1

GDB and the Chemical Space

The Chemical Space Project J.-L. Reymond, Acc. Chem. Res. 2015, 48, 722-730

1. GDB-17 and FDB-17

2. Visualization

3. Virtual Screening

4. Extension to larger molecules

Jean-Louis Reymond

19 April 2017, BigChem BCN Meeting

gdb.unibe.ch

2

Molecules

166,443,860,262

Hydrocarbons

5,422,153

Graphs

114,304,569,097

Skeletons

1,330,958,530

GDB-17

Rod Sphere

Disc

T. Fink et al. Angew. Chem. Int. Ed. 2005, 44, 1504-1508,

J. Chem. Inf. Model. 2007, 47, 342-353 (GDB-11)

L. C. Blum, J.-L. Reymond, J. Am. Chem. Soc. 2009, 131, 8732-3 (GDB-13);

L. Ruddigkeit et al., J. Chem. Inf. Model. 2012, 52, 2864-2875 (GDB-17)

Ueber die analytischen Figuren, welche in der Mathematik

Bäume genannt werden und ihre Anwendung auf die Theorie

chemischer Verbindungen. E. Cayley, Chem. Ber. 1875, 8, 1056-

1059. Practical Graph Isomorphism. B. D. McKay, Congressus

Numerantium 1981, 30, 45-87.

ChEMBL-17

Fragment Database FDB-17

3Fragment Database FDB-17. R. Visini, M. Awale, J.-L. Reymond., J. Chem. Inf. Model. 2017, doi:10.1021/acs.jcim.7b00020

GDB-17: Molecules

166,443,860,262

Hydrocarbons

5,422,153

2) Unsaturations

Graphs

114,304,569,097

1) Ring strain

Skeletons

1,330,958,530

3) Heteroatoms

4.6 G Subset: Fragments

4,606,078,406

4) Limit complexity

5) Even sampling

FDB-17

10,101,204

Limit Complexity

4

Table 1. Filtering criteria to reduce GDB-17 to its 4.6G fragment subset.

Scaffolds FG Density Problematic/Superfluous FGs

3 rings 5 Nitrogen + Oxygen atoms No aldehydes No aromatic ring > 6 atoms

2 small (3- or 4-

membered) rings

1 positive charge at neutral pH No epoxides, aziridines 1 CN (cyanide)

2 quaternary centers 1 negative charge at neutral pH No O-(C=O)-O (carbonate) No non-aromatic C=C

4 stereocenters 3 H-bond acceptor atoms No O-C=N (imidate) No CC (triple bonds)

3 rotatable bonds 2 H-bond donor atoms No NO2 (nitro) No halogens

Even Sampling

5

a) b)

c) d)

1000

10000

100000

1000000

10000000

100000000

1000000000

0 50 100 150

Num

be

r o

f co

mp

ound

s

Bin by size

1000000

10000000

100000000

1000000000

10000000000

0 1 2 3 4

Num

be

r o

f co

mp

ound

s

Stereocenters

1000000

10000000

100000000

1000000000

10000000000

≤1 2 3 4 ≥5

Num

be

r o

f co

mp

ound

s

Heteroatoms

1000000

10000000

100000000

1000000000

10000000000

≤11 12 13 14 15 16 17

Num

be

r o

f co

mp

ound

s

Heavy atom count

4.6G fragment subset

FDB-17

value triplets (HAC, heteroatoms, stereocenters)

Property Profiles

6

0

10

20

30

40

50

60

0 1 2 3 4 5

% o

f d

ata

se

t

Hydrogen bond donor

0

10

20

30

40

50

60

70

6 10 14

% o

f d

ata

se

t

Heavy atom count

0

5

10

15

20

25

30

35

40

45

80 180 280

% o

f d

ata

se

t

Mass in dalton

0

1

2

3

4

5

-4 -3 -2 -1 0 1 2 3 4 5

% o

f d

ata

se

t

Clog P

0

10

20

30

40

50

60

0 1 2 3 4 5 6

% o

f d

ata

se

t

Ring count

0

20

40

60

80

0 2 4 6 8

% o

f d

ata

se

t

Stereo centers

0

10

20

30

40

50

0 2 4 6 8 10

% o

f d

ata

se

t

O+N count

0

5

10

15

20

25

30

35

0.0 0.2 0.4 0.6 0.8 1.0

% o

f d

ata

se

t

fsp3 ratio

0

5

10

15

20

25

30

35

0 1 2 3 4 5 6 7

% o

f d

ata

se

t

Rotable bond count

a) b) c)

e)

d)

g)f)

h) j)i) k)

