactors, AutonoMS for high-throughput mass spectrometry, and LGEM+ for relational learning to improve genome-scale metabolic models. It demonstrated automated abductive hypothesis generation and model refinement, predicting growth in minimal media and identifying 2,094 candidate model improvements with formal logic and FBA integration.

ABSTRACT

The cutting edge of applying AI to science is the closed-loop automation of scientific research: robot scientists. We have previously developed two robot scientists: `Adam' (for yeast functional biology), and `Eve' (for early-stage drug design)). We are now developing a next generation robot scientist Genesis. With Genesis we aim to demonstrate that an area of science can be investigated using robot scientists unambiguously faster, and at lower cost, than with human scientists. Here we report progress on the Genesis project. Genesis is designed to automatically improve system biology models with thousands of interacting causal components. When complete Genesis will be able to initiate and execute in parallel one thousand hypothesis-led closed-loop cycles of experiment per-day. Here we describe the core Genesis hardware: the one thousand computer-controlled $μ$-bioreactors. For the integrated Mass Spectrometry platform we have developed AutonoMS, a system to automatically run, process, and analyse high-throughput experiments. We have also developed Genesis-DB, a database system designed to enable software agents access to large quantities of structured domain information. We have developed RIMBO (Revisions for Improvements of Models in Biology Ontology) to describe the planned hundreds of thousands of changes to the models. We have demonstrated the utility of this infrastructure by developed two relational learning bioinformatic projects. Finally, we describe LGEM+ a relational learning system for the automated abductive improvement of genome-scale metabolic models.

Motivation & Objective

• To automate systems biology research by replacing human-led hypothesis cycles with a closed-loop robot scientist.
• To develop a scalable infrastructure for executing 1,000 hypothesis-led experimental cycles per day in systems biology.
• To enable automated, iterative model improvement using AI, robotics, and formal knowledge representation in eukaryotic systems like yeast.
• To integrate multi-omics data (metabolomics, transcriptomics) with AI-driven hypothesis generation and validation.
• To demonstrate that AI-driven automation can outperform human scientists in speed, cost, and model accuracy for complex biological systems.

Proposed method

• The system uses a network of 1,000 computer-controlled μ-bioreactors to run parallel experiments on S. cerevisiae under varying conditions.
• AutonoMS automates high-throughput mass spectrometry: experiment execution, data processing, and analysis.
• Genesis-DB provides structured, software-agent-accessible storage of biological knowledge and experimental data.
• RIMBO ontology formalizes hundreds of thousands of model revision proposals for biological models.
• LGEM+ converts genome-scale metabolic models (GEMs) from SBML into logical theories using first-order logic, enabling automated abductive reasoning.
• Deductive inference via iProver on LGEM+ theories predicts strain growth/no-growth in minimal media, validating model accuracy.

Experimental results

Research questions

• RQ1 Can a robot scientist automate the entire closed-loop cycle of hypothesis generation, experimentation, and model refinement in systems biology?
• RQ2 Can AI-driven hypothesis generation via LLMs and relational learning produce biologically relevant, testable model improvements?
• RQ3 Can automated high-throughput experimentation with μ-bioreactors and MS outperform traditional, human-guided systems biology approaches?
• RQ4 How can formal logic and automated theorem proving be used to validate and refine genome-scale metabolic models?
• RQ5 To what extent can LLMs contribute to scientific hypothesis generation without compromising factual accuracy?

Key findings

• Genesis demonstrated automated abductive hypothesis generation, producing 2,094 unique candidate model improvements for genome-scale metabolic models.
• The LGEM+ system successfully encoded S. cerevisiae's metabolic network into a first-order logic theory, enabling formal reasoning and prediction of growth/no-growth phenotypes in minimal media.
• Deductive inference using iProver on the LGEM+ logic model accurately predicted growth phenotypes, validating the formal model's correctness.
• Integration of flux-balance analysis (FBA) with the logic model enabled constraint-based simulation, improving model fidelity and hypothesis evaluation.
• A model-driven experimental design strategy was demonstrated using differential expression analysis on the PFK2 gene, showing the system's capacity for targeted, hypothesis-guided experimentation.
• The project established a foundation for full-scale automation, with core hardware and software components (μ-bioreactors, AutonoMS, databases, ontologies) in place and functional.