Karel Berka, Ph.D.

Jindřich Fanfrlík, Ph.D.

Martin Lepšík, Ph.D.

Advanced in silico Drug Design

KFC/ADD Drug design intro

UP Olomouc, 1.-3.2. 2016

Motto

A pharmaceutical company utilizing computational drug design is like an organic chemist utilizing an NMR. It won’t solve all of your problems, but you are much better off with it than without it.

DAVID C. YOUNG

Disclaimer:

Authors of this lecture series did not designed any drug in common use.

Yet.

Course program Monday

• 13:00 Drug design – intro (Berka)

• 14:00 Databases of small molecules (ZINC), drug representations (Berka)

• 15:00 Properties of biologically active compounds (Fanfrlik)

• 16:00 Biomolecular targets (Lepšík)

• 17:00 Training – target database search (Berka)

Tuesday

• 9:00 Noncovalent interactions in drug-target binding (Fanfrlík)

• 10:00 Molecular docking intro (Berka)

• 11:00 Training – visualization and simple docking (Berka)

Tuesday

• 14:00 Advanced docking – flexibility, consensus, ensemble docking (Fanfrlík)

• 15:00 Advanced scoring (WaterMap, MM-PBSA, FEP, QM) (Lepšík)

• 16:00 Training – scoring (Lepšík)

Wednesday

• 9:00 QSAR, ADMET (Berka)

• 10:00 Advanced virtual screening (Fanfrlík)

• 11:00 Future of drug design (Lepšík

KFC-ADD Exam Requirements

• Project

– Drug design for selected problem

(short and clear drug design report, 1-4 pages, short intro with defined task, short methodology, results with visuals, conclusion)

• Oral exam

Discussion over project and strengths and weaknesses of selected approach

Literature

• Young, D.C. Computational Drug Design. Wiley, 2009.

• Young D.C. Computational Chemistry, a Practical Guide for Applying Techniques to Real World Problems. Wiley, 2001.

• Leach AR. Molecular Modelling - Principles and Applications (2 edition). Pearson Education, 2001.

• Alvarez J. & Shoichet B. (Eds.). Virtual Screening in Drug Discovery. Taylor&Francis, 2005.

• Berka K. & Bazgier V. Racionální návrh léčiv pomocí in silico metod. UP Olomouc, 2015 (in Czech)

History of Drug Design

1806

A. Cherkasov

Time New Sources Testing Subjects - ancient & plants, poisons (Paracelsus) humans middle ages minerals ... natural sources - 1806 morphine (first extracted) humans - 1850 chemicals (chinin) humans (prisoners) - 1890 synthetics, pigments animals - 1920 animals, isolated organs - 1970-1980 enzymes, cell lines (HeLa) - 1990 High throughput libraries recombinant proteins - 2000 chemical libraries chips, virtual screening, ADMET testing

History of Drug Design

All drugs by source, registered 01/1981 - 06/2006, FDA, n = 1184

B – biologicals, N – nature compounds, ND – nature compounds derivatized, S – synthetic compounds,

S/NM – synthetics mimicking natural compounds, S* - synthetic, with pharmacophore from natural compounds V - vaccines

Sources of Drugs

D. J. Newman and G. M. Cragg, J. Nat. Prod. 70, 461-477 (2007)

S; 30%

ND; 23% B; 14%

S/NM; 10%

S*/NM; 10%

N; 5% S*; 4% V; 4%

Drug Design Project Illness

Isolation of

cause

(many years)

Identification of

lead compound

(2-5 let)

Preclinical

testing on

animals

(1-3 years)

Formulation

(1-3 years)

Human clinical trials

(2-10 years)

FDA approval

(2-3 years)

Heal and Sell

Vocabulary • Target

– Biomolecule interacting with the drug

• Lead – Base molecular structural motif of developed drug

• Hit – Compound with positive hit in initial screening

• Candidate compounds – Selected compounds used for next testing

• Efficacy – Qualitative property – (drug heals or not)

• Activity – Quantitative property – dosage needed for effect to happen

(pM – great, nM – excellent, μM – sufficient, mM – well…)

• Bioavailability – Availability of compound in site of target in necessary concentration

Eroom’s Law

Scannell JW, Blanckley A, Boldon H, Warrington B: Nature Reviews Drug Discovery 11, 191-200 (2012) doi:10.1038/nrd3681

• Decline in pharmaceutical R&D efficiency – halved per 9 years – 'better than the Beatles' problem

– 'cautious regulator' problem

– 'throw money at it' tendency

– 'basic research–brute force' bias.

