Drug Design

Modern drug design processes include computational methods to discover, study, and modify biologically active molecules. Computer-aided drug design reduces the number of iterations in the drug design cycle and accelerates invention and optimization processes. In addition, it offers the potential to make more novel, small molecule structures available for the design process.

Due to its roots in quantum chemistry, COSMOlogic software is universally applicable across the complete organic chemistry space. Users can apply it to very different and novel chemical situations.

COSMOsim3D and COSMOsar3D use local σ-profiles as descriptors. σ-profiles are derived from quantum chemistry calculations and describe properties of molecular electronic surfaces instead of chemical structures. Thus, they naturally allow for scaffold hopping and searching for active analogues. Local σ-profiles provide information about

  • electrostatics
  • hydrogen bonding
  • hydrophobic interactions
  • shape

Alignment and active analogue search

Using a quantum-chemistry based molecular surface polarity, COSMOsim3D is a unique and very robust method for field-based ligand-ligand alignment. The COSMOsim3D similarity is a powerful descriptor of ligand similarity. This enables a good discrimination between bioisosteres and random pairs and allows for scaffold hopping and searching for active analogues.

Example for bioisoster screening with COSMOsim3D (PDF)

3D-QSAR

Arrays of local σ-profiles are a novel set of molecular interaction fields. They provide all the information required for quantifying the virtual ligand-receptor interactions, even including desolvation. In COSMOsar3D, this leads to increased predictive accuracy in 3D-QSAR studies, combined with outstanding robustness with regard to grid step size, grid positioning and random misalignment.

Example for the predictivity of a COSMOsar3D based 3D-QSAR in comparison to other methods on the Sutherland data set. (PDF)

ADME properties

With COSMOtherm or COSMOquick, molecular descriptors are derived from surface polarity information and structural parameters. Users can build their own QSAR models based on these physically meaningful descriptors. A number of such ADME models are available for direct use.