protein ligand docking     molecular modeling and computational chemistry  
  high throughput screening     Enrichment curve and enrichment factor  
 

Lead–Finder

   
   
   
protein structure preparation
 
 

Deviation of the predicted free energy of ligand binding plotted against the experimentally measured binding energy for the set of 330 protein-ligand complexes (blue dots correspond to the training set, red dots — to the test set).







You can download complete description of binding energy benchmarking data.
Accuracy of binding energy predictions achieved by Lead-Finder

   

Accuracy of binding energy estimations

Lead-Finder uses special type of scoring function to calculate free energy of  protein-ligand binding. This scoring function includes all energy terms (described in section Technology) scaled with a set of coefficients, which were adjusted to fit experimentally measured binding energies.

A set of  330 protein-ligand complexes with experimentally measured binding constants and available 3 D structures was used for the current benchmarking study, out of which 100 complexes (the training set) were used to calibrate the scoring function. RMSD between calculated and experimentally determined binding energies comprised 1.24 kcal/mol for the training set and 1.5 kcal/mol for the entire set of  330 protein-ligand complexes. About 50 % of all protein-ligand complexes fall into ±1 kcal/mol bin and 79 % fall into ±2 kcal/mol bin of ΔG prediction accuracy. These data illustrate that Lead-Finder is the most accurate in binding energy calculations compared to other docking programs 1, 2, 3


Set of protein-ligand complexes for benchmarking studies

To make current parameterization and benchmarking studies more robust, protein-ligand complexes with known 3 D structure and experimentally measured binding constants for were chosen on the basis of maximal diversity of ligand’s physicochemical properties (molecular weight (Mw), clog P, number of hydrogen bond donors (dHB) and acceptors (aHB), net charge) and range of protein  ligand binding energies. Physicochemical properties of selected ligands fairly cover chemical subspace satisfying Lipinsky rule of Five 4. However, ligands spanning beyond this subspace (57 compounds with Mw>500; 31 compounds with clog P>5;89 compounds with (dHB+aHB)>10; 41 compounds with dHB>5) are also generously represented, which is very important according to recent findings that physicochemical properties of drugs are not tightly restricted by Lipinsky rule 5. These facts support broad applicability of Lead-Finder for binding energy calculations.


Speed of binding energy calculations

Lead-Finder provides unique speed of binding energy calculations compared to such methods as  free energy perturbation or linear interaction energy, which need long molecular dynamical simulations and generally require individual treatment of each particular task. For the set of studied 330 protein-ligand complexes average time for binding calculations with Lead-Finder comprised less than one second per compound.





Текст
 
  Drug design     Drug discovery