Presentation Title

Liquid Handling Processes Impact Computational Modeling in Drug Discovery Joe Olechno1, Sean Ekins2, Antony Williams3, Rich Ellson1 Pittcon 2013 Session 2670 3:55 PM, March 21, 2013

1. Labcyte Inc. 2. Collaboration in Chemistry 3. Royal Society of Chemistry

Agenda

• What is Acoustic Liquid Handling? • Serial Dilutions vs. Direct Dilutions • Lead Optimization and Pharmacophores

• The Impact of Serial Dilutions on Drug Discovery • Conclusions

2

Acoustic Droplet Ejection (ADE)

Comley J, Nanolitre Dispensing, Drug Discovery World, Summer 2004, 43-54

3

Acoustic Droplet Ejection (ADE) Acoustic energy expels droplets without physical contact 15.0

• Extremely precise • Extremely accurate • Rapid • Auto-calibrating • Completely touchless – No cross-contamination – No leachates – No binding

12.5 10.0

%CV 7.5 5.0 2.5 0 0.1

1

10

100

Volume (nL)

1000

10000

Comley J, Nanolitre Dispensing, Drug Discovery World, Summer 2004, 43-54

4

Agenda



5

Conventional Dose-Response Set-up by Serial Dilution

Source Plate

Assay Plate

Intermediate Buffer Dilution Plate

Serial Dilution vs. Direct Dilution Serial with Tips

Direct with Acoustics

• Equal volumes of changing concentrations

• Changing volumes of equal concentrations

• Compounds are sequentially diluted. Each new dilution is the source for the next step.

• Maximum of one dilution step

• Many “touches” with tips (or significant potential for carry-over or leachates)

• Touchless—no carry-over, leachates or binding No solute lost

• Errors are compounded

Serial Dilution • Reduced error

• Low-volume assays with high solvent concentration (or compound loss)

Direct Dilution assays with low • Low-volume

solvent concentration

7

Direct Dilution Process

Third Step Transfer 75, 25, 7.5 and 2.5 nL of each hit to four consecutive wells

12-point curves

(30, 10, 3 and one droplets, respectively) Source Plate

Assay Plate Fourth Step Transfer 75, 25, 7.5 and 2.5 nL of each diluted sample to four consecutive wells of the assay plate (30, 10, 3 and one droplets, respectively)

First Step Transfer 252.5 and 2.5 nL to two wells in an intermediate plate

Second Step Dilute intermediate plate with 25 mL DMSO in each well

Intermediate Plate

Intermediate Plate

Agenda



9

Traditional Scaffold Modifications Fibrinogen Receptor Inhibitor IC50 = 29 µM

Poor stability, poor bioavailability, nonpatentable

IC50 = 3 µM

Poor stability, poor bioavailability

IC50 = 0.15 µM

Poor oral availability

IC50 = 0.067 µM

Excellent oral availability, good stability 10

But what to do if the structures are dissimilar?

Both compounds bind strongly to the GABAA receptor.

Diazepam

CGS-9896

These compounds are extremely different in structure but both have the same effect. Is there a way to reconcile this and generate information to make new drugs?

Pharmacophores • Describes the optimal binding of a protein to a ligand. • Shows how different structures bind to same site. • Designed from screening data.

12

GABAA Receptor Pharmacophore Hydrogen bond acceptor

Hydrogen bond donor Hydrophobic pocket

GABAA Receptor Pharmacophore Hydrogen bond acceptor

Hydrogen bond donor Hydrophobic pocket

Agenda



15

Real World Data – EphB4 Receptor Compound # IC50 Acoustic (µM) IC50 Tips (µM) 5 4 7 W7b 8 W5 6 W3 W1 9 10 W12 W11 11

0.002 0.553 0.003 0.146 0.003 0.778 0.004 0.152 0.004 0.445 0.006 0.087 0.007 0.973 0.012 0.049 0.014 0.112 0.052 0.170 0.064 0.817 0.158 0.250 0.207 14.400 0.486 3.030 14 compounds with structures and IC50 data.

Ratio IC50Tip/IC50ADE 276.5 48.7 259.3 42.5 111.3 13.7 139.0 4.2 8.2 3.3 12.8 1.6 69.6 6.2

Barlaam et al., WO2009/010794 Barlaam et al., US 7,718,653

16

Real World Data – EphB4 Receptor 2

Log IC50-tips

1

0 -3

-2

-1

0

-1

-2

1

2

The acoustic technique always provided a more potent IC50 value. The greater the distance from the red line, the greater the difference in IC50 values.

-3

Log IC50-acoustic 17

Experimental Process Flow

Acoustic Model 14 Structures with Data

Generate pharmacophore models for EphB4 receptor Tip-based Model

Initial data set of 14 WO2009/010794, US 7,718,653 18

AZ Pharmacophores Pharmacophore

Hydrophobic features

Hydrogen bond acceptors

Tip-based

0

2

1

0.80

Acoustic based

2

1

1

0.92

Tip-based pharmacophore

Hydrogen Observed vs bond donors predicted IC50

Acoustic-based pharmacophore

Experimental Process Flow Results

Acoustic Model 14 Structures with Data

Generate pharmacophore models for EphB4 receptor Tip-based Model

Acoustic Model Test models against new data Tip-based Model

Results Initial data set of 14 WO2009/010794, US 7,718,653

Independent data set of 12 WO2008/132505

20

Compounds Tested with Tip-based Pharmacophore Name

Tip-based IC50 Prediction (mM)

Tip-based IC50 Actual (mM)

W084.1

0.3488

0.297

W084.2

0.3806

0.456

W084.4

0.6994

0.374

W082.2

0.8392

0.808

W082.4

1.4989

6.270

W083

2.8229

0.198

W084.3

2.9119

0.473

W082.1

3.3829

1.120

WO81

NOT RETRIEVED

38.300

WO82.3

NOT RETRIEVED

1.780 Barlaam wo2008/132505

22

Tip-Based Pharmacophore – Predicted vs. Measured 8

R² = 0.0002

1.000 0.1

1

10

Measured Rank Order

Measured Tip-based IC50

10.000

7

R² = 0.1837

6 5 4 3 2

0.100

1 1

Predicted Tip-based IC50

2

3

4

5

6

7

8

Predicted Rank Order

The pharmacophore developed from tip-based data is an extremely poor predictor of measured activity. 23

Results of Testing Pharmacophores

Acoustic Pharmacophore

Tip-based Pharmacophore Poor correlation (R2