The integration of microarray information in the drug development process

643

The integration of microarray information development process Scott Inthe

Braxton*

and Tod Bedilion?

past year, microarray technologies

stage. Microarrays

being used for genome-wide

expression

polymorphism

screening

large nuniher

have moved

beyond the proof-of-principle large-scale

in the drug

si\,ely

are now

csprcssion

monitoring,

of

and mapping, and for

of probes

p:irallcl

ci”“ntific~ltion.

niolccular

biology

sc:llin,q technologies.

the evaluation of drug candidates. Addresses Syntenl Inc., 6519 Dumbarton Circle, Fremont, CA 94555, USA

IIIIO\VS ;I rcse:ircher

‘:e-mail:

[email protected]

tcncc ofniiiltipl~

‘e-mall:

[email protected]

sinj$e

Current

Opinion

in Biotechnology

1998,

of

hlicroarrays

step

to nic;lsurc

EST

information

is ciirrcntly

RA SNP

single

nucleotide

high-density

arc

oligonllcleotitlcs

‘1%~ pharmaceutical

scqiicnti:ill~

intiustr)

is bcinfi

the incrc:iaecl cost ofqztting

profitabilit);

the number

aping. dccreasc dc\-clopnient dccrc:ise nitics.

the Liilurc ‘I‘licre

dc\~clopnicnts lcnis l:or

thy

arc dcvcl-

clecrc~sc clinical

costs,

riitc. and de\,elop better drug opportu-

ha1.e hccn

;I

in rcccnt )ears

csamplc. the cuplosion an unprcccdcnted

for drug development. prlt \crccning

I>ct\vccn

and rcduccd

drug companies

of drugs

timr,

nnnil,er

of

technological

that ha\,c sol\,cd some proI,-

in the driig dc\ clopnient

produced

scjuee/.ed

;I drug to market,

maintain

~1ill need to incrc:isc

funnel.

but created others.

of genomic number

inforn~ation

of potential

\\‘ithin

in medical diag-

used

nianagcincnt

of patients.

has

Ius reduced

and IIOW ultr:l-tl’l’PLS

the tinic to screen \,alidated targctc

and

in

:icti\,c

Ilk,4

hybridization

iisc.

One

s)-nthcsizcd masked

dircctl~

photolabile

[ 1.11. the other, IISCS long l)Nr\ 3rd

by IY:K,

[Si].

‘Ike

short his

gcnc chips

cIsc\\ here [&I?]. stanci;trdizing dticiblc focuses

chcniistq

t)pically

substrate

gcncr-

using robotics

of microarray

and

and rc\.icwcd cxtcnsi\-cl? are being made in

g each tcchnolojiv

assays [13-1.5.16”].

ircliahlc

on the utility

using

:id\mces

and clii:intif!-in

:intl

on :i substrate

history

clcscribcd

Significant

short

fragments.

technological Bern

tech-

of

nuclcotidc

on ;i suitable

printed

ma!-

consists

ofniicroarra~-

into repro‘i’his

information

rc\ic\\

in the driig

de\~elopnicnt process.

targets

‘I‘hc dc\~elopment of high-through-

(H’I’I’S).

\vill I,c

Microarray technology ‘Ii\.0 prini:ir)-

polymorphism

Introduction

‘Ii)

of 1 iable drug

drug toxicit);

tag

nologies

drug rc\‘cnncs.

;I

(;enc

being iised to hc1p 1 al-

to help in the sclcction

and for the clinical

esis-

c:irryorit

of conditions.

unti to p-edict potmtial

nostic tests

expressed sequence rheumatoid arthritis

‘I’his

csprcssion

to probe for thr

for thousands

the nc\-t dec~ide. niicroarra!-s

Abbreviations

mKNX

or e\entiiall~

diagnosis

c Current

669

the

gene niut;itions.

candidates.

0958-I

no\xl engineering-interisi~.~ that t~sed to be done

with

http://biomednet.com/elecref/0958166900900643 Ltd ISSN

gene

the po\vcr

H~hridizations

idate drug targets.

