Tema 4.

Introducció Disseny de
fàrmacs assistit per ordinador
Dr. Carles Galdeano
Novembre 2023

cgaldeano@ub.edu
• Molecular recognition

• 3D structures

• Computational hit generation

- Structure-based: Molecular Dynamics simulation

- Structure-based: Virtual Screening (docking)
- Unknown binding site
- Ligand-based drug discovery
Computational Methods in
Drug Discovery

Informa on Computa onal

Technologies Chemistry

Artificial
Intelligence
Drug Discovery
 Molecular Mechanics vs Quantum Mechanics
QM
MM
• E=E(R,r), with R and r being nuclear and
• E=E(R), with R being nuclear coordinates
electron coordinates
• The effect of electrons is included implicitly
• The effect of electrons is treated in an
by means of empirical parameters
explicit way
• Deterministic: subject to Newton’s laws
• Probabilistic: subject to Schrödinger’s
equation

H Y = EY

Advantages of QM
Formally rigorous, without empirical parameters A priori valid to any chemical system
Very accurate at the highest levels of theory
Allows to study chemical reactions and spectroscopic properties

Shortcomings of QM
Very expensive, therefore valid only for molecular systems of small size
 Computer-aided drug design (CADD)
– Hit discovery
Known ligand(s) No known ligand
Known protein

Structure-based drug
design (SBDD) De novo design
Binding site
structure

identification
Protein-ligand docking
(Virtual screening)
Rational Design
(MD simulations)

Ligand-based drug design (LBDD)

protein structure

1 or more ligands
• Similarity searching CADD of no use
Several ligands Need experimental
Unknown

• Pharmacophore searching data of some sort

Many ligands (20+)
• Quantitative Structure-Activity
Relationships (QSAR)
 Computational Methods in Drug Discovery

CADD can be classified into two general categories:

Structure- Ligand-
based based

Structure-based CADD relies on Ligand-based CADD exploits the

the knowledge of the target knowledge of known active and
protein structure inactive molecules through

to select compounds based on ▪ chemical similarity searches

their binding energies.
or

▪ construction of predictive,
quantitative structure-activity
relation (QSAR) models.
 Computational Methods in Drug Discovery

Structure- Ligand-
based based

generally preferred generally preferred

where high-resolution structural when no or little structural

data of the target protein are information is available, as
available, i.e., for soluble proteins often for membrane protein
that can readily be crystallized. targets.
 Obtaining protein structures
• Protein homology
(when structure is not elucidated)

• Protein Data bank:

- X-ray structures
- NMR structures
- cryo-EM Structures
• Alphafold?
 Alphafold 2
(https://alphafold.ebi.ac.uk)
 Alphafold 2
(https://alphafold.ebi.ac.uk)
Obtaining X-ray protein structures
Protein 3D Structures
Cryo-electron microscopy

14
Computational Methods in
Drug Discovery

STRUCTURE-BASED
 Identification of novel hits using CADD
Binding site
Known Unknown

1.Rational design 2.Virtual screening 3. De novo design

(visual inspection and MoA) (input: compound libraries)

BOMB, BREED, LUDI…

DOCKING softwares
Pymol, chimera, VMD.. softwares
Identification of novel hits using CADD

1.Rational Design
Exemple of inhibitors of acetylcholinesterase
Molecular RecogniOon
Molecular RecogniOon
Basic
BasicPrinciples
Principles

• Ligand-Receptor recognition is highly

Ligand-Receptor
• selective & specific
recognition is highl
selective & specific
• High affinity can only be achieved
High affinity
• through cancomplementarity:
excellent only be achieved
through excellent(ionic,
• Electrostatics complementarity:
H-bonds)
Electrostatics (ionic, H-bonds)
• • Steric
• • Hydrophobic
Steric
• Hydrophobic

Given that the rules of molecular recognition

are understood, qualitative assessment of
DDGGivencan
that
bethe rules of
obtained molecular
from recognition
visual inspection
bind
are understood,
of complexes qualitative
or approximate assessment
scoring functions of
DDGbind can be obtained from visual inspectio
of complexes or approximate scoring functio
. Structural waters

Interstitial & displaceable water molecules

• Remaining water molecules must form a Ody

network of interacOons
• Intrinsic preferences + ligand-specific effects

Do we want to interact or displace the highlighted water?

