Skip to main content
All solutions

Solutions · 4 workflows

Developability and ADMET triage

Take a hit list and find the ones worth developing: clean the structures, predict ADMET with applicability-domain flags, gate on liabilities and drug-likeness, then map metabolism, off-targets and whether the building blocks can be bought.

Not sure which? Describe your problem
[Developability and ADMET triage]Fig. 08
Workflows
4
Stages
6
Tools
8
Each line is one workflow. Lines share a station where they run the same step; select a workflow below the map to jump to its details.

Workflows

Pick the one that fits.

Open any of them in the builder: signed out you get a read-only preview, signed in you can add your inputs, see the estimate for your plan, and run. The cost cap is the default; you can change it at launch, and a run pauses instead of spending past it.

  • simple4 steps

    Standardize a hit list and triage it on ADMET liabilities

    Standardize → ADMET liabilities → drug-likeness

    Clean a SMILES hit list (salts, charges, tautomers), predict 41 ADMET endpoints with applicability-domain flags, then keep the molecules whose predicted hERG, Ames and DILI probabilities are below 0.5 and inside each endpoint's training domain, and apply the Lipinski, Veber and PAINS/Brenk drug-likeness gates. These are ML estimates: a molecule outside an endpoint's domain is dropped, not assumed clean.

    You provide
    Molecules (SMILES)
    Runs on
    CPU
    Cost cap
    $5.00 default
    Open in the builder: Standardize a hit list and triage it on ADMET liabilities (opens in a new tab)
  • moderate5 steps

    Map where hits are metabolized and whether their building blocks are in stock

    Standardize → metabolites and in-stock building blocks

    Standardize a hit list and take up to 500 molecules (the first 500 if you give more), enumerate their rule-based Phase I and Phase II metabolites, and propose a one-step retrosynthesis for each; then look the proposed building blocks up in a pinned ZINC20 in-stock snapshot. Metabolites are predictions from reaction rules, and the stock check is a catalogue lookup, not a quote.

    You provide
    Molecules (SMILES)
    Runs on
    CPU
    Cost cap
    $5.00 default
    Open in the builder: Map where hits are metabolized and whether their building blocks are in stock (opens in a new tab)
  • simple4 steps

    Profile off-targets for the hits that clear the ADMET gate

    Standardize → ADMET gate → off-target profile

    Standardize a hit list, keep molecules whose predicted hERG and Ames probabilities are below 0.5 and inside each endpoint's training domain, then rank the human targets whose known ChEMBL actives each survivor most resembles. The off-target table is a report to read beside the ADMET results: it does not gate the list, because a molecule resembles its intended target's actives too.

    You provide
    Molecules (SMILES)
    Runs on
    CPU
    Cost cap
    $5.00 default
    Open in the builder: Profile off-targets for the hits that clear the ADMET gate (opens in a new tab)
  • moderate6 steps

    Rank hits by developability and pick a diverse shortlist

    Standardize → ADMET gate → QED band → diverse shortlist

    Standardize a hit list (docking hits included: give their SMILES), keep molecules inside the ADMET applicability domain with a predicted hERG probability below 0.5, then keep those that pass the drug-likeness gates with QED of at least 0.5 and a synthetic-accessibility score of at most 4, and pick the 10 most structurally diverse (MaxMin, Morgan fingerprints). The cutoffs are ML-based screening thresholds, not a verdict on any molecule.

    You provide
    Molecules (SMILES)
    Runs on
    CPU
    Cost cap
    $5.00 default
    Open in the builder: Rank hits by developability and pick a diverse shortlist (opens in a new tab)

Related tools

Run any step on its own.

The tools behind these workflows also run individually.

  • Molecule Standardizer

    Salt-strip, neutralize, tautomer, and stereo cleanup to parent structures

  • ADMET Predictor

    Predict 41 ADMET endpoints, including solubility, permeability, CYP, hERG, and toxicity, from SMILES

  • Drug-likeness & Structural Alerts

    Score and filter molecules with Lipinski, Veber, QED, SA score, and PAINS/Brenk/NIH alerts

  • Metabolite Enumerator

    Predict ranked phase I and phase II metabolite structures with parent and pathway provenance

  • Retrosynthesis

    Propose ranked one-step or short-route retrosynthetic disconnections and building blocks from target SMILES

  • Purchasability Screen

    Check which molecules are listed as in stock in ZINC20, with vendors and catalog numbers, by exact, parent or connectivity InChIKey match

  • Off-Target Predictor

    Rank the human protein targets a molecule most resembles known actives of, with calibrated probabilities and the evidence behind each

  • Fingerprint Similarity & Clustering

    Tanimoto neighbors, Butina and Leiden clustering, MaxMin diversity picks, and UMAP maps from fingerprints