Biology

or

combine

10,000 gcncs simultaneo~isly,

c\-prcssion

9:643-649

‘I’hcb- pro\,ide niassequrncing

can no\v be done 10,000 at 3 time.

one-Lit-3-time Ic\zl

arc attached.

h~,bridization-based

find potential

lead

Genetic analysis Recent applications analysis

support

in

the

the effort

arca of Lirray-hascd

to derix

practical

genetic

inforniation

htit has produced an c\-plosion of c01~~p0111~ds for the cntirc drux disco\xr) process, from mapping discase genes to stratifying patient populations in clinical to c\xluatc for dc\-clopnient. I:inally, many dnigs ;irc onl! defining allele-specific therapeutics. fractionall? cffccti\ c. or effecti\ e in only certain s11l~p0p11- trials. :incl ultiniatel~ lations and positi\c clinical outconic ma)- not Ix possihlc Genome-wide mapping \\.ithout siihscttinfi the population.

compo~~~idc,

\I’inzclcr

\I;in) steps in dtq de\~clopmcnt ;irc incft‘icient :rnd ne\\ nicthocls riced to Ix cmplo~cd that ;irc miich niorc prcdicti\.c of the cffccts of drilxs on humans. ‘l‘hcsc nicthods riced

nrc)

to md

he

rclati\-cl!

incrpcnsi\.e

(compared

deployable cad? in driif

to tlrug

kiil-

yclist

CT o/.

using

inap of the entire niiiltidrng

of the effects (icnctic

analysis

sti-atifying and

on cells,

is us&l

patients

3rc cspccted sion

of drugs for

are a poIvci+ul tissues.

for mapping cffccti\e

to be the dominant

or whole disease

treatment.

four

are small

t\~o-dimcnsioi~~li

genes and for cxprcs-

de\ elopnient.

surfaces

was then iised

l0c11s. ‘I’lie authors

Iv idcntificd

:ipplicd

confirming

gcncs. and localizing

kno\Vn

of mapping

3 high

in

resolution to map II the new

the loca-

the prcviousl>

gcnc to a 57 kh region that LV;IS subsequent-

II? ;I c;indid;itc

;ippro;ich

and dclction

anulysis.

animLlls.

\licroarrays

tcchnolog>-

gcnctic :inaI> sis in pharmace~itical

hlicroarra!s

nic;isurc

the cast

generated

map to a set of fi\.c Ioci. siniultancousl~ rinkno\\-n fifth

(;lolxil cl~angcs in jiene expression

‘Ilie!

genome which

rcsist:ince

tion of the

dc~~clopnicnt.

[ 17’1 dcnionstr;itc

oligo arrqs.

to ~7hich 3

Single nucleotide I’ol~niorj’hisiii has :ilso

polymorphisms

disco\.cr)

(typically

and chips

screening

found broad applic:ition

approach of c:it:iloging phisins

and

with

using

and assayin :: indi\.idunl

sinfilc

niicleotide

niicroarra)-s

the niorc

directed polynior-

pol~inorphisiiis

644

Pharmaceutical

Figure

biotechnology

1

(4

tcca

ram

THP.1 Control

(d)

loa

10332

1OuOrn

THP-1+ PM?!

63)

10

1w

lam

lum

lm!m

Thp.1 Control Current Opinion

Expression monitoring for 10,000 genes in a model of inflammation. (a) THP-1 cells were differentiated to a macrophage-like state with PMA for 48 hours. (b) Several genes characteristic of neutrophils disappear (cathepsin G, monocyte chemoattractant protein [MCPl] receptor, GCSF receptor), and macrophage-associated genes are upregulated (osteopontin, IL1 -/3, IL6, CD-I 4). (c) Subsequent stimulation with lipopolysaccharide (LPS) for four hours produced

a

I”

Eiwtechnology

profound change in gene expression patterns. (d) Several cytokines including SCYAP, MIPl-0, LD78, IL6, and MCPl are highly upregulated. In addition, the drug target of NSAlDs, prostaglandin G/H synthase, is highly upregulated. Ten of the top 23 upregulated genes are ESTs. The reproducibility of the assay is demonstrated in (e) where the same sample is hybridized against itself; 99% of the elements gave differential expression values < +/- 1.4.fold.