 Identification of novel hits using CADD
Rational Design
A. Protein flexibility → Is the right conformation?
Binding Site Flexibility

HSP90 example
(1BYQ vs. 1UY6)

VISUALIZATION:
Identification of novel hits using CADD
Rational Design
A. Protein flexibility → Is the right conformation?
Molecular Dynamic Other: e.g. Monte-Carlo calculations
simulations
Random movements,
Based in molecular mechanics tend low energy conformations
Molecular Dynamics Simulation
 Identification of novel hits using CADD
Binding site
Known Unknown

1.Rational design 2.Virtual screening 3. De novo design

(visual inspection and MoA) (input: compound libraries)

BOMB, BREED, LUDI…

DOCKING softwares
Pymol, chimera, VMD.. softwares
Identification of novel hits using CADD
2. Virtual Screening
- Virtual screening (VS) refers to the use of computational tools to select
compounds for screening in biochemical assays.

- VS is typically much cheaper than running a high-throughput screening. It

can also be significantly faster and equally or more successful than HTS. VS
cost around 10-fold less than HTS and takes about half the time.

- Academic and small biotech consider VS in their projects.

On demand collections: a new era in drug discovery
 SCREENING VIRTUAL: filtros computacionales
2. Virtual Screening workflow

1. Preliminary filters
to the library

2. Obtation 3D structures
(conformers)

3. Docking process
Identification of novel hits using CADD
2. Virtual Screening
C. Docking
 Protein-ligand docking
• Computational method that mimics the binding of a ligand to a
protein
• Given...

• Predicts...
• The pose of the molecule in
the binding site
• The binding affinity or a
score representing the
strength of binding
 Pose vs. binding site
• Binding site (or “active site”)
– the part of the protein where the ligand
binds
– generally a cavity on the protein surface
– can be identified by looking at the crystal
structure of the protein bound with a known
inhibitor
• Pose (or “binding mode”)
– The geometry of the ligand in the binding
site
– Geometry = location, orientation and
conformation
• Protein-ligand docking is not about
identifying the binding site
 Uses of docking
• The main uses of protein-ligand docking are for
– Virtual screening, to identify potential lead compounds from
a large dataset (see next slide)
– Pose prediction

• Pose prediction
• If we know exactly where
and how a known ligand
binds...
– We can see which parts are
important for binding
– We can suggest changes to
improve affinity
– Avoid changes that will ‘clash’
with the protein
 Components of docking
•
software
Typically, protein-ligand docking software consist of
two main components which work together:

• 1. Search algorithm
– Generates a large number of poses of a molecule in the
binding site

• 2. Scoring function
– Calculates a score or binding affinity for a particular pose

• To give:
• The pose of the molecule in
the binding site
• The binding affinity or a
score representing the
strength of binding
 Search Algorithms
• We can classify the various search algorithms
according to the degrees of freedom that they
consider
• Rigid docking or flexible docking
– With respect to the ligand structure
• Rigid docking
• The ligand is treated as a rigid structure during the
docking
– Only the translational and rotational degrees of freedom are
considered
• To deal with the problem of ligand conformations, a large
number of conformations of each ligand are generated in
advance and each is docked separately
• Examples: FRED (Fast Rigid Exhaustive Docking) from
 The perfect scoring function
Molecular docking
will…
• Accurately calculate the binding affinity
– Will allow actives to be identified in a virtual screen
– Be able to rank actives in terms of affinity

• Score the poses of an active higher than poses of an

inactive
– Will rank actives higher than inactives in a virtual screen

• Score the correct pose of the active higher than an

incorrect pose of the active
– Will allow the correct pose of the active to be identified

• “actives” = molecules with biological activity

 Identification of novel hits using CADD
Binding site
Known Unknown

1.Rational design 2.Virtual screening 3. De novo design

(visual inspection and MoA) (input: compound libraries)

BOMB, BREED, LUDI…

DOCKING softwares
Pymol, chimera, VMD.. softwares
 Identificació de nous hits basats en
l’estructura amb eines computacionals
3.Disseny deennovo
DISEÑO basado la ESTRUCTURA de novo AUTOMATIZADO

Programa LUDI, Böhm 1992

Las estructuras que se obtienen son

completamente novedosas y en algunos
casos un reto para los químicos orgánicos
sintéticos.

El programa Ludi tambien se puede utilizar en diseño basado en el ligando a partir de

una serie de ligandos cuando no se conoce el receptor
 Identificació de nous hits basats en
l’estructura amb eines computacionals
3.Disseny de novo
DISEÑO basado en la ESTRUCTURA de novo AUTOMATIZADO

Ejemplos de
fragmentos usados
por LUDI

Proceso de
unión (LUDI)

Ejemplos de
puentes
moleculares
(LUDI)
The druggable proteome

Protein with approved

GPCR
drugs
Nuclear Receptors

Ion Channels

3%
Enzymes

Transporters

Kinases

Others

[Oprea et al. Nat. Rev. Drug. Discover 2018]

 Identification of novel hits using CADD
Binding site

Known Unknown
Static
(GRID, Fpocket…) Binding site prediction

Dinamic
(MdMix…) 3D farmacòfor

1.Disseny racional 2.Virtual screening 3. De novo design

BOMB, BREED, LUDI…

DOCKING softwares
Pymol, chimera, VMD.. softwares
 Pharmacophore definition

A pharmacophore is the ensemble of steric and electronic

features that is necessary to ensure the optimal
supramolecular interactions with a specific biological
target structure and to trigger (or to block) its biological
response. A pharmacophore does not represent a real
molecule or a real association of functional groups, but a
purely abstract concept that accounts for the common
molecular interaction capacities of a group of compounds
towards their target structure.

