Computational Tools in Prediction of Drug Absorption and Distribution

Computational Tools in Prediction of Drug Absorption and Distribution Christel Bergström, PhD Department of Pharmacy, Uppsala University 0.00 -0.20 ...
Author: Jonah Parsons
1 downloads 0 Views 621KB Size
Computational Tools in Prediction of Drug Absorption and Distribution Christel Bergström, PhD Department of Pharmacy, Uppsala University

0.00 -0.20

L-Cystine L-Arginine Thiamine hRiboflavin L-Leucine Niacinamid L-Isoleuci L-Serine L-Methioni L-Proline Pyridoxine L-Lysine D-Calcium hPyridoxal L-Threonin L-Valine L-Tyrosine Glycine L-Phenylal L-Histidin L-Glutamic D-Glucose Choline ch Sodium Bic L-Aspartic L-Glutamin i-Inositol Para-Amino L-Hydroxyp L-AsparagiPapp L-Tryptoph Ascorbic a Hypoxanthi Tw Sodium eenNitac 80 Ace Sodium Pho Retinol DL-a-tocop Menadione Calciferol Magnesium Ferric Nicotinic Sodium Pyr Biotin Calcium ChGlutathion Magnesium L-Cysteine L-Alanine Sodium Pho Phenol Red SodiumFolic Chl Acid Potassium Vitamin B1

-0.30

O

-0.20

-0.10

0.00

0.10

10 t[2 ]

w *c [2 ]

0.20

IMD

0

-20

0.20

-10

0

Media component_PCA_061031_corr.M5 w *c[1] (PLS), m4-lowest VIP VIP[Comp. 2]

OH

DME RPM MCC

NCT

-10

10

t[1] Media component_PCA_061031_corr.M5 (PLS), m4-lowest VIP, PS-Media component_PCA_061031_corr YPredPS[Comp. 2](YVar Papp (cm/s))/YVarPS(YVar Papp (cm/s))

2.50

M19

1.20-05

2.00

OH O

Y V a r P S (P a p p (c m /s ))

1.10-05

1.50 1.00

N ia c in a m id C h o lin e c h T h ia m in e h D - C a lc iu m

L - A s p a r t ic L - G lu ta m ic G ly c in e L - S e r in e

0.00

R e tin o l a c D L - a - to c o p M e n a d io n e Tw een 80 S o d iu m A c e L - H is tid in M a g n e s iu m F e r r ic N it H y p o x a n th i

0.50

L - C y s t e in e R ib o fla v in A s c o r b ic a D - G lu c o s e C a lc ife r o l

N

O

N H

V IP [2 ]

H N

20

L - G lu ta m in L - L e u c in e V it a m in B 1 M a g n e s iu m L - P r o lin e L - M e th io n i L - A la n in e L -A s p a ra g i G lu ta t h io n

H N

Media component_PCA_061031_corr.M13 (PLS), Excl medium M199 t[Comp. 1]/t[Comp. 2]

Media component_PCA_061031_corr.M13 (PLS), Excl medium M199 X w*c[Comp. 1]/w*c[Comp. 2] Y

O

1.00-05 9.00-06 8.00-06

RPM DME

7.00-06 6.00-06

MCC

IMD NCT 6.00-06

7.00-06

8.00-06

9.00-06

1.00-05

1.10-05

1.20-05

YPredPS[2](Papp (cm/s))

Var ID (Primary)

N

Christel Bergström christel.bergstrom@ farmaci.uu.se

RMSEP = ---

PhysChem Forum 3, June 13

Outline • Background; ADMET(absorption, distribution, metabolism, elimination/excretion, toxicity) modeling • Computational tools used – Descriptor generation – Statistical methods

• Quality of experimental data • Case studies: – – – –

Aqueous solubility Membrane permeability Active transport mechanisms Absorption

• Conclusion Christel Bergström christel.bergstrom@ farmaci.uu.se

ADMET modeling

O

H N H N

N

O

N H O

OH

OH O

w *c [2 ]

0.20 0.00 -0.20

L-Cystine L-Arginine Thiamine hRiboflavin L-Leucine Niacinamid L-Isoleuci L-Serine L-Methioni Pyridoxine L-Proline L-Lysine D-Calcium hPyridoxal L-Threonin L-Valine L-Tyrosine Glycine L-Phenylal L-Histidin L-Glutamic D-Glucose Choline ch Sodium Bic L-Aspartic L-Glutamin i-Inositol Para-Amino L-Hydroxyp L-AsparagiPapp L-Tryptoph Ascorbic a Retinol DL-a-tocop Menadione Magnesium Ferric Hypoxanthi Tw Sodium eenNitac 80 Ace Sodium Pho Calciferol Nicotinic Sodium Pyr Biotin Calcium ChGlutathion Magnesium L-Cysteine L-Alanine Sodium Pho Phenol Red SodiumFolic Chl Acid Potassium Vitamin B1