The integration of microarray information in the drug development process Braxton and BedIllon

(SNPs]). oligo

the

Initial

implemcntlitions

arrays covering mitochondrial

membrane

ha\,e

of the HI\‘-1 genome

segments gcnome

conductance

included

[l].

the

regulator

cystic

gene

fibrosis

645

Figure 2

short

[1X], trkms-

[ 191, and the beta

globin gcnc [ZO]. Hacia ul N/. [21] describe a refined method using tvwcolor competiti\-e hybridization to scan the 3.45 kb cxon 11 of the RR(X1 gene. ‘I’he high information density and ability to scan for nerv polo-morphislns v,irh

these

arrays

is particularly

important

in allelicall>

genes. such 3s 13R(Al. whew heterogcneows expected I0 be extcnsi\,e and unpredictable. ‘1%~ next step to extend many genes nndcrtaking

is nicely described

SW

screening

variation

and analysis

\

Tamoxifen

is

across

illustrated b)- the large collaborati\,e I~\- \I’ang @ l/l. [22’]. ‘I’his massi1.e

effort scanned almost 17,000 sequence tagged sites (Z/3 of \rAiich \\ere expressed secluence tags [WI’s]) across a total of 2 \Ib of sequence, and required 1-D unique fabrications of short oligo microarrays. Because of assav scnsiti\,ity and complexity constraints in the human genomc, probes were gcncrated from pools of IY:R products requiring specific primers for each of the 17,000 sequences snnqed across each of sc\zn indi\ iduals. ‘I’he rcsnlt turned up approximately 3000 SNP candidates. ‘I‘he survev information ;1 dedicated genotyping arr+ for.2 test subsct

\\‘a~ reduced to of .5.58 SNPs.

(:oniplimentary mapping utility is predicted for the method of genomic mismatch scanning (GhlS). GhlS selects rhc population of fragments that are identical b)- descent among members of a pedigree or lvirhin a population by a process of solution hybridization and enzymatic selection of heteroduplex-hybrids containing no mismatches. High resolution mapping information is derived from labeling and hyhridization to high densit); microarrays if there is ;I sufficient number of mapped and ordered clones. I:irst de\,eloped in the Iess complex )-cast genome [23,21], current refinements show promising hybrid enrichment in human samples [25,X], vhich could facilitate disease gene mapping [27-311 and genome-wide scanning for relevant allelic \wiation \+,here large pre-cataloged

sets of SNPs do not already

exist.

‘I’he potential value of these approaches incited rhc current expansion of commercial efforts and alliances based on col&ring and typing polymorphism information. Studies and statistical approaches using SNP and polymorphism data are becoming more sophisticated [32-341. Array-based methods clearly add needed leverage, but because of the current limitations on sensitivity, throughput, and accurac); the most effective formats probably remain to be developed.

Expression analysis Small arrays to large arrays to time-course studies Heller et a/. [3.5] provide the first example of profiling genes involved in disease progression using small 5%cI1NA microarrays, which include genes previousI) implicated in this process in the literature. In a model of inflammation, cytokines and chemokines were upregulatcd as expected. Comparison of primary chondrocytes and

Ethinylestradiol Current Op~monI” Biotechnology

Cladistic representation of Inflammation. U937

for different steroid treatments in a model cells were differentiated with 100 nM phorbol

mynlstlc acid (PMA) for 48 hours, followed by the combination of lipopolysaccharide (LPS) (10 ng/ml) and TNF-c( (0.1 ng/ml) for 24 hours with or without

one of 10 steroids.

Differential

expression

Information was generated using 10,000 gene microarrays. To compare the relatedness of different steroid treatments, a distance matrix was generated differentially experiments. the publically using

expressed Clustering avallable

Drawtree.

Inflammation hydrocortisone)

from the correlation

Philip

package.

