Use cases - Aura

How can Aura help

At Fastfold, we believe the future of discovery belongs to AI agents: purpose-built agents that reason, retrieve, predict, and simulate, reducing barriers to entry while scaling scientific capabilities.

Aura is our response: an AI agent fleet that empowers users across pharma, biotech, and academia to iterate, test, and validate hypotheses across modalities and molecular scales.

Core Capabilities

Protein and Ligand Modeling

Aura supports structure prediction and analysis using:

• AlphaFold2, ColabFold, and ESM for folding

• Boltz-1 for co-folded protein-ligand complexes

• Boltz-2 for combined affinity prediction and structure in one pass

• ProteinMPNN for sequence redesign and stability optimization

Affinity Prediction with Boltz-2

Boltz-2 predicts binding affinity (log(IC50)) and outputs interaction probabilities per model in an ensemble. Its performance rivals traditional FEP methods while running approximately 1000× faster, making it ideal for lead discovery and screening.

Aura automatically parses results and visualizes structure, predicted ΔG, and probability, all within a single workflow.

Generative Design Workflows

Aura enables multi-objective optimization and rational design with:

• De novo generation tools (Evolla, Denovo-Pinal)

• Control via structural priors (e.g., helical content, radius of gyration)

• Composable pipelines with feedback from structural, energetic, and developability metrics

Users can design protein binders or AMPs, optimize them for affinity and stability, and automatically generate candidates for downstream testing.

Molecular Dynamics and Simulation

Aura integrates molecular dynamics (MD) simulations using:

• OpenMM engine for structural stability

• Fine-grained atomistic simulations for peptide-membrane interactions

• Visualization and scoring of conformational ensembles

These tools are accessible via prompts like:

"Simulate this protein-ligand complex with OpenMM for 10ns and return interaction energy and RMSD."

Functional Genomics with Tahoe-100M

Fastfold integrates Tahoe-100M, a 100M-cell atlas of cancer perturbation data, directly into Aura.

Users can query the dataset in natural language:

• "Compare TP53 expression in LoF vs non-LoF samples in skin cancer."

• "Find gene expression changes after compound X treatment."

Aura instantly returns curated tables, plots, and stats, abstracting away database wrangling or code writing.

This unlocks large-scale single-cell data for functional analysis, drug screening, and mechanistic insight.

Agent Modes and Scientific Reasoning

Aura adapts to domain-specific needs via specialized agents:

Aura-X-assistant: Structure & function (Folding, design, PAE/pLDDT, function prediction)

Aura-Z-thinking: Advanced modeling (MD, interaction analysis, structural scoring)

Aura-TX-gemma: Therapeutics (Toxicity, DTI, PubMed search, antibody affinity)

Aura-Tahoe-100M (in beta): Genomics & transcriptomics (Natural language access to perturbation data)

Aura-designer-1: Use Boltz-2 to design protein binders with multiple objective functions.

Aura agents chain tools together intelligently, combining reasoning, prediction, and validation without manual orchestration.