There are 2D pharmacophore and 3D pharmacophore

 Example of 2D pharmacophore

Steric and electronic features

 Pharmacophore: chemical features
• The chemical features can be hydrogen bonds acceptors, hydrogen bond
donors, charge interactions, hydrophobic areas, aromatic rings, positive or
negative ionizable group.) The shape or volume is also considered.

Hyd Acc

Acc Aro

Acc & Don Start to be complex !!!

• Pharmacophores represent chemical functions, valid not only

for the curretly bound, but also unknown molecules.
 Identification of novel hits using CADD
Binding site

Known Unknown
Static*
(GRID, Fpocket…) Binding site prediction

Dynamic
(MdMix…) 3D farmacòfor

1.Disseny racional 2.Virtual screening 3. De novo design

BOMB, BREED, LUDI…

DOCKING softwares
Pymol, chimera, VMD.. softwares

* Static can de divided in: template-based, geometry-based and energy-based methods

Finding druggable hot spots with MDMix
MDMIX identifies high affinity interactions spots in the surfaces of macromolecular systems
by means of molecular dynamics (MD) simulations using solvent mixtures as solvation conditions

Binding site detection

Portable hot spots

Characterization of waters
 LEAD OPTIMIZATION:
• SAR – Binding affinity prediction
(FEP, TI, MM/PBSA,…)

• ADME prediction
 Computer-aided drug design (CADD)
– Hit discovery
Known ligand(s) No known ligand
Known protein

Structure-based drug
design (SBDD) De novo design
structure

Protein-ligand docking
(Virtual screening)
Rational Design

Ligand-based drug design (LBDD)

protein structure

1 or more ligands
• Similarity searching CADD of no use
Several ligands Need experimental
Unknown

• Pharmacophore modelling data of some sort

Many ligands (20+)
• Quantitative Structure-Activity
Relationships (QSAR)
 Ligand-
based
Selection of novel
compounds based on Construction of a QSAR
chemical similarity to model that predicts
known active biologic activity from
compounds using some chemical structure
similarity measure
QSAR weights the features of the
chemical structure of the compounds
according to their influence on the
biological activity of interest.

Chemoinformatics (Statistical)
VS based
in pharmacophore

Ligand-based
Pharmacophore modelling
 Chemoinformatics
Substructural Similarity Searches
Searches

Link to Property
Other DBs Predictors
mical Interpre
he te
C

r
DataBase
QSAR
Models
Annotations
3D converter
Docking
 Chemoinformatics

Typical cheminformatic questions

• What bioactivities are available for a compound?
– Characterize a screening hit
• Which compounds are similar to a given compound?
– Test or find bioactivities for similar compounds.
• What is the common structural motif in a set of compounds?
– Which motif is responsible for a property?
• How to decompose a compound into fragments?
– Calculate a property of a compound using quantities assigned to
fragments of the compound. (clogP, 3D structure,…)
• How to select a subset of compounds that are likely to hit a wide
range of protein targets?
– Set up a screening collection
Ligand based approaches
Identificació de nous hits basats en el
lligand
Databasis
D’on s’obtenen les dades d’activitat?

• Selected chemical and biological

databases
– ChEMBL
– PubChem
– ZINC
– DrugBank
 Chemoinformatics:
Public Databases & Resources

ChEMBL– Focus on:

• Published biological activities
Ligand-
based
Ligand-based
Pharmacophore modelling
Ligand-
based Construction of a QSAR
model that predicts
biologic activity from
chemical structure

Among the computational chemistry methods:

Quantitative Structure Activity Relationships (QSARs).

QSARs:

equations that help to predict biological “activity” from

chemical structure of ligands.
Ligand- Construction of a QSAR
based model that predicts
biologic activity from
chemical structure

Set of Compounds

Activity Data (Y) Molecular Descriptors (Xi)

QSAR
Y = f(Xi)
Prediction
Interpretation
Tema 4. Introducció Disseny de
fàrmacs assistit per ordinador
Dr. Carles Galdeano
Novembre 2023

cgaldeano@ub.edu