How it works
Three levels. One reaction-centric twin.
SPADE Twin links molecular representation, reaction-level kinetics reasoning, and biological relevance. This page explains the public scientific spine - not an implementation blueprint.
01
Molecular level
Represent drug systems so the twin can learn structure-to-reactivity relationships that matter for activation chemistry.
- Structure-based encodings of the candidate
- Chemically informed descriptors (steric properties, local bonding environment, electronic effects)
- Deep learning to capture non-linear relationships between molecular profile and reaction behavior
- Outputs in plain language: activation tendency · intermediate formation · transformation pathways
02
Reaction level
Combine learned molecular representations with reaction-engineering principles to simulate how a therapeutic system changes chemically under defined conditions.
- Kinetic behavior over time
- Identification of rate-limiting (rate-governing) steps
- Conversion profile as the system transforms
- Focus on when, where, and how activation may occur
- Mechanistic reaction understanding is embedded in the twin so physics is retained while predictive scope expands
03
Biological relevance
Connect chemical transformation to downstream biology that matters for prodrug design decisions.
- Transport and disposition context
- Release kinetics of active species
- Functional activity / therapeutic relevance framing
- For prodrugs: how molecular modifications may change activation timing and efficacy-related outcomes
What different users care about
- Chemist
- Clear activation hypothesis, intermediates, rate-limiting intuition, what to synthesize next
- Computational scientist
- Structured outputs, reproducibility posture, exports for downstream analysis
- Student or PI
- Plain-language summary of the path from structure to activation relevance
SPADE Twin is a research concept. Predictions are hypotheses to be tested - not clinical guidance. Experimental validation remains essential.