High throughput solubility measurement with crystalline/amorphous ...

High throughput solubility measurement
with
crystalline/amorphous information
PhysChemForum5
19 June 2008
Kiyo Sugano
Pfizer
Kiyohiko.Sugano@pfizer.com

Outline
• Introduction of new solubility screen
• Results from 1700 measurements
• Strategy to apply the new method in drug
discovery
• Solubility calculation for oral absorption
simulation

What determines solubility?
Crystalline
Yalkowski’s equation
logSw0 Water
logSw0 = 0.8– 0.8– (m.p.-25) x 0.01 0.01 – logP logP
Lattice energy Hydration
Bile micelle partition
Amorphous
Dissociation
S
=
⎛
⎜1+
⎝
Henderson – Hasselbalch equation
Sw Sw = Sw Sw (1+[H+]/Ka)
+
log P
H C
⎞
micelles ,undissociated
micelles log Pmicelles ,dissociated
⋅10
+ ⋅ ⋅
⎟ Bile micelle
Ka Cwater
⎠
Dressman’s equation
logP logP mic,undissolved
=
mic,undissolved 0.74logP+2.29
+
[ H ] C
[ ]
+
micelles
tot
S0
10
Ka Cwater

Traditional kinetic solubility
• Start with DMSO sample solution
• Short incubation time
• Detect turbidity by nephelometry
• Precipitant is assumed to be amorphous
– This assumption is wrong

Kinetic solubility > thermodynamic sol
• Three possible reasons
– Solubilitization effect of DMSO
– Short incubation time
– Crystal/amorphous

New solubility assay
DMSO stock
1. Long incubation
2. Centrifuge
Filtration
•LC-MS
•HPLC
PLM
Final DMSO = 1%
(little effect on solubility)
Glass bottom plate

Validation using Marketed Drugs
Solubility from DMSO sample (log, μM )
1.0E+03
1.0E+02
1.0E+01
1.0E+00
1.0E-01
Incubation: 10 min
(kinetic solubility)
1.0E-02
1.0E-02 1.0E-01 1.0E+00 1.0E+01 1.0E+02 1.0E+03
1.0E+03
1.0E+02
1.0E+01
1.0E+00
1.0E-01
Incubation: 20 h
1.0E-02
1.0E-02 1.0E-01 1.0E+00 1.0E+01 1.0E+02 1.0E+03
Equilibrium solubility from crystalline (log, μM )
C rysta line
P artia ly C rysta line
N o C rystalO bse rve d

Concentration time profile
Precipitate out
Amorphous
Crystalline
~ 1000 fold
~ 4 fold
(mostly 2 fold)
From powder
(dissolve in, crystalline)
Conc.
Kinetic solubility
Incubation Time
5 min 24 hours

Solubility Ratio: polymorphs
25
Solubility Ratio
20
15
10
5
0
0 20 40 60 80
Compound Number
M. Pudipeddi, A. T. M. Serajuddin. J. Pharm. Sci. 2005, 94, 929–939.

Solubility Ratio: hydrates
25
Solubility Ratio
20
15
10
5
0
0 5 10 15 20
Compound Number
M. Pudipeddi, A. T. M. Serajuddin. J. Pharm. Sci. 2005, 94, 929–939.

Photo of 96 well plate

Auto PLM diagnostic
250
Intensity – position
C C’
1 mm
200
150
100
50
0
250
200
150
100
50
0
0 200 400 600 800 1000 1200
0 200 400 600 800 1000 1200
About 85% correct against human eye observation.

Outline
• Introduction of new solubility screen
• Results from 1700 measurements
• Strategy to apply the new method in drug
discovery
• Oral absorption simulation

In real drug discovery, does it work?
• In 2006, > 1600 compounds measured
– All pain project compounds at lead optimization
– 0.6 person x day/once a week
– Semi automation (No robot)
• Eye observation of crystalline/amourphous

All compounds (ca. 1700)
2
2
2
Log Solubility (μ m)
1
0
5
5
0
Log Solubility (μ m)
1
0
5
-1
-1
0 1 2 3 4 5
-1
1 2 3 4
ACD logP
Retention time (min)

Percentage of crystalline
precipitant
Compound number Number of crystalline Crystal %
All 1669 - -
< 150 uM a 1219 434 36
Project A a 625 248 40
Project B a 341 82 24
Project C a 130 73 56
Project D a 85 14 16
Project E a 38 17 45
a Compounds with > 150 μM solubility value were excluded from the
analysis due to uncertainty of crystal detection by PLM.

Three findings from 1700
measurements
• Solubility of crystalline is lower than that of
amorphous (Of course!).
• Solubility – lipophilicity relationship is vague.
– Even when the precipitant was amorphous.
• Percentage of crystal differed among chemical
scaffold.

Fact or Myth?
• In silico is good enough for solubility and
permeability. Let’s quite these assays”
• Similar “Myth” is also found for oral
absorption simulation
• Formulation is perfect. No worry about low
solubility.