Why in silico Drug Design? Identification of new drug:

• hard problem

– Identification of target-drug pair is not simple

– ADMET

– Side-effects

• Expensive problem

– Expenditures per 1 drug development - 1 300 000 000 USD1

+ expenses for production, patents, distribution…

New drugs are expensive >1 000 USD/dose of drug2

1 - Tufts Center for the Study of Drug Development, 2012 2 – SÚKL, 3Q 2011, average price tag for most expensive drug category in CZ (over 10kKč)

Hard Problem • Human genom ~30 000 ORF

Alternative splicing => ~500 000 proteins

~26 000 structures known in PDB (8 000 unique)

• 2 – 10 years from lead molecule identification to clinical testing (patents last 20 years)

• 1 successful out of 10 drug development projects

www.rcsb.org - 30.1.2016

Possible Obstacles • Nonexistent testing model

– Example: HIV is human disease! – Ethically not possible to test directly on people (cf. OS)

• Rare disease – orphan disease – Future sales would not pay for regular development – Orphan drug have lower requirements for registration

and individual incentives

• Too low activity of found drug – Too toxic, bad bioavailability

• Active compounds are already patented – Me2drugs – Product has to be just as good as the one from

competition and patentable under our name

Illness Type • Enzyme overproduction - some cancer types

– Inhibition (e.g. kinase inhibitors)

• High response of receptor – COX in pain – Antagonists (e.g. pain relievers)

• Low response of receptor – neurological GPCRs – Agonists (e.g. serotonin receptor agonists)

• Regulation peptide – CGRP peptide in migraine – Antibodies (e.g. biologicals)

• RNA – RNAi, RNA aptamers… – Emerging field

Small ligand with protein

Most Typical Mechanism • Lock and Key Analogon, 1894

Expensive Problem

David C. Young - Computational Drug Design: A guide for computational and medicinal chemists. Wiley-Blackwell, New York, 2009, ISBN 978-0470126851

Experiment Cost per 1 compound

Virtual screening 100 Kč

Biochemical analysis 7 000 Kč

Cell culture testing 75 000 Kč

Acute toxicity on mice 250 000 Kč

Protein structure evaluation with X-Ray 2 000 000 Kč

Efficiency testing on animals 5 500 000 Kč

Chronic toxicity on rats 14 000 000 Kč

Clinical testing on volunteers 10 000 000 000 Kč

Lower price tag allow testing of more drug candidates

Possibilities of in silico Drug Design

Known ligand Unknown ligand K

no

wn

ta

rge

t

str

uctu

re

Un

kn

ow

n ta

rge

t

str

uctu

re

Structure-based drug design

(SBDD)

Docking

Ligand-based drug design

(LBDD)

1 or more ligands

• Similarity search

Several ligands

• Pharmacophore

Large number of ligands (20+)

• Quantitative Structure-Activity

Relationships (QSAR)

De novo design

CADD not possible

some experimental

data needed

ADMET filtering

SIMILARITY SEARCHES

19

Similarity Search

Search for similar structures to lead –

• 2D Sub-Structures

• 3D Sub-Structures

• 3D Conformational flexibility

20

N

NH2

HO

H

N

N(CH3)2

H

S

HN

O O

H3C

5-Hydroxytryptamine (5-HT)Serotonin (a natural neurotransmitter

synthesized in certain neurons in the CNS)

Sumatriptan (Imitrex)

Used to treat migrain headaches

known to be a 5-HT1 agonist

Similarity to Natural Ligand

21

2D Sub-Structure

• Functional groups

• Connectivity

Example. Halogen on aromatic ring together with carboxylic group

[

F

,

C

l

,

B

r

,

I

]

O

O

N

O

O

Cl

O

O

Cl

N

N

N

O

O

F

F

O

F

22

3D Sub-Structure

• Distances in 3D play more important role

• Bioisostericity – similar groups in 3D

• Usually storage only lowest energy structure

C

(

u

)

O

(

s

1

)

O

(

s

1

)

A

A

[

O

,

S

]

O

3.6 - 4.6 Å

3.3 - 4.3 Å

6.8 - 7.8 Å

360300240180120600

0

1

2

3

4

5

6

Dihedral angle

Ste

ric

Ene

rgy

(kca

l/mol

)

23

Bioisostericity

Young, D.C. Computational Drug Design. Wiley, 2009. 24

How to calculate molecular similarity

)&()()(

)&(),(

yxByBxB

yxByxT

n

i

ii yxyxE1

2),(

Sequences/vectors of bits, or numeric values that can be compared by distance functions, similarity metrics .

E= Euclidean distance T = Tanimoto index (distance in XYZ) (similarity of bit vectors)

Quantitative assessment of similarity/dissimilarity of structures

• need a numerically tractable form • molecular descriptors, fingerprints, structural keys

Paradox of Similarity

Aminogenistein (x cystic fibrosis)

7-Hydroxy-2-(4-nitro-phenyl)-chromen-4-one

Pargyline (x hypertensis)

N-benzyl-N,1-dimethyl-2-propynylamine

It is not that simple

26

PHARMACOPHORE

Pharmacophore

• structural motive

(geometrical

restrictions on

functional groups)

important for

biological - mostly

pharmacological

activity

• Analogous to

chromophore

Bojarski, Curr. Top. Med. Chem. 2006, 6, 2005.

Search for pharmacophore • Prerequisition:

– All active compounds in training set bind to common active site

• Pharmacophore query preparation

– Identification of characteristic functional groups (hydrogen bonds acceptors and donors, lipophilic groups, charge distribution)

• Pharmacophore search

– Search for pharmacophore query in all molecules in DB

– Scaffold hopping possible

– No need for receptor structure (can be useful for query generation)

VIRTUAL SCREENING

Virtual screening • Multiple ligands => Pipelining (Pipeline Pilot, KNIME, etc.)

Pipeline Pilot (Accelrys)