-0.30

-0.20

-0.10

0.00

0.10

Media component_PCA_061031_corr.M13 (PLS), Excl medium M199 t[Comp. 1]/t[Comp. 2]

X Y

20 10 t[2 ]

Media component_PCA_061031_corr.M13 (PLS), Excl medium M199 w*c[Comp. 1]/w*c[Comp. 2]

N

IMD

0

DME RPM MCC

NCT

-10 -20

0.20

-10

0

Media component_PCA_061031_corr.M5 w *c[1] (PLS), m4-lowest VIP VIP[Comp. 2]

10

t[1] Media component_PCA_061031_corr.M5 (PLS), m4-lowest VIP, PS-Media component_PCA_061031_corr YPredPS[Comp. 2](YVar Papp (cm/s))/YVarPS(YVar Papp (cm/s))

2.50

M19

1.20-05

2.00 Y V a r P S (P a p p (c m /s ))

V IP [2 ]

1.10-05

1.50 1.00

N ia c in a m id C h o lin e c h T h ia m in e h D - C a lc iu m

R e tin o l a c D L - a - to c o p M e n a d io n e Tw een 80 S o d iu m A c e L - H is tid in M a g n e s iu m F e r r ic N it H y p o x a n th i

L - C y s t e in e R ib o fla v in A s c o r b ic a D - G lu c o s e C a lc ife r o l

L - A s p a r t ic L - G lu ta m ic G ly c in e L - S e r in e

0.00

L - G lu ta m in L - L e u c in e V it a m in B 1 M a g n e s iu m L - P r o lin e L - M e th io n i L - A la n in e L -A s p a ra g i G lu ta t h io n

0.50

1.00-05 9.00-06 8.00-06

RPM DME

7.00-06 6.00-06

MCC

IMD NCT 6.00-06

7.00-06

8.00-06

9.00-06 YPredPS[2](Papp (cm/s))

Var ID (Primary) RMSEP = ---

Christel Bergström christel.bergstrom@ farmaci.uu.se

1.00-05

1.10-05

1.20-05

Typical ADMET properties assessed in early process of drug discovery • Dissolution/solubility • Passive membrane permeation BCS (permeability) • Transport mechanisms • CYP metabolism • Specific toxicity Christel Bergström christel.bergstrom@ farmaci.uu.se

Computational tools: Descriptors

Christel Bergström christel.bergstrom@ farmaci.uu.se

Molecular descriptors – 2D

– – – – – – – –

Christel Bergström christel.bergstrom@ farmaci.uu.se

Size Ring structure Flexibility H-bonds Polarity Electronic environment Charge Lipophilicity (ClogP)

F N

H N

N N

O

Molecular descriptors – 3D Polar surface area (PSA): O, N and H bound to these heteroatoms Non-polar surface area (NPSA): Total surface area – PSA Non-polar atoms (C, halides etc) Partitioned Total Surface Area (PTSA): Surface area of a specific atom/functional group (eg. –COOH) sp3-O (-OH) and sp2-O (=O) plus H’s Christel Bergström christel.bergstrom@ farmaci.uu.se

Computational tools: statistics

Christel Bergström christel.bergstrom@ farmaci.uu.se

• Rule-based system/ decision tree • Correlations: linear/ sigmoidal • Multivariate problems – linear: • Multlilinear regression (MLR) • Principal component analysis (PCA) • Partial least squares projection to latent structures (PLS) • Multivariate problem – non-linear: • Non-linear PLS • Neural network (NN)

Multivariate data analysis: PCA & PLS • PCA: Multivariate data analysis of characteristics of the observations: – Groups/outliers/trends – Correlationa obs./descr. och descr./descr. • PLS; relates two matrices, X and Y, to eachother by linear mathematics • Instead of all variables that are included in the matrix, super variables with condensed information are extracted and used for prediction.

Christel Bergström christel.bergstrom@ farmaci.uu.se

• PLS-DA (discriminant analysis) uses the supervariables to separate between two groups

Quality of data; typical ”literature” data • Mixed sources and methods; introduces large noise • Understanding the applicability domain: usage of training set and the transparency of the models Training set n = 797 Test set n = 1587 r2 = 0.92 r2 = 0.67

Christel Bergström christel.bergstrom@ farmaci.uu.se

Huuskonen data set

MERCK data set

ADMET modeling: the flow Use training set/test set Method specifications: pH, temp, time points, additives

Experimental data Dataset selection

Drugs Diversity Range

1D, 2D, 3D christel.bergstrom@ farmaci.uu.se

new unknowns

PLS MLR Statistics NN

Molecular descriptors

Christel Bergström

VALIDATION Pred. vs Obs.