0

10

20

30

40

50

60

70

he

tero

aro

ma

tic

aro

ma

tic

he

tero

cyclic

ca

rbo

cyclic

acyclic

% o

f D

ata

se

t

GDB-17

4.6G fragment subset

FDB-17

commercial fragments

0

10

20

30

40

50

60

70

0 1 2 3 4 5 6 7 8

% o

f d

ata

se

t

Hydrogen bond acceptor

Nearest Neighbor Searches

7

fencamfamine rimantadine levetiracetamgabapentin

a)

b)

c)

d)

0

500

1000

1500

2000

2500

3000

3500

4000

0.5 1 1.5 2

num

ber

of com

pounds

ROCS score

GDB-17 MQN

4.6G frag MQN

FDB-17 MQN

FDB-17 Xfp

FDB-17 Random

Commercial

0

100

200

300

400

500

600

700

800

900

0 0.2 0.4 0.6 0.8 1

num

ber

of com

pounds

TSfp

0

500

1000

1500

2000

2500

3000

3500

4000

4500

0.5 1 1.5 2

num

ber

of com

pounds

ROCS score

0

10

20

30

40

50

60

70

80

90

0 0.2 0.4 0.6 0.8 1

num

ber

of com

pounds

TSfp

0

500

1000

1500

2000

2500

3000

3500

4000

0.5 1 1.5 2

num

ber

of com

pounds

ROCS score

0

100

200

300

400

500

600

700

0 0.2 0.4 0.6 0.8 1

num

ber

of com

pounds

TSfp

0

500

1000

1500

2000

2500

3000

3500

4000

4500

0.5 1 1.5 2

num

ber

of com

pounds

ROCS score

0

10

20

30

40

50

60

70

0 0.2 0.4 0.6 0.8 1

num

ber

of com

pounds

TSfp

Understanding Molecular Diversity

88

1. GDB-17 and FDB-17

2. Visualization

3. Virtual Screening

4. Extension to larger molecules

9

Property 1

Property 2

Property 3

Property 4

Property 5

Property 6

PCA

PC1

PC2

n-dimensional Space

search

2D (or 3D) projection

Visualise

10

Molecular Quantum Numbers

Polar groupsH-Bond donor atoms 3 1H-Bond donor sites 3 1H-Bond acceptor atoms 3 4H-Bond acceptor sites 3 7Positive charges 1 0Negative charges 0 1

BondsAcyclic single bonds 3 8Acyclic double bonds 0 3Acyclic triple bonds 0 0Cyclic single bonds 18 11Cyclic double bonds 4 3Cyclic triple bonds 0 0Rotatable bonds 0 4

TopologyAcyclic monovalent nodes 3 6Acyclic divalent nodes 0 2Acyclic trivalent nodes 0 2Acyclic tetravalent nodes 0 0Cyclic divalent nodes 8 6Cyclic trivalent nodes 9 6Cyclic tetravalent nodes 1 13-Membered rings 0 04-Membered rings 0 15-Membered rings 1 16-Membered rings 4 17-Membered rings 0 08-Membered rings 0 09-Membered rings 0 0 10 membered rings 0 0Atoms shared by fused rings 7 2Bonds shared by fused rings 6 1

AtomsCarbon 17 16Fluorine 0 0Chlorine 0 0Bromine 0 0Iodine 0 0Sulphur 0 1Phosphor 0 0Acyclic nitrogen 0 1Cyclic nitrogen 1 1Acyclic oxygen 2 4Cyclic oxygen 1 0Heavy atom count 21 23

HO

O

HO

NMe

H

H

Morphine

N

S

CO2H

HN

O

O

Penicillin G

R. van Deursen et al. J. Chem. Inf. Model. 2010, 50, 1924-1934

GDB-17

11Visualization and Virtual Screening of the Chemical Universe Database GDB-17. L. Ruddigkeit, L. C. Blum, J.-L. Reymond, J.

