Skip to main content

Overview

The De Novo Drug Design tool generates novel small molecules that fit a target protein’s binding pocket. It uses state-of-the-art generative models to create drug candidates for hit discovery.

Workflow

1

Upload protein structure

Provide a PDB file by uploading, pasting, or using the example structure. The 3D structure is visualized in the viewer.
2

Define binding pocket

Either:
  • AI detection — The AI automatically identifies the binding pocket
  • Manual coordinates — Specify pocket center coordinates (x, y, z) and radius
3

Configure generation

Select a model and parameters:
  • Model: Pocket2Mol, TargetDiff, DiffSBDD, or DecompDiff
  • Count: Number of molecules to generate
  • Temperature: Creativity parameter (higher = more diverse)
4

Generate

Click Generate. The job runs and produces candidate molecules.
Alex’s screenshot note: Take a screenshot of the De Novo setup wizard showing the protein structure viewer with the binding pocket highlighted.

Results

Generated molecules are displayed in a sortable table:
ColumnDescription
2D DepictionMolecular structure drawing
SMILESSMILES string
QEDDrug-likeness score (0-1)
SASynthetic accessibility (1-10, lower = easier)
Vina ScorePredicted binding affinity
MWMolecular weight
LogPLipophilicity

Molecule Detail

Click any molecule to see:
  • Enlarged 2D depiction
  • All properties
  • SMILES string (copyable)
  • Edit & Re-score — Modify the SMILES and recalculate properties
  • Fragment Growing — Extend the molecule from specific attachment points

Export

FormatDescription
SDFStandard molecular data format
SMILESPlain text SMILES strings

Jobs History

View past generation jobs with:
  • Job ID and date
  • Model used
  • Number of molecules generated
  • Status (running, complete, failed)
Click any job to reload its results.

Guided Demo

Click Guided Demo for a step-by-step walkthrough using an example protein that demonstrates the full workflow.
Alex’s screenshot note: Take a screenshot of the results table showing 5-6 generated molecules with their 2D depictions and property scores.