Polypharmacology Browser 4

Predict active and inactive targets across ~8,000 ChEMBL targets from a chemical structure.

~8,000 targets 5.8M interactions ChEMBL 35 active + inactive models

Query molecule

Prediction settings

Targets the molecule is likely to hit, and those it is likely not to.

Predictions typically run in < 5 seconds

What it does

PPB4 predicts which protein targets a small molecule is likely to interact with, based on multitask deep neural networks trained on ChEMBL bioactivity data.

What's new in PPB4

  • Confirmed-inactive prediction
  • ChEMBL 35 (5.8M interactions)
  • MAP4 fingerprint support
  • Consensus model
  • Combined model predicting active, inactive or untested

Run it locally

The application and its Dockerfile are on GitHub, and the trained models can be downloaded separately.

MaedehDarsaraee/PPB4

Cite us