We use state of Active-learning techniques to perform virtual screening of large libraries. We routinely perform virtual screen of our 10 million compounds from vendor library (e.g. eMolecules, Molport, Enamine Real). The list of techniques include

  • Docking based virtual screen using Active learning
  • 3D Shape based virtual screen
  • 2D fingerprint based virtual screen
  • Pharmacophore based virtual screen
  • Mountain

    Active Learning for Virtual screening

    To be able to screen larger number of compounds using high precision docking we perform an Active learning strategy. Our active learning strategy is adapted from literature. Active learning strategies are run on Azure cloud infrastructure to utilize high performance computing using state of art GPU enabled virtual machines. Depending on the needs of the project explore and exploit parameters are chosen to screen compounds with right balance between activity and diversity of compounds. We perform range of cheminformatic analysis techniques to filter synthetically tractable compounds to be progressed for single shot or dose dependent IC50 determination.