AI Laboratory
Apply artificial intelligence to explore, accelerate, and discover pathways from basic chemistry to the origins of life. Each module focuses on a different stage of abiogenesis.
Prebiotic Chemistry AI
Use AI to discover and optimise abiotic synthesis pathways for the building blocks of life
Pathway Discovery
Search for novel synthesis routes from simple molecules (H₂O, CO₂, NH₃, HCN, H₂S) to amino acids, nucleobases, sugars, and lipids using graph-neural-network guided exploration.
GNNRetrosynthesisMonte Carlo Tree SearchYield Optimisation
Given a reaction network, use Bayesian optimisation to find temperature, pH, mineral catalyst, and concentration conditions that maximise monomer yield.
Bayesian OptGaussian ProcessThermodynamic Feasibility
Evaluate Gibbs free energy (ΔG) along every branch of the reaction network. Highlight kinetically trapped states and suggest catalytic bypasses.
ΔG PredictionTransition StateMineral Catalyst Screening
Screen mineral surfaces (pyrite, montmorillonite, zeolites, iron-sulfur clusters) for catalytic activity using ML-predicted adsorption energies.
DFT SurrogateAdsorption EnergyAutocatalytic Discovery AI
Find self-sustaining chemical cycles — the foundation of metabolic networks
RAF Set Detection
Use reflexively autocatalytic food-set (RAF) algorithms accelerated by graph attention networks to identify self-sustaining subsets within reaction networks.
RAF TheoryGraph AttentionCycle Enumeration
Enumerate all autocatalytic cycles up to a given length. Rank by thermodynamic driving force and kinetic accessibility using reinforcement learning.
RLCycle RankingNetwork Growth Prediction
Predict how a small autocatalytic set expands over time by recruiting new reactions. Simulate compositional inheritance via stochastic models.
Network ExpansionCompositional GenomeMembrane Assembly AI
Model and optimise the spontaneous formation of lipid vesicles and protocell compartments
Critical Micelle Concentration
Predict CMC values for mixtures of prebiotic amphiphiles (fatty acids, isoprenoids, polycyclic aromatics) using molecular descriptor QSPR models.
QSPRCMC PredictionVesicle Stability Analysis
Evaluate bilayer stability under varying pH, ionic strength, temperature, and wet-dry cycles. Predict membrane permeability to small molecules.
MD SurrogatePermeabilityDivision Dynamics
Model vesicle growth-and-division cycles. Use physics-informed neural networks to predict when osmotic pressure or surface tension triggers fission.
PINNDivision TriggerReplication & Heredity AI
Explore how information-carrying polymers can copy themselves and evolve
Template-Directed Polymerisation
Simulate non-enzymatic RNA copying on mineral surfaces. AI optimises monomer activation chemistry, template sequence, and environmental cycling.
Sequence OptimisationKinetic ModelError Threshold Analysis
Calculate Eigen's error threshold for the current system. Determine maximum genome length sustainable given the observed copying fidelity.
QuasispeciesError CatastropheRibozyme Evolution
Run evolutionary search for catalytic RNA sequences. Fitness landscape exploration via genetic algorithms with structure prediction (secondary structure folding).
Genetic AlgorithmRNA FoldingHeredity Emergence
Model the transition from statistical (compositional) inheritance to template-based (digital) heredity. Track information content over generations.
Information TheoryPhase TransitionProto-Metabolism AI
Discover and analyse primitive energy-harvesting cycles that powered early life
Energy Coupling Discovery
Find thermodynamically favourable reaction couplings: pair exergonic reactions (e.g. FeS oxidation) with endergonic biosynthesis (e.g. peptide bond formation).
ΔG CouplingRedox PairingProto-TCA Cycle Search
Search for reverse-TCA-like cycles that could operate abiotically. Evaluate feasibility on mineral surfaces (iron-nickel-sulfide catalysis).
Reverse TCACarbon FixationChemiosmotic Gradient
Model proton/sodium gradients across protocell membranes at alkaline vents. Predict early chemiosmotic energy harvesting potential.
pH GradientProton Motive ForceSelection & Evolution AI
Simulate Darwinian selection among populations of protocells with heritable variation
Fitness Landscape Mapping
Generate and visualise multi-dimensional fitness landscapes for protocell populations. Identify adaptive peaks, valleys, and neutral networks.
NK ModelLandscape VisualisationGroup Selection Dynamics
Model multi-level selection between competing protocell lineages. Track cooperation vs parasitism in populations sharing resources.
Multi-Level SelectionParasite ControlMajor Transitions Detector
Identify major evolutionary transitions in simulation data: compartmentalisation, replicator integration, division-of-labour emergence.
Transition DetectionAnomaly DetectionEnvironment Optimiser AI
Search parameter space for environmental conditions most conducive to life's emergence
Multi-Parameter Sweep
Bayesian optimisation over temperature, pH, salinity, mineral composition, UV flux, and wet-dry cycling frequency to maximise protocell complexity.
Bayesian OptMulti-ObjectiveScenario Comparison
Compare alkaline vent, warm little pond, iron-sulfur world, and RNA world scenarios. Rank each by probability of generating self-replicating systems.
Scenario RankingSensitivity AnalysisDay-Night & Seasonal Cycles
Investigate how periodic environmental forcing (temperature cycling, UV pulses, tidal wet-dry) drives polymerisation and selection dynamics.
Periodic ForcingOscillatory ChemistryFull Synthesis Pipeline
End-to-end AI-driven exploration from simple chemistry to a self-replicating protocell
Step-by-Step Abiogenesis
Run the complete pipeline: prebiotic synthesis → monomer accumulation → polymer formation → autocatalysis → compartmentalisation → replication → selection. AI guides each transition.
Full PipelineGuided SearchMulti-StageBottleneck Identification
Analyse the full pathway and identify the rate-limiting step. Is it monomer synthesis? Membrane formation? Replication fidelity? AI pinpoints the weakest link.
Rate LimitingSensitivityNovel Hypothesis Generation
Use large-language-model reasoning over simulation results, literature embeddings, and thermodynamic data to propose novel hypotheses for untested origin-of-life pathways.
LLM ReasoningLiterature MiningHypothesisReport Generation
Compile results from all AI modules into a structured research report with figures, tables, and citations. Export as PDF or Markdown.
Auto-ReportPDF / Markdown