Modell new prediction unkowns

USE

Conc (M)

Case I: Models of aqueous drug solubility Solubility range determined; tamoxifen intrinsic 2.9 nM verapamil pH-dep 2 M

300 rpm

time (min)

ultracentrifugation

HPLC analysis

Bupivacaine

35

6 30

logS (uM)

S0 (µg/ml)

25 20 15 10

5 4 3

5 0 0.05

Christel Bergström christel.bergstrom@ farmaci.uu.se

0.1

0.2

0.5

mL

1.0

2 2.5

5.0

7.5

10.0

12.5

pH

Bergström et al., Pharm Res 2002

Global Models for Larger Datasets 2D descriptors logSobs (M)

2.5 0.0

3D descriptors 2.5

R2=0.75

0.0

2.5

R2=0.57

-2.5

-2.5

-5.0

-5.0

-5.0

-7.5

-7.5

-7.5

-7.5

-5.0

-2.5

0.0

2.5

-10.0 -10.0

-7.5

-5.0

-2.5

0.0

2.5

-10.0 -10.0

logSpred (M)

logSobs (M)

logSpred (M)

2.5

Consensus

0.0

R2=0.80

R2=0.78

0.0

-2.5

-10.0 -10.0

2D+3D descriptors

-7.5

-5.0

-2.5

0.0

logSpred (M)

-2.5 -5.0 -7.5 -10.0 -10.0

Christel Bergström christel.bergstrom@ farmaci.uu.se

-7.5

-5.0

-2.5

0.0

2.5

logSpred (M) Bergström et al., JCICS 2004

2.5

Local Models Result in Higher Accuracy

1,2

RMSEtr

1 0,8 0,6 0,4 0,2

christel.bergstrom@ farmaci.uu.se

bases

steroids

non-proteolytes

xanthines

ampholytes

barbituric acids

acids

ß-receptor antagonists

Christel Bergström

global consensus model

0

Bergström et al., JCICS 2004

Identifying poorly soluble compounds

8 6 4

Albe

tPS[2]

2

Tolf

0

Indo Felo

-2

Rimo

Glyb Glim Trog Carv Aste

-4

Tamo Dana

-6

-10

0 tPS[1]

christel.bergstrom@ farmaci.uu.se

Terf

Cinn

-8

Christel Bergström

Itra

Bergström et al., submitted

10

Case II: Models of membrane permeability (passive)

100

-4

80

-5

60 -6

40

-7

20

-8

0 -8

Christel Bergström christel.bergstrom@ farmaci.uu.se

-7

-6

-5

-4

Predicted log Pc (cm/s)

-3

-8

-7

-6

-5

-4 -3 -2 logPapp (cm/s)

Bergström et al., JMC 2003

FA (%)

Observed log Pc (cm/s)

-3

In vitro systems for prediction of fraction absorbed

Christel Bergström christel.bergstrom@ farmaci.uu.se

In silico prediction of permeability

In vitro

Christel Bergström christel.bergstrom@ farmaci.uu.se

In vitro

In silico

Case III: Active transport . Breast Cancer Resistance Protein (BCRP) Blood-brain barrier

Mammary gland Small intestine

Blood-placenta barrier • Liver • Kidney • Stem cells

Christel Bergström christel.bergstrom@ farmaci.uu.se

Experimental assay: Efflux inhibition

Christel Bergström christel.bergstrom@ farmaci.uu.se

No inhibitory effect:

Inhibition:

mitoxantrone is effluxed Î low intracellular fluorescence

mitoxantrone is retained in the cells Î high intracellular fluorescence

Computational modeling

PLS-DA

Common features pharmacophore modeling

Christel Bergström christel.bergstrom@ farmaci.uu.se

Matsson et al., submitted

Case IV: BCS prediction • To investigate accuracy of absorption prediction when both solubility and permeability are incorporated. • Used the WHO list of essential drugs • 15 compounds used for developing the two models, validated with 21 compounds.

Christel Bergström christel.bergstrom@ farmaci.uu.se

A BCS extended for the drug discovery process • High solubility: maximum dose given orally is soluble in 250 ml, pH 1-7.5 • High permeability: >80% absorbed • Low permeability:

Suggest Documents