Budget-Aware Adaptive Search
Search priorities evolve with every acquired measurement: explore broadly, exploit a promising region, complete a partial profile, or revisit a candidate
SMALL-MOLECULE DISCOVERY
More promising molecules. Fewer expensive evaluations.
MolScout learns from every result to select the next molecule–target measurement, making the most of limited evaluation budgets across large molecular libraries.
Its active-learning approach combines complementary molecular representations, probabilistic target models, and multi-step acquisition planning to balance chemical-space exploration with focused evaluation of promising candidates.
CHEMICAL SPACECANDIDATE SELECTION
Candidate spaceEvaluated candidates
A conceptual projection of molecular neighborhoods. Each acquired result informs the next molecule–target measurement.
01 / FEATURES
The Search Advantage
About Molecular Adaptive SearchLearn from every acquired measurement. Update the search landscape. Put a finite evaluation budget where it creates the most discovery value.
Cross-space search
Bring structural, physicochemical, learned, and response-driven neighborhoods into one coherent search
Multi-target search
Treat each molecule as a partially known profile. Decide which target or endpoint is most valuable to evaluate next
Profile optimization
Express discovery goals through potency, selectivity, molecular properties, and target-profile requirements
Interactive research
Inspect trajectories, candidate provenance, predictions, and the evidence behind each search decision
Agentic Experiment Control
A machine-readable view of experiments, with explicit permissions for observation and intervention
02 / BENCHMARKS
Discovery Efficiency
Explore experiment results across discovery objectives and molecular datasets.
DUAL-TARGETA2AR + D2R
04Enamine Ro4 11M
Experiment in progress
Benchmark results could not be loaded.
· budget
03 / ARCHITECTURE
Fits Your Discovery Stack
Your molecules. Your evaluators.
MolScout decides what to evaluate next.
The result: prioritized candidates with the evaluation evidence behind each compound.
Molecular librariesRegistries · compound collections
GeneratorsInternal · external chemistry
ADAPTIVE DECISION LAYER
Adaptive search · Experiment state · Evaluation budget
Adaptive search · Experiment state · Evaluation budget
EVALUATORS
Docking
FEP
QSAR / ML
Assays
Evidence updates MolScout state & models
SEARCH OUTPUT
Prioritized candidates
Complete molecular profiles · Evaluation evidence
04 / SHOWCASE
See MolScout in Action
This sped-up recording shows MolScout searching a library of 260,000 molecules for selective activity between two targets: JAK2 and LCK.
Each evaluation measures one molecule against one target. A partial profile means only one target has been evaluated. A complete profile contains results for both targets, allowing their activity to be compared.
The frontier is the pool of molecules MolScout is currently considering for evaluation. In each round, it chooses 300 molecule–target pairs from this pool—starting profiles for some molecules and completing profiles for others.
As results arrive, MolScout updates the frontier and decides what to evaluate next. The search continues until it reaches a total budget of 5,000 individual evaluations.
View Details