Current Thinking in Ensuring the Quality of

Compound Collections

Zoe Blaxill

Zoe.Z.Blaxill@gsk.com – for any questions

Property of GlaxoSmithKline

mailto:Zoe.Z.Blaxill@gsk.com

Introduction

Historical situation

–Emphasis on:

–Ensuring compounds had high purity and stability

–Minimising water uptake

–Less emphasis on structural properties

–Numerous examples of similar compounds or analogues in collections

–Compound long-term storage at single (10mM) concentration

Preferred Properties of Compounds are Evolving

Property of GlaxoSmithKline

500

5

Limit of “Lipinski Rule of 5” Space”

Optimal drug-like space

Fragment space 200

0

350

Lead-like space

3

Many older screening compounds in

“drug-like” space

Now a shift to lead-like space to help

deliver better candidates

Lead-like compounds may need to be

screened at higher concentrations

In the extreme, Fragment Based Drug

Discovery (FBDD) uses very small

molecules at very high concentrations

Leeson, Nat. Rev. Drug Disc., 2007, 881

Wenlock, J. Med. Chem., 2003, 1250

Ritchie, Drug Discovery Today, 2009, 14, 1011

Three key components of FBDD

SCREENING &

CONFIRMATION

Low MW (~1mM)

Crystallography

And other high conc. assays

Computational Chemistry

Fragment to Lead

Medicinal Chemistry

FRAGMENT LIBRARY

STRUCTURE-GUIDED

OPTIMISATION,

LIGAND EFFICIENCY

Chessarie and Woodhead, Drug Discovery Today, 2009, 14, 1011

Safety profiling built into screening cascades to try

and reduce late stage attrition Reasons For NME Termination By Stage

2005-2009 Industry Portrait

Hepatotoxicity Genotoxicity

GreenScreen

assay licensed to

identify

Genotoxicants

Cell Health assay

detects 70% of

known

hepatotoxicants

Impact on Compound Management and Analytical QC

Requirement to store and

process high concentration

DMSO solutions.

Implications for process and

inventory systems.

Need to consider stability and

solubility in DMSO.

Need to consider precipitation.

Compounds are more polar.

Compounds have fewer or

weaker chromophores.

RT≤0.2 RT 0.2 < X ≤ 0.4

2009 2010

2011 2012

Impact on LC-MS Generic QA Method

High Concentration DMSO solutions

Generally, the need for high concentrations is driven by the limited DMSO

tolerance in biological assays.

Examples: Fragment-based screening & Tox assays (e.g. hERG)

1 mM 10 mM 100 mM

2% 20 µM 200 µM 2 mM

1% 10 µM 100 µM 1 mM

0.5% 5 µM 50 µM 500 µM

0.1% 1 µM 10 µM 100 µM

Top Assay Concentration for Stock Concentration of:DMSO

Tolerance

1

10

100

1000

Cip

rofl

oxa

cin

Spar

flo

xaci

n

Sem

atili

de

Pro

cain

amid

e

Nif

ed

ipin

e

Ce

tiri

zin

e

Ph

en

yto

in

Gre

paf

loxa

cin

Sota

lol

Eryt

hro

myc

in

Eryt

hro

myc

in i.

v.

Cla

rith

rom

ycin

Dip

he

nh

ydra

min

e

Cib

en

zolin

e

Nit

ren

dip

ine

Dilt

iaze

m

Me

flo

qu

ine

Fexo

fen

adin

e

Am

itri

pty

line

Fle

cain

ide

Imip

ram

ine

Flu

oxe

tin

e

Ted

isam

il

Ke

toco

naz

ole

Dis

op

yram

ide

Ch

lorp

he

nir

amin

e

De

sip

ram

ine

hER

G I

C5

0 (µ

M)

Adapted from de Bruin et al., Eur Heart J, 2005 Hajduk and Greer, Nat Rev DD, 2007

What are we looking for in high concentration solutions?

Stability

Ilouga et al, JBS 2007

6mM DMSO solutions

Zitha-Bovens et al, JBS 2009

2mM DMSO solutions

Blaxill et al, JBS 2009

10mM DMSO solutions

What are we looking for in high concentration solutions?