The four steroids

(fluocinolone, gave quite

by the tight clustering

coefficients

for genes

by greater than threefold in any one of the was calculated using the Kitch program In The diagram

commonly

dexamethasone, stmllar expression

was drawn

used to treat prednisone, patterns,

and as shown

above.

s)-no\ iocytes from human rheumaroid arthritis (RA) tissue gave remarkably similar profiles upon srimulation \vith ‘I’NP and 11,-l, thus demonstrating the utility of primary cell culture as a model for authentic human disease samples. (:omparati\,e analysis of R.4 and inflammatory bowel disease revealed both qualitative similarities, underlying the similar inflammatory nature of both diseases, and distinct differences, representing inflammaGon in intlammarorl;

the more subdued level of bowel disease. ‘l’he repro-

ducibility of this assay and the underlying biology is demonstrated by the fact that samples from se\zral Ii.4 patients gave similar results. as did rcpcat samples from the same patient. DeRisi ectrl. [ 1.51used larger microarrays containing a mixture of known and anonymous genes to identif! the genes on chromosome 6 responsible for tumor suppression. Probes were derived from the 1YACC-903 melanoma cell line and ir’s non-tumorigenic counterpart IJACC-903(+6) suppressed b) addition of a normal chromosome 6. Of the 870 unique genes arrayed, 1.7% were downregulated and 7.3% were uprcgulated in rhe non-tumorogenic strain as compared to the melanoma cell line. In each case, Northern analysis verified the microarray rcsuhs. ‘I’he authors speculate that the specific genes that change their expression and camsc tumor suppression are candidates for therapeutic intervention.

646

Pharmaceutical biotechnology

Figure 3 Expression profile for several drugs and one Non-specific activator

15

Lead candtdate

Optimized

lead

Known

toxin

Best drug

toxin. HepG2

cells were treated with a variety of

agents, and the expression using a 10,000

(a)

(b)

(d

Cd)

In each treatment

(e)

represents

the target gene: 5-, 4-, respectlvely.

proflle IS captured

gene microarray. The first gene the upregulatlon

12.,14.,

To slmpllfy the representation,

genes with a dlfferentlal expression than twofold up- or down-regulated In addition to upregulating deacylating

of

15.fold only

of greater are shown.

the target, a histone

agent control (a) caused a variety of

genes to be turned on and off. (b) The lead candidate

only caused fourfold upregulation

of

the target. (c) Medlclnal chemistry produced compound which upregulated the target by 1 P-fold, and was consldered candldate

for development.

to be an excellent Microarray analysts

revealed that several unexpected signlflcant

changes

a

genes had

Cd)The

In expression.

expresslon

profile matched closely to the pattern

generated

by a known toxic agent. The cellular

toxlclty of the optimized lead was vertfted using an in vitro assay. Medicinal chemistry efforts, starting with a different lead compound,

Gene index Current Op,n,on ,n B,atechnology

;\n:tlyzing ,qencs

esperinicnrs cliallcngcs

on

the

microarra!s

in\,estigator

10.000

recruit

other

out

tlndcr

uviind

hcding.

from

and espind rhc information set for wcn the niosI well de\ eloped systcnis. I.sing such arrays in model ofintlamm~ltio~~. ‘I’HI’-1 cells (a monoc~tic ccl1 line) ucrc diffcrcnriatcd I0 ;I macroptiajic-lilcc state \vith phorrhc

lamp’

street

the 3

I~01 niyristic acid (I’hl.Ii) ~intl then actiutcd I\ ith 13 (:ocks. (; I’ortcr, lil~o~~ol~sacch~iritlc (S Braxton, .I Scilhanicr. iinpul~lished data). Ikspitc the well-explored mxiin of classic infl:ininiation moclels, se\xxrl l
rcasonablc

drug

of

gents,

one-at-a-time. pcrwnd cd

upon

t>> follmving

tibroblasts

not

authors

siniilar

of three

response distinct

sccrctc

applied

to

pitwrns of the

gents

through

tinic

to

resol~cd

inro

he

‘I’hc

first

of transcription response.’