Chem. Inf. Model. 2013, 53, 56-65.

PubChem Ro5 + Oligomers

12

Oligosaccharides

DNA

Diamondoids

GraphenesSmall Molecules

Acyclic Alkanes

Carbon Fraction

(Polarity)

Peptides

13

Oligosaccharides

Diamondoids

Graphenes

Small Molecules

Acyclic Alkanes

Fraction Cyclic Atoms

(Rigidity)

DNA

Peptides

Ro3

Ro5

14

> N-dimensional fingerprints:

– APfp: 20 counts for atom pairs (shape)

– MQN: 42 molecular quantum numbers

– SMIfp: 34 counts for SMILES characters

– Xfp: 55 counts for atom pairs with properties (pharmacophore)

– Sfp: 1024-bit binary substructure fp

– ECfp4: 1024-bit binary extended connectivity fp

> Tools at gdb.unibe.ch

– Web-browsers for DrugBank, ChEMBL, ZINC, PubChem, GDB-11, GDB-13,

GDB-17

– Mapplets (downloadable Java applications, ca. 100 Mb)

– Similarity Mapplets (tailored mapplets by e-mail)

– WebDrugCS (3D-viewer, platform independent, DrugBank)

– WebMolCS (3D-viewer, platform independent, up to 5000 molecules from

user)

Visualisation and subsets of the chemical universe database GDB-13 for virtual screening. L. C. Blum, R. van Deursen, J.-

L. Reymond, J. Comput. Aided Mol. Des. 2011, 25, 637-647

A multi-fingerprint browser for the ZINC database. M. Awale, J.-L. Reymond, Nucleic Acids Res. 2014, 42, 234-239

The MQN-Mapplet: Visualization of Chemical Space with Interactive Maps of DrugBank, ChEMBL, PubChem, GDB-11 and

GDB-13. M. Awale, R. van Deursen, J. L. Reymond, J. Chem. Inf. Model. 2013, 53, 509-518

Similarity Mapplet: Interactive Visualization of the Directory of Useful Decoys and ChEMBL in High Dimensional Chemical

Spaces. Awale M, Reymond JL, J. Chem. Inf. Model., 2015, 55, 1509-1516. 14

15

MQN-Mapplet

The MQN-Mapplet: Visualization of Chemical Space with Interactive Maps of DrugBank, ChEMBL, PubChem, GDB-11 and

GDB-13. M. Awale, R. van Deursen, J. L. Reymond, J. Chem. Inf. Model. 2013, 53, 509-518

16

a) b)

c) d)

Similarity Maps

17Similarity Mapplet: Interactive Visualization of the Directory of Useful Decoys and ChEMBL in High Dimensional

Chemical Spaces. Awale M, Reymond JL, J. Chem. Inf. Model., 2015, 55, 1509-1516

Similarity Maps

18Similarity Mapplet: Interactive Visualization of the Directory of Useful Decoys and ChEMBL in High Dimensional

Chemical Spaces. Awale M, Reymond JL, J. Chem. Inf. Model., 2015, 55, 1509-1516

18

Adenosine A2a receptor - aa2ar (Sfp: 78 %, 7 %)

dis

tan

ce

co

rre

lati

on

co

eff

.

