computer-aided drug design

Question 1

ligand-based methods

Answer 1

• list of molecules known to have some activity
  
  • sometimes also inactive ones
• develop QSAR models
• screen QSARs and test for activity
• evaluate best hits

Question 2

QSAR

Answer 2

• quantitative structure activity relationship
• linear or 3D
• equation connecting chemical structure to biological activity
• uses experimental data and set of variables to describe structure

Question 3

use of QSAR models

Answer 3

• establish correlation between activity and properties of drugs
• determine contributions of properties
• predict activity of untested drug molecules
• only work on congeneric datasets
  
  • common core and different substituents
  • not scaffold hopping

Question 4

3D QSAR

Answer 4

• superposed set of molecules
• comfa program
• use steric relationship between pharmocophore and ligand
• correlation between 3D structure of set of molecules and activity
  
  • limited to similar molecules

Question 5

cresset

Answer 5

• determines where receptor-interacting features are likely to be
• decide which ligand shape fits best
• use in 3D QSAR to generate new molecule and evaluate
• need shape of field and molecule

Question 6

INDDEx

Answer 6

• 3D QSAR method
• series with known activity
  
  • fragment into substructures
  • ML to derive rules for QSAR
• logic-based
  
  • molecule is active if:
    
    • positive charged centre
    • sp2 orbital nitrogen 5.2A away
  • better than equation

Question 7

ligand based screening

Answer 7

• use QSAR to find new active molecules
• create QSAR for set with known activity
• build chemical library of possible molecules for testing
• take each QSAR in set and predict those likely to be high scoring
• test for activity

Question 8

ligand based screening

small molecule databases

Answer 8

• ZINC
  
  • mathematical library of 20m molecules
  • 10m drug like (worth screening)
• databases store:
  
  • molecule formula and 3D coordinates of lowest energy structure

Question 9

ligand based screening

evaluation

Answer 9

• training data and QSAR
• from zinc take out molecules you’ve learnt on
  
  • distinct training and testing set
• test best hits
  
  • top of list has good number of active molecules
  • is this by chance?
  • how many times better than by chance are these molecules here?
• enrichment factor

Question 10

DUD

Answer 10

• database of useful decoys
• 40 typical drug targets
  
  • each has <500 active molecules
  • 500 - 15,000 inactive decoys
• provides challenging testing set
• decoys:
  
  • similar MW and number of H bonds
  • focused and difficult to test on
  • if successful here the program is powerful
• don’t waste time on easy areas

Question 11

enrichment factor

Answer 11

• (no of actives in top X%/total in top X%)/(total no actives/total no of molecules)
• ie fraction of actives in top X% divided by fraction actives in whole set

Question 12

ligand and receptor docking

Answer 12

• when receptor structure known but ligand unknown
• docking algorithm
• can consider all locations or restrict to a binding site/pocket (e.g. enzyme)
  
  • ideally want to minimise number of molecules to test

Question 13

docking algorithm

Answer 13

• explore internal rotatable ligand bonds
• may not always adopt lowest energy conformation once bound
• sample range of favourable conformations upon docking
• not lock and key
  
  • protein side chains can rotate
  • some algorithms alter main chain too

Question 14

ligand receptor docking

workflow

Answer 14

• get structural info
  
  • px/NMR
  • homology modelling (if high identity, esp at active site)
• identify potential binding regions
• generate ligand-protein conformations
• identify preferred binding conformations
• optimise conformations and score

Question 15

docking

degrees of freedom

Answer 15

• with rigid body docking:
  
  • 6 DoF
  • free movement around an axis
  • translaitonal and rotational freedom
• flexible docking:
  
  • many more DoFs to search

Question 16

FPOCKET

Answer 16

• narrows down search by pcoket identification
  
  • most active sites on largest surface cleft
• alpha spheres that touch 4 protein atoms
• flatter surface gives larger sphere
• search for cluster of spheres of the right size to identify pockets
• put coordinates into server tp determine pocket location

Question 17

docking

scorig

Answer 17

• different methods consider different features
  
  • VDW and electrostatics
  • hydrophobic effect
• simple schemes based on no of contact between different atom types
  
  • e.g. methyl with methyl
  • derived empirically by counting how many times this occurs
  • favourable score if often
  • more lenient

Question 18

AutoDock vina

Answer 18

• simplified scoring function
• models steric, hydrophobic and H bonding
• allows rotation of bonds in ligands and side chains
  
  • not main chain
• dock in 10 mins

Question 19

virtual screening/docking

benchmarking

Answer 19

• how many hgihs coring molecules are active
• dock ligand to protein and create series of poses
• each has associated score
• test each one
  
  • how close to px structure is best scoring pose
  • how close are top 10 scoring poses
    
    • improves prediction quality
• involve EFs
  
  • highest fold improvement at 0.5% EF
• predict binding affinities

Question 20

docking

limitations of evaluation

Answer 20

• target protein structure is in correct conformation for docking
  
  • don’t know the structure after docking
    
    • unrealistic scenario
• overcome with blind trials
  
  • dock into closed conformation (bound to a different ligand)
  • evaluate correlation of predicted and actual binding affinity
• still can’t cover all 1 billion molecules
  
  • 1 million/day

Question 21

known docked structures

process

Answer 21

• solve structure series of protein and ligand docked
• examine computationally with molecular graphics
• detailed energy calculations
  
  • VDW, electrostatics
• map pharmacophore to suggest substitutions
• suggest ligand modifications to alter binding
• explore different ligands with docking
• e.g. gleevec (leukaemia)

Question 22

allosteric modulation

Answer 22

• recent method of interest
• drug in active site = orthosteric
• but modulators can bind distal sites = allosteric
• opens up drug space
• high specificity
  
  • surface more variable
• can modulate

Question 23

current status

Answer 23

• chemical HTS
• docked structures successfully solved
• docking for suggesting inhibitors
• ligand based screening complements