Screening

1
Screening a Virtual Compound Space
ChemAxon Ltd. Máramaros köz 3/a 1037 Budapest
Hungary www.chemaxon.com
Szabolcs Csepregi Ferenc Csizmadia Szilárd
Dóránt Nóra Máté György Pirok Zsuzsanna
Szabó Jeno Varga Miklós Vargyas
2
Drug research
Finding or making a needle in the hay stack?
virtual screening JChem Screen
de novo design JChem AnalogMaker
3
Drug research
Finding or making a needle in the hay stack?
virtual screening JChem Screen
de novo design JChem AnalogMaker
4
Virtual Screening
Find something similar to a fistful of needles
5
Molecular similarity
How to tackle it?

• Quantitative assessment of similarity/dissimilarit
  y of structures
• need a numerically tractable form
• molecular descriptors, fingerprints, structural
  keys

Sequences/vectors of bits, or numeric values that
can be compared by distance functions, similarity
metrics.
6
Virtual screening using fingerprints
Multiple query structures
01000101000111010100001100001010000100110000101000
00000100100000 00011011100111011111101000001000100
00110110110000000100110100000 01000101001101000100
00000010000000010010000000100100001000101000 01011
10100110101010111111000010000011111100010000100001
000101000 0001000100010100010100100000000000001010
000010000100000100000000 0100010100010100000000000
000101000010010000000000100000000000000 0101010101
11110011111010000000000001101010001110010000110010
1000 010001010001100001000001100000000001000100000
0110000000001100000 000000010000000001000010000000
0000001010100000000100000100100000
01011101001101010101111110000100000111111000100
00100001000101000
queries
hypothesis fingerprint
metric
00000001000011010000001010100000000001100000100001
00001000001000 01000101100100100101100110100111001
11101000000110000000110001000 01000101000111010100
00110000101000010011000010100000000100100000 00011
01110011101111110100000100010000110110110000000100
110100000 0100010100110100010000000010000000010010
000000100100001000101000 0100011100011101000100001
011101100110110010010001101001100001000 0101110100
11010101011111100001000001111110001000010000100010
1000 010001010011110101000010001000000001001000001
0100100001000101000 000100010001010001010010000000
0000001010000010000100000100000000 010001010001001
1000000000000000000010100000010000000000000000000
01000101000101000000000000001010000100100000000001
00000000000000 01010101011111001111101000000000000
11010100011100100001100101000 01000101000110000100
00011000000000010001000000110000000001100000 00000
00100000000010000100000000000001010100000000100000
100100000 0100010100010100000000100000000000010000
000000000100001000011000 0001000100001100010010100
000010100101011100010000100001000101000 0100011100
01010001000010000100111001001000001000110000000010
1000 010101010001010001010010000000000001001000001
0010100100100010000
targets
target fingerprints
7
Optimized virtual screening
Parameterized metrics
asymmetry factor
scaling factor
8
How good is optimized virtual screening?
ß2-adrenoceptor antagonist
9
Is virtual screening a discovery tool?
Scaffold hopping
10
Drug research
Finding or making a needle in the hay stack?
virtual screening JChem Screen
de novo design JChem AnalogMaker
11
JChem AnalogMaker
Workflow
Lead Candidates
12
Fragmentation
Examples
Fragmentation rules
Original molecule
Generated fragments
Amide
Ester
13
Fragmentation
RECAP rules
1 amide
2 ester
3 amine
4 urea
5 ether
6 olefin
7 quaternary nirogen
8 aromatic N carbon
9 lactam N carbon
10 aromatic carbon aromatic carbon
11 sulphonamide
Xiao Qing Lewell, Duncan B. Judd, Stephen P.
Watson, Michael M. Hann RECAP retrosynthetic
combinatorial analysis procedure a powerful new
technique for identifying privileged molecular
fragments with useful applications in
combinatorial chemistry. J. Chem. Inf. Comput.
Sci. 1998, 38, 511522
14
JChem AnalogMaker
General algorithm
create building block library
generate pharmacophore hypothesis of active
compounds
create several starting compounds by random
combination of some building blocks
select parent structure
generate ? variants of parent
15
Variant generation
Example TOPAS modifier
G. Schneider et al, J. Comput.-Aided Mol. Design,
14(2000) 487-494 G. Schneider et al, Angew.
Chem. Int. Ed., 39(2000) 4130-4133
16
Drug research
Finding or making a needle in the hay stack?
virtual screening JChem Screen
de novo design JChem AnalogMaker
17
Drug research
Finding or making a needle in the hay stack?
virtual screening JChem Screen
de novo design JChem AnalogMaker
? ?
18
Drug research
Screening a virtual compound space
de novo design JChem AnalogMaker
virtual screening JChem Screen
random virtual synthesis JChem Synthesizer
19
Screening a virtual compound space
Smart reactions

• Generic (simple)
• the equation describes the transformation only
• few hundred generic reactions can form the basic
  armory of a preparative chemist
• Specific (complex)
• chemo-, recognizes reactive and inactive
  functional groups
• regio-, "knows" directing rules
• stereo-, inversion/retention
• Customizable
• to improve reaction model quality

20
Smart reactions
Chemoselectivity
REACTIVITY !match(ratom(3), "6N,O,S1N,O,S
", 1)
21
Smart reactions
Chemoselectivity
REACTIVITY !match(ratom(3), "6N,O,S1N,O,S
", 1) !match(ratom(3), "N,O,S1C,P,SN,O
,S", 1)
22
Smart reactions
Regioselectivity
SELECTIVITY -charge(ratom(1)) TOLERANCE 0.0045
23
Smart reactions
Regioselectivity
SELECTIVITY -charge(ratom(1)) TOLERANCE 0.0045
24
Smart reaction library
Example
Baeyer-Villiger ketone oxidation
SELECTIVITY charge(ratom(2), "sigma")
25
Smart reaction library
Baeyer-Villiger ketone oxidation
Generic reaction
26
Smart reaction library
Example
Baeyer-Villiger ketone oxidation
27
JChem Synthesizer
Workflow
Virtual compound space
Available chemicals
Smart reaction library
28
JChem Synthesizer example
Dopamine D2 actives
29
JChem Synthesizer example
Virtual hits
similarity 2D pharmacophore fingerprint,
weighted Euclidean metric optimized for 20
random d2 actives
30
JChem Synthesizer example
Best virtual hits
9.88
9.82
9.53
9.73
31
JChem Synthesizer example
Synthesis path
step 1
Knoevenagel-Doebner condensation
32
JChem Synthesizer example
step 2
Baylis-Hillman vinyl alkylation
33
JChem Synthesizer example
step 3
Lawesson thiacarbonylation
34
JChem Synthesizer example
step 4
Dess-Martin alcohol oxidization
35
JChem Synthesizer example
Software and performance data

• virtual reactions 500-1000 reactions/s
• random synthesis 10-20 structures/s
• pharmacophore fingerprint generation 100
  structure/s (includes pharmacophore point
  perception)
• metric optimization 57 sec (13 parameterized
  metrics, 20 structures in training set, 50
  spikes)
• virtual screening 7500 structure/s
• pure Java
• client P4 1.6GHz, RH Linux, java 1.4.2
• database server P4 2.4GHz, Windows XP, MySQL

36
Acknowledgements
François Petitet
Alex Allardyce
ChemAxon
37
Contact
Miklós Vargyas mvargyas_at_chemaxon.hu office
36 1 453 2661 mobile [phone]