Solubility in DMSO

Solubility in Buffer

0%

10%

20%

30%

40%

50%

% o

f co

mp

ou

nd

s

<5%

5-1

0%

10-2

0%

20-5

0%

50-8

0%

80-1

20%

>120%

Ratio of buffer to DMSO

concentrations, %

Popa-Burke et al., Analytical Chem. 2004

3mM solutions, 5% DMSO in pH7.4 PBS

GSK in-house data, Holyoak et al.

Looked at Ratio of Buffer Conc/ DMSO Conc and compared

this ratio to the Measured Aqueous solubility

Balakin et al, 2004

Up to 20% of compounds in

commercial libraries are poorly

soluble in DMSO at 10mM

0

10

20

30

40

50

0-20% 20-50% 50-80% 80%+

% of compound in Buffer

% o

f co

mp

ou

nd

s

0-30 uM Aq.Solubility (LOW)

30uM+ Aq. Solubility

How is “stability” measured?

UV peak detected at the same retention time as an ion chromatogram corresponding to the expected mass.

Purity percentage calculated from DAD chromatograms.

Purity measured at Tzero and Tnow.

Defined as a reduction in the initial purity value.

DAD-UV Autosampler

(384-well plate)

UPLC

(separation) MS Purity

Stability

Compound Selection

Molecular Weight

Selected compounds

400 @ 10mM

250 @ 40mM

Out of:

40,000 @ 10mM

8,500 @ 40mM

80

0

10mM

40mM

Molecular weight

80-89

90-99

100

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

80

Init

ial p

uri

ty

Current purity

40 mM 10 mM

80-89

90-99

100

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

80

Init

ial p

uri

ty

Current purity

Very reproducible method.

No difference between overall rate of degradation of 10mM stocks vs. 40mM stocks.

Overall results

Fragments (2000 compounds)

General discovery compounds (500 compounds)

Solubility: analysis of precipitated samples

Two main types of 100mM samples investigated

95.2%

4.8%

84.3%

15.7%

Fragments

4.8% precipitates

No 10mM solutions made

General discovery

15.7% precipitates

No ppt from compounds when a

corresponding 10mM solution was available

-1%

1%

3%

5%

7%

9%

11%

13%

15%

17%

19%

21%

23%

25% 30

0

MW

All

Precipitates

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

70

0

MW

All

Precipitates

Fragments Discovery Compounds

0%

4%

8%

12%

16%

20%

24%

28%

32%

36%

40%

7

clogP

All

Precipitates

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

7

clogP

All

Precipitates

9%

2% 3% 0% 1%

2%

25%

57%

4% 2% 2% 1% 2%

4%

21%

-10%

0%

10%

20%

30%

40%

50%

60%

% C

om

po

un

ds

Precipitates

All

Purity

Fragments Discovery Compounds

Initial purity measurement same

for precipitated and non-

precipitated samples.

Clear differences in

concentrations for the precipitated

samples.

No apparent phys-chem

properties differences between

precipitated and non-precipitated

samples for fragments.

0% 3%

8%

21%

68%

6% 3%

7%

17%

67%

0%

10%

20%

30%

40%

50%

60%

70%

80%

% C

om

po

un

ds

Precipitates

All

Purity

8% 10%

12% 13%

26%

30%

1%

27% 29%

22%

8% 6%

2%

5%

0%

5%

10%

15%

20%

25%

30%

35%

% C

om

po

un

ds

Concentration

All

Precipitates

High Concentration DMSO solutions

Fragments @ 100mM Discovery Compounds @ 100mM

Level of precipitation lower for fragments versus general discovery compounds.

Majority of “precipitates” are actually solids not dissolving in DMSO, rather than

solutions that precipitate out in time and/or F/T cycles.

Precipitates cause tip clogging and carry-over in Compound Management.

Precipitated samples could not be solubilised through sonication, but dilution and

sonication solubilised all samples.

Brooks’ Tube Auditor™

The Tube Auditor (TA) is designed for high-speed, non-contact volume

measurement and precipitate detection for SBS format microtubes.

TA uses high resolution camera to acquire images of tubes

TA software then analyses defined “regions of interest” to identify the height of

the meniscus, and to confirm the presence or absence of precipitate and tube

caps

Measure

Volume

Identify

Empty

Tubes

Detect

Precipitate

Check if

Caps

Present

Rectangular regions of interest

superimposed on tube image

Immediate Processing Quality Control (IPQC)

at Time Zero

Compound

synthesis or

supply from

other source

Registration,

dissolution &

loading into STS

Downstream

compound handling

& repeat dispensing

Mandatory publication

of compound QA

• Confirmed compound

identity and purity at

point of registration

• Original QA data

available for

comparative analysis

Downstream UV testing

• Analysis of downstream

samples, or after repeat

dispensing from Small

Tube Store.

