Computational Design Studio
Molecular Design Studio
Configure an in silico design brief, tune target descriptor windows, and query the learned molecular prior through a glass-lab interface built for medicinal chemistry exploration.
Learned molecular prior
RDKit descriptor screening
3D conformer generation
Protocol Summary
Batch size
50
Candidates sampled per run
Top hits
5
Returned to the analysis lab
The current pipeline uses a learned generative model to sample structures, then ranks candidates against your descriptor envelope before producing viewable 3D conformers.
Design Brief
Protocol Input
Tune descriptors, then launch the run
Design Notes
Prompt heuristics
Terms like oral, CNS, brain, polar, soluble, lipophilic, and fragment-like are currently interpreted as soft biases during scoring.
Screening behavior
The backend keeps molecules inside practical descriptor windows before computing rewards, so unrealistic settings will naturally collapse the candidate pool.
QED
Composite drug-likeness objective
0.90 score
01
logP
Lipophilicity and membrane affinity
2.0 log units
-16
TPSA
Topological polar surface area
70 A^2
0150
Molecular Weight
Mass window for tractable leads
320 Da
150550
Execution Mode
Sampling 50 candidates and returning the top 5 ranked hits.