;I series

of

:inal\-zc

an algorirhm

richness

phases.

in csprcssion

‘reprojiramming

feasible

expression ‘I’hc

10,000

healing

by chxijics as the

is

commllnicatioii).

\\otind image

it

so the

Ixiscd

gcncs

phase.

third

in

rhc

tilxd)lasts

In addition

dvnamic

rcsunx e

rcniodclin~

their

of

prolitkition

in

;111

~iind cktentling

verifying

pukis work on gcncs inwl\cd in w011ncl haling, the\ found a nunilxx of gcncs and parhv ays u host role in 1, oiin;l healing \\‘;I\ iincyxcrccl, but casity rationalid. ‘l’his i5 ;I proniincnt indicarion that microarray will pro\ iclc ddirional ~~~~l~~al~lc inforniation foi- e\en \\-cl]-charactcri/.cd svsum5. In addition, :I \xr! large set of nox~l gcncs \\ crc found I0 Ix diffcrcntially Iregiilatcd during the \\oiind-healing rcsponsc.

:is iman\ of die iinkiio~~~ii genes clusrcr \I itli uell-kno\\n patterns pro\idc the tii-st hints :il)our hwncs, their expression their

fiinction.

‘1%~ :it)ility

scs about

the

qqxo;ich

to functiond

of a

tiinction

art: three analyiis

‘I’tiere

list

‘I’he

I0 generate

more

no\ cl bs’l

\\,ill

assignnicnt

of

na-row allo\\.

no\ cI

h! pothc-

2 s;\‘swni;itic

I
targets.

‘I‘he II\Cof thcsc 10,000 gcnc urrays mxs extended (Iycr otc//.. personal CoiiiiiiiliiiCatioii) 1,) cuniining the time coiirsc rcspoiisc of fihrot~lasts to scruim. O\cr 3000 gcncs chaiijied expression II) tu otdd or more. and about half of thcsc gcncs arc i
cells invol\~~l

containing

‘step

to

produced a better drug candldate, as evidenced by the specific expresslon profile in (e).

In

chc

cell-signaling

phase

them to clustcl

(Kiwi dara

d N/..

possible. tion

iindcrlying

of

I:ii-e,

connecting

I>er

concepts that ni;ikc the funcc’f 01. (personal ~omm~inicarion~

high-density

about

10,000 rinic

niicroarra) gcncs

course

expcrinicnts

pro\ ides cnotigh data with dissect w~iind healing into dkxinct infi tools \vor!i to group roqxhcr csprcssion

patterns, of the

and

genes

s pro\.idc

Second. \\ ith

niicro;irr;i\~

sufficienr rcsoliition to phases. ‘I’hird. clustergcncs u ith siniilai

in so doin,q

in the

infol-nia-

~iliiIllt;inCoIIsI!..

analysis

funcCon

gcncrat-

dIouul a

tional

giw

hinrs

ahout

rinic.

Ross

the

group.

the

conipellin~

is dominated

factors,

rcfcrrccl

second

phase.

the

to

1n0lec11le~

I0

In contrast

to gcncraCng

(persona1

communic~ition),

niicroarra~s, NC:1

rithms

collcctcd

Iiinioi-

of

cell

l’iscn

patterns

through also

gene

e\prcssion

using

10.000 data

for

p:incI. Applyinfi the clusterinfi 07 u/. (personal commlinicatioii)

line

cut N/. fcnc

the

00

algothe\

The integration

demonstrarcd

rhat rissucs

of microarray

sources

from similar

information

had similar

in the drug development

clcotidc niicroarra~s

aII yeast gcncs were used to

containing

cyprcssion profiles. ‘Ike method is robusr enough to idcnriflporcntial misclassification of scleral cell lines. (Il~lstering bad on cxprcssion patterns iuatchcd \\ell I0 classification lxiscd on scnsitidty to wcr hO.000 drugs

ohser\,e the effects of thcsc conipo~~nds in )-cast.