• Confirms identity &

conc. Detects significant

purity changes

t=0 test after STS load

• Confirmed compound

identity/purity and conc.

after dissolution/loading

into Small Tube Store.

• Generates comparative

reference data for t>0

testing

• All published literature indicates routine compound QC at a single time-point (except for stability studies).

• All analysis is on the stock solutions, generally when the solution is first made.

• Is that analysis indicative of what is actually screened?

• If compounds crash out of solution (water absorption or low DMSO solubility), decompose in storage, are being under- or over-dispensed on liquid handling equipment, are being switched during handling, basically everything that could happen post-stock creation,

The initial QC will not tell you anything about what was actually screened!

QC of Assay-Ready plates

UV scanning method

Receive

dry

sample

Make

10mM

stock

Make 2D tubes

For long-term storage

DMSO

Serial dil n Mother

Analytical

QC plate Baseline

(T 0 ) UV - Vis

Identity, purity, concentration

determination

All scans will be

compared to it Retrieve tubes

from store for

cherry-picking

Cherry-pick

into mother plate

Serially-dilute

mother plate

Replicate into

assay-ready plates

Tassay

UV-Vis

0

0.2

0.4

0.6

0.8

1

1.2

1.4

280

290

300

310

320

330

340

350

360

370

380

390

400

410

420

430

440

450

460

470

480

490

500

Wavelength (nm)

Ab

so

rba

nc

e

Tassay

T0

Slope => direct correlation concentration differences

R2 (or other match measure) => direct correlation with identity

Correct compound and conc. => slope = R2 = 1.0

280-500nm, every 10nm

y = 0.9962x

R2 = 0.9993

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 0.2 0.4 0.6 0.8 1 1.2 1.4

Absorbance Tassay

Ab

so

rba

nc

e T

0

Absorbance of Tassay vs T0

slope

The method

Tref is always magenta

Tassay being compared blue

All other scans are green

Plate map

Link to Mismatch Report

(download to Excel

option)

Links to

each well

spectra

Software

30 failures in a Tzero

plate. Mismatch report

indicated that all were

for 30 identical LNB’s

on that plate –

registration error

corrected.

Examples

• Multiple assay plates made.

• 2 look fine, 1 is an empty well.

• Error could be corrected prior to sending plate for screening.

Examples

Empty well

Examples

• Program team observes discrepant data and assumes due to different compound sources.

• Re-orders compounds from all solutions available to screen in parallel. • Scanned all Tassay plates that contained these compounds. • Program team found the cause to be due to assay artifacts.

How precipitation affects serial dilutions

13 100mM compounds with precipitate and identical 13 compounds at 10mM

without precipitate were put in column 1.

The supernatent only was hand pipetted and put in column 1.

A 3-fold serial dilution was done on the FX to create a plate.

Columns 8 and 17 were found to have similar concentrations of about 100uM.

UV scan was used to compare the compounds after the serial dilution.

Wavelength

Ab

so

rba

nce

7 were a perfect

match

(R2 ~1 and slope ~1)

Supernatant Plate

The presence of precipitate in the first column did not make any difference to

the serial dilution.

Surprisingly, the concentrations of the precipitated solutions were lower than

what was calculated from nominal.

Conclusion

Compound Management and Analytical Chemistry have an ongoing challenge to

be able to meet the demands of a changing compound collection.

Alternative QA methods for polar compounds need to be investigated.

Stability of higher concentration DMSO solutions has not been found to be an

issue.

Less precipitation has been observed in fragments than general discovery

compounds at high concentrations.

No apparent physical chemistry property differences between precipitated and

non-precipitated samples for fragments.

Serial dilutions are not affected by precipitates.

UV scanning of assay plates is a good method of process control.

Acknowledgements

Ioana-Popa-Burke – SMTech RTP

Neil Hardy – SMTech – UK

Ian Churcher – Discovery Sciences - UK

UV scan team – all sites

SMTech – all sites

Many others

Property of GlaxoSmithKline