Screen3D: A Ligand-based 3D Similarity

Search without Conformational Sampling

Wei Deng (David), Adrian Kalaszi

International Conference and Exhibition on

Computer Aided Drug Design & QSAR

Oct 29th, 2012

Chicago, IL, USA

ChemAxon Fingerprint

• Head Office in Budapest, Hungary

• Cheminformatics software – since ‘98

• Pharma, biotech, agrochem, academia

• Global sales – US, Europe, Asia, RoW

• Database management tools, drug design and discovery, structure analysis

• >30 implementation partners, global agents, consultancy, training

3 / 25

Wide Range of Functionalities

4 / 25

in 3D

3D Structure generation – released in 2002

3D Alignment – released in 2009

3D Similarity – ligand based virtual screening 2010 Q3

multi-core support by 2012 Q4 (version 5.12)

5 / 25

Example: Thrombin inhibitors

1dwc 1dwd

6 / 25

Quaternion Fit

Two molecules: translate & rotate one as rigid molecule

Pros:

Minimize RMSD on selected atom pairs (best alignment)

Fast (one step) – analytically solving the equation

Cons:

No flexibility

Need to know the atom mapping

EA. Coutsias et. al, J Comput Chem 25: 1849–1857, 2004

7 / 25

Flexibility: dihedral angels are optimized

Alternate between quaternion and conformation optimization

Pros:

Very fast

Robust: usually finds the best alignment

Cons:

Two molecules per alignment

Still need to solve the atom/atom mapping

Quaternion & Flexibility Hybrid

8 / 25

Calculations before Atom-atom Match

Select atoms

Colors (extended atom types / pharmacophore types )

Topological features (e.g.: longest chain start/end/center)

Ring centers (aromatic, aliphatic)

Calculate

Min/max internal distance ranges

Atomic histograms for selected atoms

Once in the lifetime for every molecule

9 / 25

Min and Max Distance

info

Watch the video here: http://www.youtube.com/watch?v=FgMsvDmtGkA

10 / 25

Atomic Histograms

d1-min

d1-max

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

binned distribution of distance 1 2 3 4 5 6 7 8 9 10 11 12

binned distribution of distance

1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 131 2 3 4 5 6 7 8 9 10 11 12 13

11 / 25

Distance Range Tanimoto

d2-min

d2-max

d1-min

d1-max

T=

),min(),max(

),max(),min(

min2min1max2max1

min2min1max2max1

dddd

ddddT

For atom pair comparison

12 / 25

Triangle Inequality

B

A

C C’

B’

A’

X= common range between AB and A’B’

= (distance range of AB) ∩ (distance range

of A’B’)

Y= common range between AC and A’C’

Z= common range between BC and B’C’

Then they must follows:

X+Y > Z

X+Z > Y

Y+Z > X

13 / 25

Find the best atom mapping between two molecules

Pairwise search:

1. Same colors of atoms for the pairs

2. Distance ranges OK for any pairs of mapped atoms

3. Triangle inequality for any triplet of maps

4. Quaternion Flexible Hybrid Alignment

Guaranteed: No good solution from the conformational space is lost!

14 / 25

Example: Thrombin inhibitors

Same colors 6,981,488

After filtering by distance range & triangle inequality 542

1dwc 1dwd

15 / 25

Output after 3D alignment

1. Atom/atom mapping

2. Molecules aligned

3. SCORE: Tanimoto of atomic histograms of the mapped atoms

Current version use default scoring weights (1, 0)

It’s possible to add customized weights according to the

atom colors

16 / 25

Test 1 on DUD

ADA Adenosine deaminase 39/927

CDK2 Cyclin dependent kinase 2 72/2074

DHFR Dihydrofolate reductase 410/8367

ER Estrogen receptor antagonist 39/1448

FXA Factor Xa 146/5745

HIVRT HIV reverse transcriptase 43/1519

NA Neuraminidase 49/1874

P38 P38 mitogen activated protein kinase 454/9141

THR Thrombin 72/2456

TK Thymidine kinase 22/891

TRY Trypsin 49/1664

Giganti et al. J. Chem. Inf. Model. 2010, 50, 992

ROCS

Surflex-sim

FlexS

ICMsim

DOCK

Surflex-Dock

FRED

FlexX

ICM

17 / 25

Test 1 on DUD

1% Enrichment

0

5

10

15

20

25

30

35

40

ADA CDK2 DHFR ER FXA HIVRT NA P38 thrombin TK trypsin

Perc

en

t o

f th

e a

cti

ves f

ou

nd

Surflex-sim

ROCS

FlexS

ICMsim

CXN-H

Giganti et al. J. Chem. Inf. Model. 2010, 50, 992

18 / 25

Test 1 on DUD

0

5

10

15

20

25

30

% o

f th

e a

cti

ves r

etr

iev

ed

Average of 1% Enrichments

Giganti et al. J. Chem. Inf. Model. 2010, 50, 992

19 / 25

Test 1 on DUD

0

10

20

30

40

50

60

70

80

90

100

% o

f th

e a

cti

ves r

etr

iev

ed

Average of 10% enrichments

Giganti et al. J. Chem. Inf. Model. 2010, 50, 992

20 / 25

Test 1 on DUD

Speed

average time per compound (s)

CXN-H 0.07

ROCS 0.5

FRED 1.0

ICMsim 2.4

Surflex-sim 6.7

FlexS 6.9

Surflex-dock 14.6

FLEXX 15.6

ICM 17.7

Intel Xeon 2.4 GHz

Intel Q6600 2.4 GHz

Giganti et al. J. Chem. Inf. Model. 2010, 50, 992

0

5

10

15

20

25

30

35

40

45

50

EF1%

Venkatraman et. al. J. Chem. Inf. Model. 2010, 50, 2079–2093

Test 2 on DUD

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

AUC

Venkatraman et. al. J. Chem. Inf. Model. 2010, 50, 2079–2093

Test 2 on DUD

Calculation Time

• Xeon E5-2670@2.60GHz, 2x8 cores

• DUD: 128374 structures

• Total time: 212 minutes

• 99 ms/molecule (38ms/molecule with distance range)

Test 2 on DUD

24 / 25

Conclusion

Screen3D:

Further developments: improve scoring

Comparable to other methods

Fast screening

Evaluators are welcome!

BACKUP SLIDES

26 / 25

Flexible Alignment

Function

atomic coordinates f(x)

Gradient f(x) /

Kinematics: infinitesimal rotational tensor

http://mathworld.wolfram.com/InfinitesimalRotation.html

Hurst, J. Chem. Inf Comput. Sci. 1994, 34, 190-196

MDS-CG (Multi-Dimensional Search in Conjugate Gradient)

Farkas, Ö., Schlegel, H. B. J. Mol. Struct.-Theochem 2003, 666, 31-39.

27 / 25

Experiment

Watch the video here: http://www.youtube.com/watch?v=22lfa5HeCts

28 / 25

Internal Distances

0

1

2

3

4

5

6

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65

rms

d

(An

gstr

om

)

molecule #

distance

coordinate

29 / 25

Distances between Molecules

Function atomic coordinates or distances f(x)

Gradient f(x) /

Add six degrees of freedom on molecules that we wish to

translate (3) & rotate (3)

Optimization on mixed internal coordinate / Cartesian domain

30 / 25

Jump to external distances

Watch the video here: http://www.youtube.com/watch?v=zNtxblx0-Dw

31 / 25

Jump to external distances

Watch the video here: http://www.youtube.com/watch?v=8KJqz5Emco0

32 / 25

Jump to external distances

Watch the video here: http://www.youtube.com/watch?v=kXfoukkAmQQ