PIGenet2

Predict protein–ligand binding affinity, reporting the total binding energy and its component terms.

API/CLI tool ID: pigenet2

Understanding inputs and outputs

A ligand is the molecule whose binding to the protein is evaluated. The current input accepts one molecule in one 3D pose per SDF file. Supply multiple candidates as separate files in ligandSdfs. The model evaluates the supplied protein–ligand arrangement; connectivity-only data or an arbitrarily positioned molecule is not a prepared binding pose.

total_affinity and its energy terms use kcal/mol; in this model, a lower total predicts stronger binding. The terms represent modeled contributions from close atomic contacts, hydrogen bonds, metal coordination and interactions involving water-avoiding regions. They are not measured binding constants or clinical effects. Their units and direction differ from PharmacoNet scores, so their magnitudes cannot be compared directly.

Input settings

SettingFormatRequiredDefaultDescription and constraints
proteinPdbpdbRequired—Protein structure — Extensions pdb · cif
ligandSdfssdfRequired—List of SDF files, each containing one molecule in a single 3D pose. — Extensions sdf, JSON array

Output metrics

ColumnDisplay nameRangeUnitDescription
total_affinityTotal predicted affinity—kcal/molSum of energy terms. Lower values predict stronger binding.
vdw_energyvDW energy—kcal/molVan der Waals interaction term
hbond_energyHydrogen-bond energy—kcal/molHydrogen-bond term
metal_energyMetal coordination energy—kcal/molMetal–ligand coordination term
hydrophobic_energyHydrophobic energy—kcal/molHydrophobic interaction term

See Reading results for metric definitions and interpretation. A range shows possible values; whether higher or lower is better depends on the metric.

Execution availability

Open an execution plan in your project to see available tool versions and costs. Review and approve the inputs, settings, and terms before running.

License

Code, model weights, and accompanying data have their own terms. The table summarizes terms from official sources; check your execution plan for the configuration you will use. License texts are available in Third-party notices.

ComponentLicenseCommercial useSourceNotes
CodeMITAllowed, subject to notice requirementsLicense text—
Model weightsMITAllowed, subject to notice requirementsRepository READMEFour pretrained weight files (pda_0~3.pt) are included in src/ckpt in the code repository. No separate terms are specified for the weights.
DataCC-BY-4.0Conditional; see notesDistribution informationThe Zenodo benchmark materials use CC BY 4.0 and require attribution. They are derived from PDBbind, CASF and DUD-E; obtaining the original PDBbind data requires checking its separate academic-use terms.
CodeApache-2.0Allowed, subject to notice requirementsLicense textApplies to the reduce program used to add hydrogens to protein structures.
CodeLicenseRef-PyMOL-PermissiveAllowed, subject to notice requirementsLicense textpymol-open-source requires retaining copyright and permission notices and prohibits using the Schrodinger name in advertising.

Checked on 2026-10-22.

References

On this page