0.4

0.5

0.6

0.7

0.8

0.9

1

1 2 3 4 5 6 7 8 9 10 11 12

cu

mu

lati

ve

va

ria

nc

e

PC

0

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10 11 12

a) b)

PC

0.4

0.5

0.6

0.7

0.8

0.9

1

1 2 3 4 5 6 7 8 9101112

MQN

SMIfp

APfp

Xfp

Sfp

c)

0

20

40

60

80

100M

…

A…

S…

% of DB covered by single occupancybins in original fingerprint-space

% of DB covered by single occupancypoints in 3D-space

% of DB covered by single occupancypixels in 2D-space

% o

f d

ata

ba

se

(D

B)

0

20

40

60

80

100

MQN SMIfp APfp Xfp Sfp

Ori

3D

2D

DrugBank in 3D Chemical Spaces

Web-based 3D-visualization of the DrugBank chemical space Awale M and Reymond JL, J. Cheminform., 2016,

doi:10.1186/s13321-016-0138-2.

a b

c dc

WebMolCS

21WebMolCS: a Web-Based Interface for Visualizing Molecules in 3D Chemical Spaces. Awale M, Reymond JL, J.

Chem. Inf. Model., 2017, doi:10.1021/acs.jcim.6b00690

Does this work?

22

1. GDB-17 and FDB-17

2. Visualization

3. Virtual Screening

4. Extension to larger molecules

23

GDB-17

166.4 G

xLOS

3D pharmacophore + shape

Sorted VS Hit list

Reference

Ligand(s)

MQN-neighbors

Xfp-neighbors

1.0 G

20 M

Docking

Synthesis/Testing

3D-Shape and pharmacophore (xLOS)

24

𝐱𝐋𝐎𝐒 𝐬𝐜𝐨𝐫𝐞 = 𝒆−𝒅𝒊𝒋

𝟐𝒏𝒅𝒊=𝟏

𝒎𝒅𝒋=𝟏

𝒎𝒅 + 𝒏𝒅

+ 𝒆−𝒅𝒊𝒋

𝟐𝒏𝒂𝒊=𝟏

𝒎𝒂𝒋=𝟏

𝒎𝒂 + 𝒏𝒂

+2 × 𝒆−𝒅𝒊𝒋

𝟐𝒏𝒉𝒊=𝟏

𝒎𝒉𝒋=𝟏

𝒎𝒉 + 𝒏𝒉

0.00

0.20

0.40

0.60

0.80

1.00

0 0.5 1 1.5 2 2.5 3

ov

erl

ap

sco

re (

e(-

dij2))

Atom-Atom Distance dij, Å

dij

Atom i

Atom j

Seed (n atoms)

Query (m atoms)

Optimization of TRPV6 Calcium Channel Inhibitors Using a 3D Ligand-Based Virtual Screening Method. C Simonin,

M Awale et al., Angew. Chem., Int. Ed. Engl. 2015, 54, 14748-14752.

0

200

400

600

800

1000

1200

1400

1600

top scoring 1000 cpdsfrom 5 referencemolecules

MW

no

. o

f cp

ds

1,3,5

4

2

p2 (y)

p3 (z)

p1 (x)

25

xLOS

TS

fp

no

. o

f c

pd

s

0

50000

100000

150000

200000

250000

300000

350000

0.00

0.20

0.40

0.60

0.80

1.00

0.00 0.40 0.80 1.20 1.60

xLOS vs Sfp for tested cpdsxLOS vs Sfp for hit cpdsxLOS score histogram

9

10

Table S1: In vitro pharmacology profile of cis-22a on several ion channels.a

Channel % Inh. @ 10 µM

TRP channels

TRPV1 (agonist effect) (h) -15.5 ± 0.7c

TRPV1 (antagonist effect) (h) 29.6 ± 8.6

TRPV3 (antagonist effect) (h) -11.5 ± 2.0

TRPV5 (antagonist effect) (r) 79.9 ± 1.4 (IC50 = 2.4 M)

