1
Q
Overview of Computer Aided Drug Design
A
- drug design based on ligand shape and x-ray structure
- virtual screening
- pharmacophores
- rigid or flexible docking
- lead compounds identified
- assess compounds by assays or x-ray structures
2
Q
CADD Steps
A
- medical need identified
- relevant mechanism determined using structure function studies
- mechanism or high throughput based screens of libraries
- hits and lead compounds found
- drug candidates identified
- preclinical and clinical testing
3
Q
Identify New Hits
A
- high throughput screening
- virtual screening
- or a combination of both
4
Q
High Throughput Screening Limitations
A
- mostly used in pharma but is expensive and yields false positives (hit appears to inhibit the protein when it actually doesn’t)
- even largest libraries of compounds are miniscule in comparison with the possible chemical diversity space (many missed options)
5
Q
HTS Method
A
- HTS dispenser (computer/automatic)
- biological activity meter (assays)
- interaction analysis (isothermal titration/SPR/yeast genetic interaction
- X-ray crystal structure analysis/docking simulation
6
Q
Computer Aided Drug Discovery
A
- in silico screening
- ligand based or structure based drug design
- structure based methods require 3D information of target
- ligand based methods used when target structure is not known : uses information about the known binding molecules
7
Q
LBDD Steps
A
- virtual combinatorial library design (select molecular scaffold)
- pre processing library
- filtering
- virtual HTS
- LBDD virtual screen
- similarity searching : 2D/3D fingerprint, volume, pharmacophore
- hit selection and biological evaluation
8
Q
SBDD Steps
A
- non combinatorial library
- pre processing library
- filtering
- virtual HTS
- SBDD screen
- preparation of targets
- docking
- hit selection and biological evaluation
9
Q
Ligand Based drug discovery
A
- takes into account different kinds of molecular information like 3D molecule shape, molecular and electronic properties, and 3D pharmacophore
- similarity search of a drug database and ranking of hits
10
Q
LBDD Strategies
A
2D fingerprint:
- identify sequence and the molecular and electronic properties
- classify library compounds as active/inactive
3D fingerprint:
- molecular shape and pharmacophore
- ranking library compounds
11
Q
LBDD Methods
A
- similarity searching
- pharmacophore mapping
- machine learning
12
Q
Similarity searching
A
- structurally similar molecules tend to have similar properties
- given active reference structure rank compounds based on similarities
- top ranking compounds tested for activity
- need a way to measure similarity between compounds quantitatively
eg. B-lactam ring in compounds
13
Q
Similarity measure
A
- molecular descriptors like 2D properties and physicochemical properties
- maximum common substructures like 3D properties
- similarity coefficient measures the similarity between two sets of molecular descriptors
14
Q
2D Fingerpting
A
- fingerprint is a part of the molecule similar throughout the search
- each bit in the bit string represents one molecular fragment
- the bit string for a molecule records presence as 1 and absence as 0 for each fragment in the molecule
- similarity assessed by comparing number of bits that are common between the two structures
- no information about arrangement of the groups or their 3D structure though
15
Q
Hashed Fingerprint
A
- chemical hashed fingerprint of the molecule is a bit string containing information on topology/structure
- stereochemistry not considered
- score of a compound is based on which groups are near what/set parameters for topology
- careful of bit collisions: pattern for describing the same group many times (ie. same bit set by many patterns)
16
Q
Similarity Coefficient
A
- Tanimoto coefficient
- set the coefficient to be what you want for similarity between original and hit compounds
= number of c bits set in common in reference + database structure / (number of a bits in reference structure + number of b bits in hit structure - number of c bits)
17
Q
Scaffold hopping
A
- find molecule as active as query one but with different core structure
- requires 3D searches
- rather than being based on chemistry the similarity is based on shape
- ie. molecules giving the same effect with different core structures
- need to identify the pharmacophores
18
Q
3D Fingerprinting
A
- identify groups of interest and find similar groups giving the same properties
- search how functional groups link in 3D space
- identify triangles of topology: triplets of atoms and their associated distances
- many triangles are identified with different properties like H bonding, hydrophobic/philic, etc
- pharmacophore pairs and triplets
- take into account valence and torsion angles
19
Q
Pharmacophore
A
- substructure of molecule responsible for its pharmacological activity
20
Q
