Rule-based cheminformatics tool

Molecular Randomizer

Generate one-step small-molecule analogues, prioritize them with RDKit descriptors, and run benchmark lists for method validation.

Run Type

Lists accept SMILES or name,SMILES, with an optional header. Web limits: 50 benchmark inputs or 20 comparison inputs. Use the command-line runner for larger studies.

Molecule Input

Use this field for one detailed analogue-generation run.

Run Settings

In benchmark mode, this controls the detail table shown for each molecule. The all-variants CSV keeps every unique valid variant.

Stepwise Change Types

Terminal replacement exchanges an end atom for a selected group in one step. Protection rules still apply.

Heavy Atom / Fragment Choices

Sampling changes which edits are tried. Scores are calculated descriptors, not predictions of biological activity.

Custom Fragment
Preparation and chemical constraints

All matching atoms are frozen, including attachments at those atoms.

Preparation changes are recorded. Atom numbers refer to the prepared canonical parent. These heuristic rules do not prove stability or synthesizability. Newly created stereocenters may remain unassigned.

Ranking and score sensitivity

Pareto fronts maximize QED and parent similarity and minimize SA and alert count. Incomplete evaluations are placed last. These descriptors do not predict target activity.

Comparative study settings

Select all desired change types above. Complete enumeration runs once per method. Confidence intervals resample parent molecules, averaging repeated seeds within each parent. CReM and external comparator imports are available through the command-line runner.

Optional reference collection

Exports report exact membership and nearest-neighbor similarity. Absence applies only to this collection; it does not establish patent novelty.

Optional measured analogue evaluation

Use one consistent assay and unit per series. Evaluation reports known-analogue recovery and measured improvements among top results. Unmeasured products are not classified as inactive. Keep evaluation series separate from rule and parameter development.

Output

Each generated result applies exactly one stepwise change. The score is descriptor-based prioritization, not a direct claim of biological activity.

Version 2.1.0. Every run saves an actual seed, settings, operation log and source fingerprints in a downloadable ZIP.