Drug Discovery Solutions Drug Discovery Solutions
down

Compound Prioritization

Compound Prioritization in the R&D Process

Compound ideas can emerge from different sources, such as computational methods, the brilliant mind of a medicinal chemist, or a team brainstorming session in the afternoon. They can even be the results of a virtual screening campaign of thousands or millions just waiting all to be purchased. The proposals are on the screen and the hardest part comes next: Since resources within a company are also finite, one must choose a manageable number of options from all possibilities to move forward with.

The best candidates must be given higher priority.
During compound prioritization, many parameters are considered that influence the decision-making process. Here, we discuss the key aspects to keep in mind when making decisions and how to significantly accelerate the process.

Compound Prioritization in Virtual Screening:
Assessment Parameters Used in Structure-Based Drug Discovery

The result of a virtual screening is a ranking of all docked molecules based on the docking score. It would be wonderful if the docking score were the absolute truth and you could simply select the top 100 compounds to test. However, this is unfortunately not the case, as the generated docking poses are only the result of an agnostic algorithm that can lead to predictions that do not accurately reflect reality.

Therefore, it is necessary to perform a visual assessment of the higher ranks to verify the plausibility of the generated results. Additionally, it is important to consider the chemical diversity of the selected compounds to avoid getting stuck in a dead-end of only a few chemotypes that end up being inactive.

This paper addresses this topic and outlines many parameters that can be considered during the assessment of the binding mode to evaluate a pose.
3D parameters
  • H-bond network
  • Interactions with particular residues
  • Interactions with (conserved) water molecules
  • Shape complementarity to the binding site
  • Ligand clashes with the target or itself (intra- and intermolecular clashes)
  • Molecular torsions/strains
  • Similarity to known binding modes
  • Protonation/tautomerization of ligand and side chains
  • Unsatisfied ligand heteroatoms
2D parameters
  • Novelty, new chemotype
  • Lipophilic ligand efficiency (LLE), ligand efficiency (LE)
  • Structural diversity
  • Predicted binding affinity
  • ADME properties
  • Unwanted/reactive substructures
  • Solubility

Understand Your Complex and Make Informed Decisions

The parameters mentioned above are integrated into our software to enable the assessment of a generated docking pose. For instance, it is possible to visualize the contributions of individual heavy atoms to the overall binding affinity. This allows users to quickly identify regions that do not form favorable interactions with the target, such as a poor match of lipophilic groups in a hydrophilic cavity.

Molecular torsions can also be highlighted. Based on empirical data from the PDB, unusual angle arrangements are flagged, allowing high-energy conformations to be immediately identified.

A traffic light system provides a clear overview of the results in tabular form. Additionally, further filters can be applied to narrow down the search, allowing you to focus solely on the best of the best.
BioSolveIT software for virtual screening:
  • SeeSAR: Visual, drug design dashboard for computational and medicinal chemists.
    SeeSAR's Docking Mode comes in two versions: Docking can either be run on the same machine or sent to an external unit using the External Docking Mode, which utilizes its own resources for computations. Once the results are ready, they can be further analyzed in the Analyzer Mode based on user-defined criteria.
  • Chemical Space Docking™: (as SeeSAR's Space Docking Mode)
    C-S-D is a the next generation of structure-based virtual screening. In this innovative approach, ultra-vast Chemical Spaces, containing billions or even trillions of entries can be screened for the promising candidates to bind at the target structure.
  • HPSee: Scalable virtual screening workflow environment.
    HPSee allows the upload of molecule libraries that then can be subsequently used for virtual screening runs. Within the interface molecule libraries and user can be managed.
Command-line tools for virtual screening:
  • FlexX: Docking algorithm.
  • HYDE: De-solvation-aware scoring algorithm.

ADME Properties

In partnership with Optibrium, it is also possible to take it a step further and calculate ADME properties using the available models.
The models include: CYP2C9 pKi, CYP2D6 affinity category, blood-brain barrier classification and CNS penetration, HIA category, P-gp category, PPB90 category, hERG pIC50, logD, logP, logS, logS at pH 7.4.

Users can also import their own calculated models into our software to cover additional parameters.

Read more about Optibrium property prediction by reading this PDF.

SeeSAR's Analyzer Mode

The Analyzer Mode is a workspace for working with a set of molecules and filtering them according to desired criteria. For instance, it provides the ability to link docking data with biological activities in a so-called HYDE performance check to establish the correlation between calculated binding affinities and real-life results. Additionally, you can display only the best result from a series of generated poses, allowing for a more efficient review of the outcomes from a virtual screening.
In addition to a wide range of useful filters, the Analyzer Mode also allows grouping compounds into categories using predefined filters.
  • Drug-likeness (Rule-of-five/RO5): Number of h-bond donors ≤5, number of h-bond acceptors ≤10, MW ≤500, logP ≤5
  • Lead-likeness: Number of h-bond donors <4, number of h-bond acceptors <7, number of rotatable bonds <8, number ring rings <5, number of hydrogen-bond donors AND acceptors between 1 and 9, number of heavy atoms between 10 and 27, logP between 0 and 4
  • Fragment-likeness: Number of h-bond donors ≤3, number of h-bond acceptors ≤3, MW <300, logP ≤3

Available Filter Options

Ligand-based parameters:
  • Molecule name
  • Molecular weight
  • TPSA
  • Number of h-bond acceptors
  • Number of h-bond donors
  • Number of heavy atoms
  • Number of aromatic atoms
  • Number of nitrogen and oxygen atoms
  • Number of rings
  • Number of aromatic rings
  • Max ring system size
  • Number of halogens
  • Number of stereo centers
  • Number of stereo bonds
  • Number of rotatable bonds
  • Total charge
  • Covalent
Software-based parameters:
  • Visible in 3D
  • Favorites
  • Annotations
  • Active status
  • Optimization state
  • Import source
  • Estimated Affinity
  • Lipohilic ligand efficiency
  • Ligand efficiency
  • Torsion quality
  • Intra molecular clash
  • Inter molecular clash
  • LogP
  • Number of unfavorable torsions
  • Number of H-bonds
  • Scored/not scored

Filtering in 3D: Pharmacophore Constraints

Pharmacophore constraints are useful 3D filters that can either be applied at the beginning of a docking process to generate only poses that meet specific requirements or used after a docking run in SeeSAR's Analyzer Mode to filter the results based on desired binding modes.

Following pharmacophore constraints are available in SeeSAR:
  • H-bond acceptor/donor interaction contact
  • H-bond donor/acceptors
  • Acceptor interaction contact to metal
  • Aliphatic
  • Specific heteroatom type: C, N, N or O, O, halogens
  • Aromatic CH
  • Aromatic
  • Bicyclic
  • Covalent warhead
  • Hydrophobic
  • Linker constraints
  • Ring atoms
  • User-defined SMARTS
  • Spiro center

Excited for more drug discovery solutions?

Keep on exploring!