3D Shape Search
A
- molecules aligned in 3D
- similarity score based on common volume
- conformational plasticity of compound taken into account
21
Q
Cons of 3D Search
A
- lots of databases use crystal structures
- might not reflect conformation in the solvent
22
Q
Types of Similar Compounds
A
- similar in 3D not in 2D
- similar in 2D but not in 3D
- identical in 2D but not 3D
23
Q
Scoring
A
- for a reference with known activity search it against a database containing other actives and decoy compounds
- good similarity measure clusters the known actives at the top of the ranking
- not ideal for scaffold hopping
24
Q
Pharmacophore generation and searching
A
- pharmacophore is a set of geometrical constraints between specific functional groups enabling the molecule to have biological activity
- protein structure not required
- assumes that all of the known actives bind to the same location
- generation identifies pharmacophoric features and finds the geometrical arrangement
- searching finds matching database molecules in low energy conformation
25
Database search
- select representative set of actives
- prepare molecules and generate 3D structures/conformations
- use pharmacophore software to generate pharmacophores
- select models and validate
- validate by inspecting if the actives fit the pharmacophore model
- database search should be compatible with parameters used to generate pharmacophore
26
Database Search Principle
- pharmacophoric features in each ligand identified
- ligands aligned by corresponding features
- conformational space explored
- scoring system
27
Inhibition of Thymidylate Synthase and Dihydrofolate Reductase
- key enzymes in folate metabolic pathway needed for -NA synthesis
- TS catalyses dUMP methylation involving transfer/reduction of carbon with production of dTMP (DNA building block)
- DHFR catalyses first step in biotransformation with NADPH as reductant
- generated phamacophore model and search *for what*
28
QSAR
- quantitative structure activity relationships tries to establish quantitative relationships between descriptors and target property capable of predicting activities of novel compounds
- requires descriptors accurately conveying chemically relevant information to machine learning models before an assay
- based on existing functional group information
29
ChemInfo Suite
- machine learning ranks for toxicity and bioavailability
1. molecular descriptors for molecules set
2. data set generation and feature selection
3. machine learning models
4. quality measure assessment
5. virtual screening
6. similarity analysis
30
Three Methods of SBDD
- structure and known inhibitor design
- virtual high throughput screening
- de novo design
31
Structure and known inhibitor design
- known inhibitor modified to improve binding affinity/selectivity
32
Virtual HTS
- docking of small molecules in crystal structure
33
De novo design
- molecule designed from scratch to bind in active site
| - fragments docked and ligated to created full molecules
34
SBDD Principles
- solve x-ray structure of target at high resolution
- solve structure in presence of the substrate (solve binding pocket)
- use structure as template for screening of small molecules
- redesign the molecule to fit or bind better
35
vHTS
- docking
- in vitro high throughput screening or from known substrates/inhibitors to identify leads
- identify chemical groups of the ligands important for binding and specificity (pharmacophore)
- avoid side effects via modification
- test assay (enzymatic and pharmacokinetics)
- solve crystal structures to verify binding mode
36
SBDD Challlenges
- energetics of protein ligand interaction are complicated
- flexibility of protein/ligand complicates structural analysis
- many target binding ligands not good drug candidates (hydrophobicity)
- structures of many important drug targets difficult to determine
37
Molecular Structure and Properties
- narrows down the library
- 1D descriptors: chemical composition and physicochemical properties
- 2D descriptors: chemical topoloy
- 3D descriptors: shape/volume, surface area, pharmacophore
38
Structure Activity Relationship
- correlations constructed between the features of chemical structure in a set of candidate compounds and parameters of biological activity
- identify important groups of the lead compound important to biological activity
- test for biological activity and compare with original to identify whether functional groups are important
39
Lipinski's rule of Fives
- used in drug discovery
1. MW < 500
2. fewer than 5 H bond donators
3. fewer than 10 H bond acceptors
4. calculated logP between -1 and 5 (indication of hydrophobicity of compound)
- too many H bonds give too high of an affinity
- logP found by determining where the compound partitions between hydrophobic solvent and water
40
Virtual Docking Screen
- identifies potential lead compounds from a database (score, rank, filter)
