Detecting Drug Effects in the Brain

Detecting Drug Effects in the Brain

Heather Turner, Phil Brain and Foteini Strimenopoulou

Nonclinical StatisticsPfizer, UK

18 August 2011

Background

Aim: identify and characterise effect of drug on the brain

• Drug effect over timeI PK/PD model

• EEG experimentsI electrical activity in the brain

• Generalised Semi-linear Canonical Correlation Analysis(GSLCCA)

EEG

• Electrodes placed on scalp

• Monitor difference in voltage between baseline electrodeand others

• Produces virtually continuous signal

EEG Data

• EEG signal converted via FFT to power spectraI ”amount” of each frequency for each time sliceI multivariate response over time

FFT−−→

0 5 10 15 20 25 30 35

0.00

000

0.00

010

0.00

020

0.00

030

Examples of frequency periodograms

Frequency (Hz)

Pow

er (

µV2 )

●

● ● ●

●

●

●

●

● ● ●●

●

●

●●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●

●

●

● ●

●●

●●

●

●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

0−5 minutes120−125 minutes

PK/PD

• AssumptionsI drug level in brain follows pharmacokinetics model (PK)I brain response proportional to dose level (PD)

• Expected response over time follows PK model, e.g.

Double Exponential

β(exp(−k1t)− exp(−k2t))

Critical Exponentialβtexp(−k1t)

PK Models

0 10000 20000 30000 40000

−0.

20.

00.

10.

20.

30.

4

Time

Res

pons

e

Double ExponentialCritical Exponential

GSLCCA (in pictures)

0 5 15 25 350.00

000

0.00

020

Spectrum

Frequency (Hz)

Pow

er (

µV2 )

×

0 5 15 25 35

−50

00

500

1000

Frequency (Hz)

Coe

ffici

ent

Signature

−→

0 10000 20000 30000 40000

−0.

10.

00.

10.

2

Time

Res

pons

e

"Observed" Value

GSLCCA (in pictures)

0 5 15 25 350.00

000

0.00

020

Spectrum

Frequency (Hz)

Pow

er (

µV2 )

×

0 5 15 25 35

−50

00

500

1000

Frequency (Hz)

Coe

ffici

ent

Signature

−→

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●●●

●

●

●

●

●

●

●●●

●●●

●●

●●

●●

●

●

●

●

●

●●

●

●●

●●

●

0 10000 20000 30000 40000

−0.

10.

00.

10.

2

Time

Res

pons

e

Fitted Values

GSLCCA Method

• Canonical Correlation Analysis (CCA)I For matrices Y and X, finds loadings a and b to

maximisecor(Y a,Xb)

• Semi-linearI Y is the matrix of power spectraI X = X(t, θ) defined by PK model

• GeneralisedI linear coefficients b or nonlinear parameters θ may

depend on treatment factor

gslcca Package

• gslcca functionI specify PK model by name/formulaI specify which parameters vary by treatmentI control over data smoothingI partial CCA option

• plot, print, summary

Clonidine Experiment

• 4 treatmentsI ControlI Low doseI Medium doseI High dose

• 8 rats in 4-period cross-over design

• EEG data recorded for 12 hours post-dose

GSLCCA Analysis

Call:

gslcca(Y = spectra, formula = "Critical Exponential", time = Time,

subject = Rat, treatment = Treatment, separate = TRUE, ref = 1,

data = design, subject.smooth = 4)

GSLCCA based on 8 subjects

Data smoothed at subject level using 4 roots

Nonlinear parameters:

subject 35 subject 36 subject 37 subject 38

K1 Low Dose 7.5576 8.4252 7.8778 9.9125

K1 Middle Dose 7.8786 8.5137 8.0901 8.8885

K1 High Dose 8.9017 9.3213 9.0159 9.1980

subject 39 subject 40 subject 41 subject 42

K1 Low Dose 8.7217 8.0102 8.7103 8.1199

K1 Middle Dose 8.8546 8.5952 9.1854 8.3800

K1 High Dose 9.0439 9.1611 9.3047 8.9933

Fitted Modelplot(result, "fitted")

Observed + Fittedplot(result, "scores")

Signaturesplot(result, "signatures")

0 5 10 15 20 25 30 35

−50

00

500

Frequency (Hz)

Coe

ffici

ent

Signatures Corresponding to Different Subjects

Subject 35Subject 36Subject 37Subject 38Subject 39Subject 40Subject 41Subject 42

Normalised Signatures

0 5 10 15 20 25 30 35

−0.

4−

0.2

0.0

0.2

0.4

0.6

Frequency (Hz)

Coe

ffici

ent

Signatures Corresponding to Different Subjects

Subject 35Subject 36Subject 37Subject 38Subject 39Subject 40Subject 41Subject 42

Mean Signature

0 5 10 15 20 25 30 35

−0.

20.

00.

20.

40.

6

Frequency (Hz)

Coe

ffici

ent

Mean Signature

• contribution of powerat each frequency toPK curve over time

• assumed to be specificto the target drug isaimed at

Control/Inactive Signature

0 5 10 15 20 25 30 35

−0.

20.

00.

20.

40.

6

Frequency

Coe

ffici

ent

• if drug inactive, anydose ≡ control

• inactive drug has samesignature as control

Control/Inactive Signature

0 5 10 15 20 25 30 35

−0.

20.

00.

20.

40.

6

Frequency

Coe

ffici

ent

clonidinevehicle

• In this case drug clearlydifferent from control

• Drug is active - asexpected!

Comparing Active Drugs

0 20 40 60 80

−10

0−

500

5010

015

020

0

Frequency

Coe

ffici

ent

Drug ADrug B

• Two drugs targetingsame ion channel,different receptors

• Run t-tests to compareloadings at eachfrequency

Snapshot Analysis

0 20 40 60 80

0.0

0.5

1.0

1.5

2.0

2.5

3.0

Frequency

−lo

g 10(p

)

deltathetaalphabetagamma • P-values adjusted using

FDR

• Frequencies split intoconventional bands

Summary

gslcca package is in development on R-Forgehttps://r-forge.r-project.org/projects/gslcca/

Further work needed before release to CRAN, e.g.

• fitting PK/PD model to all rats simultaneously

• adjusting for control response