Skip to main content

Solutions

Start from the problem you are solving.

Each problem family groups the ready-made workflows that solve it, from the simplest pipeline to the most thorough.

  • 9 workflows · 6 stages

    Virtual screening

    Find the compounds in a library most likely to bind your target, then rescore the best hits with a stronger method.

    Explore virtual screening
    • Screen a compound library against a target
    • Train a target rescorer on your measured compounds
    • Screen a library with your target rescorer
    • Find top binding poses with diffusion docking
    • Screen an ultra-large combinatorial library
    • Rescore docking hits with a CNN scoring function
    • Re-predict docking hits with Boltz-2 affinities
    • Re-dock top hits with surface-aware diffusion
    • Validate docking protocol, then screen
    [Virtual screening]Fig. 01
    • Screen a compound library against a targetSMILES → 3D embed → dock
    • Train a target rescorer on your measured compoundsVina + GNINA + PLIP → target rescorer
    • Screen a library with your target rescorerVina + GNINA + PLIP → rescore with your model → top 20
    • Find top binding poses with diffusion dockingSMILES → DiffDock → top poses
    • Screen an ultra-large combinatorial libraryThompson Sampling → 3D embed → dock
    • Rescore docking hits with a CNN scoring functionVina triage → GNINA rescore
    • Re-predict docking hits with Boltz-2 affinitiesVina triage → Boltz-2 re-predict (batch)
    • Re-dock top hits with surface-aware diffusionVina triage → SurfDock (scatter)
    • Validate docking protocol, then screenValidate docking protocol → screen

  • 4 workflows · 6 stages

    Structure prediction to docking

    Dock against a protein with no solved structure: predict it from sequence, find its pockets, then dock.

    Explore structure prediction to docking
    • Dock into a predicted structure without choosing a box
    • Dock against a protein with no solved structure
    • Dock into a cleaned predicted structure and see the contacts
    • Compare docking against a wild type and its point mutants
    [Structure prediction to docking]Fig. 02
    • Dock into a predicted structure without choosing a boxSequence → fold → DiffDock
    • Dock against a protein with no solved structureSequence → fold → dock
    • Dock into a cleaned predicted structure and see the contactsSequence → fold → prepare → dock → PLIP
    • Compare docking against a wild type and its point mutantsPoint mutants vs wild type → fold → dock → compare

  • 7 workflows · 6 stages

    Library design

    Generate or enumerate new compounds for a target or an SAR series, then dock them.

    Explore library design
    • Dock analogs enumerated from your SAR data
    • Grow analogs from your SAR table with matched pairs, then dock
    • Propose R-group combinations with Free-Wilson, then pick the next batch to make
    • Pick the next compounds to make from your measurements, then dock them
    • Generate and dock new ligands for a pocket
    • Build a library from building blocks, filter it, then dock
    • Dock the analogs your own model scores best
    [Library design]Fig. 03
    • Dock analogs enumerated from your SAR dataFree Wilson → 3D embed → dock
    • Grow analogs from your SAR table with matched pairs, then dockMatched pairs → analogs → dock
    • Propose R-group combinations with Free-Wilson, then pick the next batch to makeFree-Wilson proposals → pick the next batch to make
    • Pick the next compounds to make from your measurements, then dock themPick the next batch from your measurements → dock it
    • Generate and dock new ligands for a pocketDrugFlow generate → dock
    • Build a library from building blocks, filter it, then dockReaction enumerate → filter → diversity pick → dock
    • Dock the analogs your own model scores bestReaction enumerate → score with your model → dock the best

  • 4 workflows · 4 stages

    Predictive modeling

    Pull measured activities for your targets, curate them into a clean dataset, train a model that predicts activity, and score your own compounds with it.

    Explore predictive modeling
    • Train an activity model from ChEMBL data for your targets
    • Train an activity model from ChEMBL data and score your library with it
    • Train a reaction yield model from your reactions
    • Enumerate products and keep the reactions your model predicts best
    [Predictive modeling]Fig. 04
    • Train an activity model from ChEMBL data for your targetsChEMBL activities → curate → train MPNN
    • Train an activity model from ChEMBL data and score your library with itChEMBL activities → curate → train MPNN → score your library
    • Train a reaction yield model from your reactionsReactions → atom-map → train a reaction model
    • Enumerate products and keep the reactions your model predicts bestReaction enumerate → atom-map → predict yield → keep the best