- pose prediction: orientation in binding site and how it interacts (residues)
- need a high resolution structure
- dock compound by selecting proper stereochemistry
41
Search Algorithms
Rigid docking
- ligand is treated as a rigid structure and only translation/rotational degrees of freedom are considered
- assumes rigid protein
- each ligand conformation docked separately
- speedy
42
Rigid Docking
- map active site with spheres
- make triangle of spheres occupying the full active site
- satisfy by docking compounds matching the shape
- orient ligand in active site using simple geometry descriptors
- plausible conformations of a ligand satisfying stereochemistry and match the binding site shape
43
Scoring
- large number of different docking poses need to be ranked based on accuracy and based on receptor affinity
- best binding pose selected based on satisfying shape and chemical groups
44
First Principle Scoring
- gives total number of interactions
- use intramolecular interactions : bond lengths/angles/dihedrals and intermolecule interactions : polar and nonpolar
E binding = E intra + E nonpolar + E polar
45
Empiriccal Scoring
- rapid scoring of ligands
* * see formula **
- quantify Gibbs free energy of binding to the active site
- uses a penalty function aimed at penalising unfavourable interactions
46
Flexible Docking
- accounts for ligand flexibility along torsion angles
- sample all types of torsion and dihedral angles
- different binding models
- protein itself is rigid
- in silico docking doesn't work for an induced fit binding model (protein changes shape) with large conformational changes
47
Docking Considerations
1. need minimum 2A high resolution structure to identify waters
- need water identified in the binding pocket
- electron density for flexible side chains weak
- hydrogen bonds missing
- pKa of active site side chains is influenced by buffer
- ligand protonation/tautomeric form influenced by buffer
2. take into account all rotations, torsion angles, rotatable bonds, ring conformations
48
Structure based fragment screening (de novo)
1. link/grow strategy: dock small compounds to satisfy site and link
2. lattice strategy: active site represented as a lattic with points to satisfy as the ligand
49
De novo building process
1. binding site comprising of three binding pockets
2. crystal screening locates fragments binding to pockets
3. lead compound designed by organising all fragments around core template
4. growing out of a single fragment
50
Lead identification via fragment evolution
- need information about binding mode of initial fragment
- lead binding fragment evolved by building away and modifying
- used as a scaffold to increase the occupancy of the site
- affinity increases as it binds more
51
Lead identification via fragment linking
- individual fragments bind to pockets and are joined by a linked group
- lead molecule now spans the full site
- linked has a increased affinity
52
Lead identification by self assembly
- only used with enzymes
- fragments self-assemble
- protein used to self select and catalyse synthesis of its own inhibitor without covalent attachment of protein to inhibitor
- add groups recognised as substrates
- adduct formation crates bonding between the fragments
53
Lead progression
- optimise or modify properties of lead compound
- reengineering to address optimisation of a particular property
- alter compound's properties to increase polarity, hydrophobicity, etc
- measure delta G in silico
54
Fragment based vs. HTS
Fragment Based
- emphasis on efficiency
- less compounds screened
- hits have clearly defined binding interactions
- protein structure based information measure binding interactions
- design intensive
- need knowledge of protein structure
HTS
- emphasis on potency
- screen many compounds
- hits can contain functional groups used for scaffolding
- need to validate hits
- resource intensive
55
Thymidylate synthase
- maintains the dTMP pool critical for DNA replication and repair
- enzyme target for cancer chemotherapeutic agents
- mimic substrate to inhibit
- identification from docking
- further in silico screening and synthesis
*** further revision *****
56
Lead Criteria
- pharmacodynamic properties
- physiochemical properties
- pharmacokinetic properties: rate of distribution, metabolism, etc.
- chemical optimisation
- patentability
- potency
- efficacy
- pharmacodynamics: mechanism of action, concentration of drug vs. effect, etc
57
Example of SBDD
- Indinavir and squanivir
- protease inhibitors designed by modelling chemical structure on computer to fit inside active site of HIV 1 proteases using crystal structure of enzyme
58
Tamiflu
- use knowledge of sailic acid binding to in silico dock and modify Tamiflu to form more bonds and bind with tighter affinity
Protein Science
flashcards
Decks in class (11)
# Cards
