Docking and virtual screening

Question 1

Pharmacophore complementary to HIV protease active site

Answer 1

Active site has a backbone carbonyl oxygen (Gly27), carboxyl oxygen (Asp25), backbone amide NH and alkyl R side chain (Ile50)
Therefore the complementary pharmacophore would have a H-bond donor to interact with each of the oxygens, a H-bond acceptor to interact with the NH and a hydrophobic group to interact with R
Then all appropriately positioned to fill the available space

Question 2

Methods for identifying suitable drug candidates

Answer 2

1. Database screening

2. De novo design

Question 3

Database screening

Answer 3

When known chemicals from a computational database are docked into the receptor target site in order to find a best fit
A scoring function for docked structures is used to rank the compounds and identify potential hit candidates

Question 4

De novo design

Answer 4

Involves generating novel compounds computationally and docking them into the receptor target site in order to find a best fit

Question 5

Molecular docking

Answer 5

Refers to a prediction of the binding mode of a compound to the active site
Also predicts the strength of association

Question 6

What must be done before virtual screening/docking can be used?

Answer 6

Need to assign the charge, tautomeric state and conformation of the ligand (will also need to be aware that these may be different in the protein environment)
Need to assign protonation state of protein
Also need to consider the flexibility of both the ligand and protein - although simple methods often neglect protein flexibility because it is time consuming

Question 7

Steps involved in the first stage of a typical docking process

Answer 7

1. Because ligands often change their conformation when binding to proteins, many programs take the torsional flexibility of the ligand into account by allowing specific bonds to be able to rotate. Alternatively, the program may perform rigid docking of a predetermined set of ligand conformations
2. Many conformations of the ligand-protein complex are generated for each molecule being considered - each generated ‘snapshot’ of the ligand-protein complex is referred to as a pose
3. Each pose is assessed using a scoring function
4. The orientations that score well are kept and the remaining low-scoring poses are disregarded

Question 8

Why can some of the large number of potential poses be immediately rejected?

Answer 8

Due to clashes with the protein

Question 9

Ideal scoring function

Answer 9

Would take the binding mode produced from molecular docking and give an accurate estimate of the free energy of binding (DeltaGbind)
However, in practical applications of virtual screening, a crude scoring function is used (such as molecular mechanics functions or empirical scoring functions)

Question 10

Empirical scoring functions

Answer 10

Developed to reproduce experimental binding affinity data
Based on relatively simple descriptions of intermolecular interactions
The scores can be computed quickly to rank large libraries of diverse compounds that have been docked into the active site of the target protein

Question 11

Example of an empirical scoring function

Answer 11

AScore

Question 12

AScore predicts DeltaGbind using terms that describe:

Answer 12

1. Van der Waals interactions: non-covalent interactions described by similar equations to those used in molecular mechanics but with summations over ligand-protein and ligand-ligand atom pairs
2. Hydrophobic effects: summation over hydrophobic ligand-protein atom pairs
3. Hydrogen bonding: H-bonding is perhaps the most important factor for the specific binding of a ligand to its receptor. The interaction occurs when 2 atoms are in close enough proximity to form a donor-acceptor pair. The geometry of the H-bond (D-H—A) is typically described by the bond angle and the bond length (i.e. distance between H and acceptor A)
4. Deformation: upon binding, both the ligand and protein are constrained in conformation compared to their free states in solution. This leads to adverse entropic changes, which is a negative effect that must be overcome during the binding process

The parameters can then be modified to improve the correlations between the predicted and experimental DeltaGbind values for a test set of related protein-ligand structures

Question 13

Steps involved in the second stage of a typical docking process

Answer 13

1. Different poses, belonging to either the same or different ligands, are ordered according to their computed score
2. The energy of the top scoring poses are recalculated using more accurate, computationally more intensive scoring functions

Question 14

SkelGen

Answer 14

A sophisticated program used for de novo drug design
Has a library of 1678 fragments generated by fragmentation of known synthetically accessible structures
A ligand is initially created by linking randomly chosen fragments from the fragment library and placing the ligand at a random position in the binding site
The structure evolves and is modified in a stochastic process - the impact of each modification is determined by feedback from docking and scoring
Run multiple times to produce a series of chemical scaffolds designed to fit into the active site

Question 15

Advantage of SkelGen

Answer 15

Provides access to around 1 trillion low molecular weight, drug-like molecules, compared to only a few million structures in a typical chemical library

Question 16

Example of when SkelGen was used

Answer 16

In the development of novel oestrogen receptor ligands
SkelGen was used to make structures by combining 1700 fragments from molecules in the World Drug Index
Docking and scoring led to 17 high-scoring diverse structures for synthesis
Of these, 5 structures with high binding affinity were generated, the best with IC50 - 340 nM

Question 17

Limitations of database screening and de novo design

Answer 17

A rigid geometry for the receptor and ligand is a major approximation - although nowadays it is common for ligands to be treated as flexible and docked in multiple conformations of the (rigid) receptor to partially overcome this limitation
There are inherent approximations and simplifications used in crude scoring functions (e.g. molecular mechanics has many approximations/assumptions)

Question 18

What type of calculation is used when validating docking programs before performing large-scale virtual screening experiments?

Answer 18

Comparing the docked structure of a ligand with the X-ray structure to see how well they match up

Question 19

ArgusDock features

Answer 19

The flexible ligand is described as groups of rigidly-bonded atoms connected by torsions
The binding site of the protein is overlaid with grids of search points
The ligand is rotated in space at each search point - any ‘bumps’ with the target are removed by small translations, and individual poses of a ligand are rejected if parts of the ligand are outside the binding site
Torsions are explored systematically for each rotation
Poses are scored with AScore and the N lowest poses are ranked (typically 50 < N < 150)

Question 20

Pose

Answer 20

Binding mode of the ligand/drug to the receptor

Question 21

Scoring function

Answer 21

A mathematical function used to approximately predict the binding affinity between 2 molecules (generally a drug and a protein receptor) after they have been docked

Question 22

Limitations of empirical scoring functions

Answer 22

Ignore key steps in drug-receptor interactions
e.g. formation of the initial complex between the binding site of the protein receptor and the drug is firstly determined by a diffusion-controlled encounter rate
Desolvation of both the drug and the binding site must then occur, alongside conformational changes of both when they bind
The free energies of each of these stages cannot be calculated easily