TRPM8 (agonist effect) (h) -7.4 ± 1.4c

TRPM8 (antagonist effect) (h) 21.2 ± 7.5

Voltage-gated ion channels

Ca2+

channel L-type, dihydropyridine (r) -1.6 ± 20.5

Ca2+

channel L-type, diltiazem (r) 18.4 ± 13.0

Ca2+

channel L-type, verapamil (r) 31.9 ± 3.0

Ca2+

channels, N-type (r) 4.4 ± 8.1

K+ channel [Kv] (r) -15.9 ± 3.8

hERG (h) 82.3 ± 4.0d

Na+ channels, site 2 (r) 52.1 ± 4.2

Ligand-gated ion channels

5-HT3 (h) -1.3 ± 5.9

GABA, central benzodiazepine (r)b -10.7 ± 8.7

Glutamate, NMDA (r) 8.2 ± 1.5

Store-operated Ca2+

channels -10.1 ± 6.0

GPCR

Adrenergic α1A (h) 94 ± 5

Adrenergic α2A (h) 31.2 ± 4.9

Adrenergic β1 (h)b 14.3 ± 3.1

Adrenergic β2 (h)b 11.5 ± 0.6

Cannabinoid CB1 (h ) b -19.2 ± 4.2

Cannabinoid CB2 (h) b

-1.2 ± 7.5

Dopamine D1 (h) 63.9 ± 0.2

Dopamine D2S (h) b 90.3 ± 0.6

Dopamine D3 (h) 50.4 ± 2.1

Dopamine D4.4 (h) 93.8 ± 0.5

Muscarinic M1 (h) 74.3 ± 2.7

Muscarinic M2 (h) 87.2 ± 2.7

Opiate µ (h) b 98.1 ± 0.6

Serotonine 5-HT1A (h) 88.5 ± 4.5

Serotonine 5-HT2A (h) 56.2 ± 2.8

Serotonine 5-HT1B (r) 28.4 ± 4

Serotonine 5-HT2B (h) 23.0 ± 4.0

T47D

LNCaP

SKOV3

HEK u

ntran

sfec

ted

HEK

cl.1

5

0.0

0.5

1.0

1.5

TRPV6 mRNA

fold

mR

NA

ex

pre

ss

ion

(re

l. to

T4

7D

)

ND 0.000271

70

80

90

100

110

5000400030002000100050030010050

% g

row

th

conc. (nM)

cis-22a

T47D

SKOV3

70

80

90

100

110

5000400030002000100050010050

% g

row

th

conc. (nM)

trans-22a

T47DSKOV3

Optimization of TRPV6 Calcium Channel Inhibitors Using a 3D Ligand-Based Virtual Screening Method. C Simonin,

M Awale et al., Angew. Chem., Int. Ed. Engl. 2015, 54, 14748-14752.

26

HRD rotated

K162

E181

Q185-R255

D274-R255

A213

F275R137

R220

HRD

K162

E181

D274-H254

A213

F275

Q185-W277

L139

L263

V147

K162

G142

A160

R195

L194L196

A273

L210

F275

A213

R220

L139

G216

A213

L263

R137

V147

K162G142

A160

F275

R195

L194L196

A273

L210

Y212

Kinase

Inhibitors

1) xLOS

2) Biochemical screen

3) Structure-based optimization

Aurora A

IC50 = 2 nM

Discovery of a Selective Aurora A Kinase Inhibitor by Virtual Screening. Falco Kilchmann et al., J. Med. Chem., 2016,

doi:10.1021/acs.jmedchem.6b00709

27

KD (nM)

- 9 795 ± 129

+ 9 -

DNA Aurora A TPX2 Merge

Con

tro

l9

(10

µM

)

1

(0.2

5µ

M)

5 μm

9 (µM) 9 rac-92

28

New Molecules

29

1. GDB-17 and FDB-17

2. Visualization

3. Virtual Screening

4. Extension to larger molecules

rod sphere

disk

www.cheminfo.org/pdbexplorer

30Jin X, Awale M, Zasso M, Kostro D, Patiny L, Reymond JL: PDB-Explorer: a web-based interactive map of the protein data

bank in shape space. BMC Bioinformatics 2015, 16, 339.

31

gdb.unibe.ch

Xian Jin

Mahendra

Awale

Ivan Dibonaventura

Tamis

Darbre Céline Simonin

Bee-Ha

Gan

Ricardo Visini

Runze He

Daniel Probst

Josep Arus

Luc Patiny, EPFL

Daniel Bertrand, HiQscreen

Matthias Hediger, UniBE