MODEL SYSTEMS 02
Specialists
in a shared world.
Each terminalBio system has a distinct job. Together, they create a composable intelligence layer for designing in biology.
MOLECULAR
INTERACTIONS
BindFM
A foundation model for molecular recognition—reasoning across proteins, nucleic acids, small molecules, and cofactors to reveal the interactions that make binding work.
What binds, why does it bind, and how can we improve it?
Affinity prediction · interface reasoning · cross-modal molecular representation · candidate ranking
Prioritizes molecules to synthesize and assay, then learns directly from measured performance.
GENERATIVE
APTAMERS
XUZU
A generative design system for aptamers that treats sequence, structure, target context, and experimental feedback as one iterative search problem.
Which sequences have the highest likelihood of becoming useful binders?
Constrained sequence generation · structure-aware ranking · diversity selection · wet-lab feedback integration
Generates a small, information-rich library—not an unranked flood of sequences.
RNA
INTELLIGENCE
RNJ / RNX
RNA structure intelligence for reading and designing a molecule whose function depends on a moving, contextual geometry—not sequence alone.
How does an RNA sequence fold, behave, and change in the context where it must work?
Secondary-structure reasoning · RNA representation learning · sequence-to-function analysis · therapeutic design support
Connects predicted structural behavior to expression, activity, stability, and experimental readouts.
COMPUTE
SUBSTRATE
TB Biocompute
The composable infrastructure beneath the model layer: a consistent way to run, compare, connect, and build biological intelligence systems.
How do specialized models become a reliable, reusable scientific system?
Reproducible evaluation · interoperable data and model interfaces · experiment tracking · agent-ready orchestration
Makes every model output traceable, comparable, and available to the next decision.