Docstoc

Virtual Screening and Experimental High Throughput Screening Are

Document Sample
Virtual Screening and Experimental High Throughput Screening Are Powered By Docstoc
					Virtual Screening and Experimental High Throughput Screening Are Complementary
György M. Keserű Head of Discovery Chemistry Gedeon Richter plc

Quest for a chemical starting point
 Estimated number of druglike compounds: 1060  Number of compounds in CAS registry: 30x106  Screening 100.000 compounds daily against 5000 human

targets at 1000 screening facilities would cost ~30 billion USD that is the ~ annual pharma R&D budget  Simple brute force is over, numbers game ended.

?

Searching needles in the haystack
 High throughput screening (HTS)  Cells or recombinant protein  Fluorescent or luminescent readout  Automated, miniaturized  10.000 or more datapoints daily (uHTS > 100.000)  Virtual screening (VS)  Ligand-based: similarity search, 2D or 3D pharmacophore search, flexible 3D search  Structure based: high throughput docking  Virtual or real libraries

Keserű et al. Drug Disc. Today 11, 741 (2006)

Relationship of HTS and VS

Keserű et al. Virtual Screening in Encycl.Pharm.Tech, Marcel Decker, 2007

Limitation of HTS approaches
 Logistic errors  Incorrect structures, salts, batch numbers  Solubility issues  Frequent hitters (aggregation)  Measurement errors  Variable target expression  Pipetting/dispensing problems  Temperature gradients, positional effects  Suboptimal readout (biological relevance)  Strategic errors  Single molecule/ mixture  Single point/ parallel measurements  Inter/intra assay variability
QSAR Comb. Sci. 2006, 25, 1153

Limitation of VS approaches
 Protein structure  Resolution (homology models)  Completeness  Binding site prediction  Flexibility (induced fit)  Binding mode prediction  Ligand flexibility  Imperfect posing  Solvent effects  Affinity prediction  Protomers, tautomers  Inadequate scoring  False positives and negatives

Integration strategies
 Common conceptual framework: limited

accuracy compensated by the throughput  Significant limitations in both HTS and VS  Do combined efforts work better?  Integration strategies
   

Focused screening Sequential screening Parallel screening Independent screening

Focused and sequential screening

VS

HTS

Screening library

Focused library

Hits

Hypothesis generation

Parallel and independent screening

VS Screening library

HTS

VS hits

HTS hits Analysis

Hits

Parallel and independent screening

VS

Screening library

HTS

VS hits

HTS hits Analysis

Hits

HTS or VS or both?
 Comparative analysis of HTS and VS results

in a parallel setup against 3 kinase targets
    

Using optimized HTS and VS protocols Screening the same library by HTS and VS Evaluating hit rates achieved by HTS and VS Analyzing false positives/negatives in HTS and VS Comparing hit lists obtained by HTS and VS

Optimization of VS protocols
 Objective: selection of known actives from a set

of decoys by ranking  Measure: calculation of enrichment factors (EF) at 1% of the ranked library e.g. EF=38 means that the protocol picked up 38% of known actives in the top 1% of the ranked library  Goal: Selection of the docking/scoring protocol with highest proportion of known actives in the top scored 1% of the library (highest EF)

Case study 1: GSK-3b
Alzheimer target: inhibition of abnormal Tau phosphorylation
 HTS test: Promega Kinase-Glo luminescent assay (in-house) 

VS test: FlexX docking with pharmacophore constraints  Library: Richter corporate sublibrary (16.300 cmpds)

Keserű et al. J. Med. Chem. 48(25), 7946-7959 (2005)

Hits for GSK-3b
2000 0 100 90 80 70

EF: 28

5 0 -5 -10 -15 -20 -25 -30 -35 FlexX score

+

TP: 21 (23%) FP: 141 (87%)

Inhibition %

60 50 40 30 20 10 0 -40 0 2000

+

FN: 69 (77%)

Hit list for GSK-3b
HTS
R1

VS

Cluster 1

O S N O R2

8
6 6
O

2
2 2 1 0 0 Hit rates:
for wet-tested cmpds

CN

Cluster 2 Cluster 3 Cluster 4

R1 S N

R2

R1 X N Y S R2
N S

HTS: 0.0055
VS: 0.129

3
R2

R

R1

Cluster 5
R3

O S X

N

4 4

O

Cluster 6

O R1 O N
+

O X O

R2

Case study 2: JNK-3
Potential target for Alzheimer’s and Huntington’s diseases
 HTS test: homogeneous AlphaScreen assay (NCBI) 

VS test: FlexX docking wit pharmacophore constraints  Library: NIH Molecular Libraries Screening Centers Network sublibrary (10.300 cmpds)

Keserű et al. JCIM 46(4), 1795-1805 (2006)

Hits for JNK-3
Virtual screening against JNK-3
80 40 0 -200 -150 -100 -50

EF: 58

-

+

TP: 13 (38%) FP: 77 (86%) FN: 21 (62%)

Efficacy %

0 50 100 150 200 0 -10 -20 Chem score -30 -40 -50 0 40 80

-

+

Limited availability of TN efficacy

Hit list for JNK-3
HTS
Cluster 1

VS 1 7 2 0 0

2 9 3 3

Cluster 2

Hit rates:
for wet-tested cmpds

HTS: 0.034

Cluster 3

VS: 0.13
Cluster 4

4
Cluster 5

Case study 3: CDK2
Potential target for cancer treatment
 HTS test: kinase inhibition assay (Deltagen Research Labs) 

VS test: FlexX docking with pharmacophore constraints  Library: Deltagen corporate sublibrary (17081 cmpds)

Aktív hely képe

Keserű et al. JCIM 46(4), 1795-1805 (2006), dataset: J. Med. Chem. 2003, 46, 4360

Hits for CDK2
10000

0 120 100 80 60 40

EF: 9

-

+

TP: 30 (8%) FP: 140 (82%) FN: 326 (92%)

Inhibition %

20 0 -20 -40 -60 -80 -100 20 10 0 -10 -20 -30 -40 -50 0 10000

FlexX score

+

Hit list for CDK2
HTS
Cluster 1

VS 4 0

37 28

Cluster 2

Hit rates:
for wet-tested cmpds

Cluster 3

23
19

1
8

HTS: 0.02

VS: 0.17
Cluster 4

Summary
 Optimized VS protocols performed well in real

screening situations (TP-EF = 8-38%)  VS provided significantly higher hit rates, 4-23 fold better than HTS  VS produced large amount of false positives (82-87% of top scored compounds)  25-40% of hit clusters were lost by VS  VS produced large amount of false negatives (missing 62-92% of true actives)

Conclusions
 HTS shows larger sensitivity than VS having 15-25% and    

62-92% FN rates, respectively Specificity of HTS is similar or higher than VS, having 4080% and 82-87% FP rates, respectively Daily VS throughput is similar or higher than HTS Cost effectiveness of VS is larger than HTS VS could not replace HTS but it is worth integrating them:




VS can decrease FP and FN rates of HTS when used on the same library VS can add new hits when used on different –even virtuallibraries

 VS and HTS are complementary rather than competitive

Acknowledgement
 Tímea Polgár, Róbert Kiss (VS)

 GSK3b (Richter), JNK3 (NCBI), CDK2 (Deltagen)

datasets  Greg Makara (Merck Research Labs)


				
DOCUMENT INFO