Even logP/D calculation is not accurate
Octanol shake flask vs in silico calculation (across all project)
•Do •Do we we need need more more sophisticated calculation principle?
•Even •Even logD logDcalculation is is not not accurate with with the the current current
in in silico silico technology, how how solubility, Caco-2,etc
prediction can can be be better better than than logD logDprediction?
•With •With in in a chemical series, series, the the predictability may may be be
better. better.
Nagoya
(Shake Flask)
Alex Avdeef, Stefanie Bendels, Li Di, Bernard Faller, Manfred Kansy, Kiyohiko Sugano, Yukinori Yamauchi (Intercompany
collaboration) J. Pharm. Sci., 2008, 2893-2909

Outline
• Introduction of new solubility screen
• Results from 1700 measurements
• Strategy to apply the new method in drug
discovery
• Oral absorption simulation

Low solubility compound increasing
120
100
80
60
Frequency
40
20
0
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
1978
1977
1976
1975
Year
Number of publications containing the concept "poor solubility
drug" as of December 2006. Carried out using SciFinder®
Sugano et al., DMPK, 2007, 225-254

Solubility line-up
Full fledged discovery decision point
Development decision
Library design
Lead seeking
Lead
optimization
Candidate
selection
API form
optimization
Formulation
study
FIH
In silico
DMSO precipitation PLM method
Apparent solubility from powder material
pH solubility profile/FaSSIF/FeSSIF
Miniscale dissolution test/DP-system
In vivo studies

Chemical modification or DDS?
Does standard formulation approach work? (Milling/Salts)
Is the compound suitable for DDS technique?
What is possible and what is not possible?
Which has higher success rate?
Which is faster to clinical trial and launch?
Which is less resource intensive (human, manufacture etc)?
Computational
simulation might
help to
understand this.

Outline
• Introduction of new solubility screen
• Results from 1700 measurements
• Strategy to apply the new method in drug
discovery
• Solubility data for oral absorption
simulation

Scheme to calculate solubility in each GI tract
Measurement
Chemical structure
Melting point Hydrophobicity
(ΔH, ΔS melting)
K sp
pK a
Intrinsic equilibrium solubility
Bile micelles water
partition coefficient
pH and bile conc. (Av/SD)
•GI position
•Fasted/fed
•Species differences
•Individual differences
Standard HH HH equation
pH pH solubility profile profile
Modified HH HH equation
Solubility in in the the intestinal fluid fluid at at each each GI GI position
•Average value value
•Individual differences
Sugano et al., DMPK, 2007, 225-254

Glomme, A.; März, J.; Dressman, J.,. In Pharmacokinetic Profiling in Drug Research, Testa, B.; Krämer, S.;
Wunderli-Allenspach, H.; Folkers, G., Eds. Wiley-VCH: Zurich, 2006; pp 259-280.
Avdeef, A.; Box, K. J.; Comer, J. E.; Hibbert, C.; Tam, K. Y.,. Pharmaceutical Research 1998, 15, (2), 209-215.
Modified Henderson-Hasselbalch equation
10
Buffer
NaTC 7.5 mM, TC/PL = 4:1
solubility (mg/mL)
1
0.1
0.01
Dipyridamol
Buffer HH
HH for Biorelavent media
Using Buffer HH for
biorelevant media
0.001
3 4 5 6 7 8
If FaSSIF solubility (at pH
6.5) was used to calculate
the pH solubility profile
using the standard HH
equation, solubility at low
pHs would be over
estimated. Please check
your program simulation.
pH
+
+
[ H ] C
log P [ H ]
⎛
micelles
C
micelles ,undissociated
micelles
S
⎜
tot
= S0
1+
+ ⋅10
+ ⋅ ⋅10
⎝ Ka Cwater
Ka Cwater
log P
− log P ≈ 1 Base
log P
micelles,undissociated
micelles,undissociated
− log P
micelles,dissociated
micelles,dissociated
≈ 2
Acid
log P
micelles ,dissociated
⎞
⎟
⎠

Other precautions for simulation
• pH change at solid surface
• Bile micelles diffusion coefficient
• Free fraction? Or drug in bile micelles absorbed?
• Precipitation
• GI fluid volume
• Hydrodynamics
• Species differences of bile conc
• …
• …
I will discuss these items at this British Pharmaceutical Conference @ Manchester.
Please come and see me again!!!

Conclusion
• The new solubility assay is a beneficial asset for drug
discovery.
• Crystalline/amorphous information is important for
drug design and compound selection.
• Strategic approach is required to fix the low solubility
issues.
– Computational simulation might help.
• The modified HH equation is required for oral
absorption simulation
– There are many other precautions for simulation

Acknowledgement
• Pfizer Nagoya members
– Teruhisa Kato (Sandwich at present)
– Kentaro Suzuki
– Shiho Torii
– Arimichi Okazaki
– Atsushi Omura
– Takashi Mano (Sandwich at present)
• Pfizer Sandwich members
– Michael Cram
– Richard Manley
– Hurst Kelly