logo
English (United Kingdom)Ukrainian (UA)
RESEARCH Molecular Design

Molecular Design

Molecular Design Laboratory of Medicinal Chemistry Department carry out investigations in the field of developing of new biologically active compounds using modern computer modeling methods.

  • hit identification using virtual screening
  • hit-to-lead optimization of affinity and selectivity (docking, QSAR)
  • lead optimization


We perform receptor-based and ligand-based virtual screening using chemical library of about 200,000 compounds. Receptor-based virtual screening can be used when 3D structure of molecular target is known. In this case such methods as molecular docking and 3D-pharmacophore modeling can be applied. For molecular docking we use AutoDock and DOCK software. Much attention is paid to ligand pre-processing that is carried out with in-house developed software. Accordingly to docking calcultions we obtain "receptor-ligand" complexes and scores. Docking results interpretation is performed in several steps including consensus scoring of different rescoring functions and visualization of "receptor-ligand" complexes. Molecular dynamic simulations allow investigate conformational changes of "receptor-ligand" complexes.  

Figure 1. Virtual screening flowchart for protein kinase FGFR1.
Figure 1. Virtual screening flowchart for protein kinase FGFR1.

Pharmacophore modeling includes several consequent steps: building of pharmacophore models, their optimization and validation and pharmacophore screening. Pharmacophore model is an ensemble of steric and electronic features that are necessary for ligand-receptor interaction. Receptor-based pharmacophore models can be generated based on the structure of molecular target active site taking into account intermolecular interaction with the ligand. 

Ligand-based virtual screening can be applied if 3D structure of molecular target is unknown, but there are data cocerning compound activity. We use three different approaches for ligand-based virtual screening: 1) QSAR, which is based on proprietary software bases and bases, collected from literature data for specific tasks; 2) ligand-based 3D-pharmacophore modeling, that is based on overlay of compounds and identification of key common properties important for interaction with the receptor; 3) machine learning which includes building of artificial neural network, Bayesian statistics and the k-nearest neighbors algorithm. 

When the structure of molecular target is unknown, but there are data on the compound activity, for example, antibacterial activity, we can use machine learning which allows on the basis of knowledge, obtained through deep analysis of data, to find new potentially active compounds belonging to novel chemical classes. Machine learning can be equally useful for primary virtual screening and optimization of already found chemical classes. 

The critical step in machine learning is intellectual work on the development of computer protocols for each spesific task, the establishment of settings and formation of training sets. Then, computer builds neural networks, Bayesian models, performs basic screening and integral ranking of results. 

Also, the goal of our scientists is the development of new, efficient methods in the field of receptor-based high-throughput virtual screening.

We have discovered and parameterized novel method for generating inexpensive and electrostatically reasonable atomic charges of organic compounds. This method is based on the principle of electronegativity relaxation of the Kirchhoff charge model. Parameters of the method, orbital electronegativities and hardnesses, are fitted to reproduce reference, ab initio calculated dipole and quadrupole moments of a representative training set of neutral and charged organic molecules cover most medicinal chemistry relevant bonding situations. Accuracy of the derived parameters are confirmed on an external test set. The new Kirchhoff charges were implemented into a virtual screening engine.

To increase accuracy of ligand-receptor interaction energy evaluation the Kirchhoff charges calculation model was used for creating a new force field YFF. This force field is obtained by joining Van-der-Waals and bonded part of well-known MMFF94 with our charge calculation scheme. The electrostatic part of YFF has been parameterized to reproduce ab initio calculated dipole and quadrupole moments. The 6-12 Lennard-Jones potential terms were parameterized against homodimerization energies calculated at the MP2/6-31 G* level of theory. YFF was used to develop preprocessing and docking programs - Preprocessor, Topbuilder and Screener.

We have developed and validated the algorithm for pharmacophore model optimization and rescoring of pharmacophore screening results. The algorithm was written in the Java programming language, has a graphical interface and a system of parallelization. The advantages of this algorithm are optimization of pharmacophore features radii, using of pharmacophore features weights (indicate the importance of pharmacophore features) and molecular descriptors (QSAR approach). The algorithm allows significantly improve the quality of pharmacophore models and consequently, the results of pharmacophore screening. It was used to develop program for pharmacophore modeling - PharmDeveloper.

Computer programs used in Department of Medicinal Chemistry:

Compound collection:

  • Jchem
  • CheD

 

Pharmacophore modeling:

  • PharmDeveloper (was developed in Medicinal Chemistry Department)
  • Pharmer
  • PharmaGist

 

Machine learning:

  • KNIME
  • R

 

QSAR:

  • PASS

 

Molecular docking:

  • AutoDock
  • Vina
  • Dock
  • Preprocessor (was developed in Medicinal Chemistry Department)
  • Topbuilder (was developed in Medicinal Chemistry Department)
  • Screener (was developed in Medicinal Chemistry Department)
  • AutoDockMapper (was developed in Medicinal Chemistry Department)

 

Molecular mechanics and dynamics, quantum chemistry:

  • Amber
  • GROMACS
  • NAMD
  • GAMESS

 

Homology modeling:

  • Modeller

 

Visualization:

  • Discovery Studio Visualizer
  • Chimera
  • VEGA ZZ
  • PMV
  • ViewerLite

 

  • Ab initio parameterization of YFF1, a universal force field for drug-design applications. Yakovenko O.Ya., Li Y.Y., Oliferenko A.A., Vashchenko G.M., Bdzhola V.G., Jones S.J.M. J. Mol. Model 2012, 18, 663-673.
  • Kirchhoff atomic charges fitted to multipole moments: implementation for a virtual screening system. Yakovenko O.Ya., Oliferenko A.A., Bdzhola V.G., Palyulin V.A., Zefirov N.S. J. Comput. Chem. 2008, 29, 1332-1343.
  • The development of algorithm for pharmacophore model optimization and rescoring of pharmacophore screening results. Starosyla S.A., Volynets G.P., Protopopov M.V., Bdzhola V.G., Yarmoluk S.M. Ukr. Bioorg. Acta. 2016, 24-34. 

Our research:

FGFR1 Inhibitors
Phenylbenzisoxazoles and aminopyrimidines Image
Phenylbenzisoxazoles and aminopyrimidines were also found from virtual...



 

ASK1 Inhibitors
Identification of 3H-naphtho[1,2,3-de]quinoline-2,7-diones as Inhibitors of ASK1 Image
Virtual screening and biochemical screening allowed us to identify...



 

CK2 Inhibitors
2-phenylisothiazolidin-3-one-1,1-dioxides Image
We have synthesized 40 new 2-phenylisothiazolidin-3-one-1,1-dioxide...



 

Molecular Design
Molecular Design Image
Molecular Design Laboratory of Medicinal Chemistry Department carry out...



 

Fluorescent Probes
Fluorescent detection of partially denatured conformation of beta-lactoglobuline Image
The native state of many proteins is known to be only marginally stable,...



 

© 2012 All rights reserved
footer logo