[.%*I.

one half of the genes affected hy hrh

‘l’hi5

suggests

c\pression

profile

rhat str:rGfication

Ix

iiiav

of minors

prcdicti\x

IIy gcnc

_p-i% ,

of the mnscripts

inany

of drug sensiti\.ir).

of tlicst‘

rationalix. ‘I’he power of this approdi ckprcssion ‘I’hus,

profiling

is ticnionstratcd

to clusitcr similar

drug wndidatcs

cm lx

of their csprcssion

pimrns

toxicities

idcnrifial

can Ix

by the kihility of

acring drugs (l’igurc

prioritized

2).

txisecl on clustering

ii toxic agents. hlininy ing and patwrn recognirion

ix~sed

tlpon

clustering

with

thcsc rich data sets with clim3methods

is only in its infancy. In

the ncu fi\c years. de\~clopnient of these kinds ofalgorithmr

applied the pow3

of high-density

inicroamivs to nicmui-c ,g:cnoiiic-nide in vc;i\t kinder the nietahlic shift

respiration.

l-or

cxprcssion

profile

the

first

time.

has lxxn

c1)N.A

esprcssion changes from fernicntation I0 organism’s coniplctc

an

mctalx~lic

have dtered

ahout 700 gcncs upregulared

dou nregulkitcd

more

firnctions

than

t\wfold.

changed

in

cspression

and 1000 gents

(kncs

u,ith

eupccted

kmm n

is up-regulared

information

;I genetic

against honiolog

nilitant

product

is

has re\xxilcd

of

.3-lo-fold

upon

diausic

in ;I pdiwa); and that up-regulation to C:IIISC 3 functional up-rcgulution

so

or assigned funcrion,

(~oexpression

widcncc

shift.

precisicl>- Ixxause

\vcrc also altcrcd

general

of the entire

and abour

can Ix

the silwtion

site

inacri\ ad

might

of the mniendws

Additionally

the authors

the

step

againsr

will

u,hich

ix

midi

types

~ipproxh

rhese complcn regiilaumy

\\as used

:IS the basis for iteratiw and biological

c\,clin-dependent

sulxet

of the

I0 die drug nutniutuil.

is ;I

‘I’his

or

us):

cell

can

the

lx

used

dru g candidates

closely

I [sing a

kinmc-2.

analog

directed chemical librm

resulting

in potent inhibitors

High-density

proniisc

hetwecn

can Ix

oursel\xs

uprcgillate S 13raxton,

a particular unpublished

sion

of an optinii~ed

profile

discmwing

from

fact confirmed pounds

lx

Icad

compo~~~ici

and disco\wcd

in

\I ho were

~wuld \I I,earncd, llic cxpresfo

(I:igLirc

drug

a stud)

that

esprcssion

3). they

umuld protxiblv

on mother

candi-

data \I ith

niolcculc

conipoi~nds

ia ~~+/n analwis.

efforts

Ixxter

profiles.

(I’ Bawcrlc. 13~ comparing

able to infer that the conipound da clopnient

I>rugs of lhc

ar ‘liil;lrik.

;I small

receptor data).

mxic

nieusured.

rcalizcd

and scienrists

in

of

fingerprint

different

ken

using

profile

3s a ‘gold standard’

niicroarr;i~

has

inmxxtecl

fingerprints

cliniination

:ind, thcrcforc.

intcgraring

decisions

acri\xz

antiscnsc

esprcssion

match the expression

of

de\~elopnicnr

‘I’hc gene

deletion.

stratcgics.

rranscript

casts,

standal-d \vill bc more specific

‘l’hc

functions.

\vcrc

bc toxic, :I

‘I’hc\ then f~Auseci their srructural

;I much better

drug

class of conicandidate.

and

of integrating proccss.

approach, a purine

synrhesis

asa!;

method to

nct\\orks by dclction

into the drug discovers chcniistr)

arc possible.

one poRmtial

an example

nicclicin:~l

that it is

depdi of~:~“o”i”-\\,icle

Chips and drugs CT 01. [37’] demonstrated

(;ray

of

iniportdnt

uncharacter-

of anatyx5

gtmes.

rrditional

pro-

niorc straightfor\~ard

c‘obligator)

dominant-ncjiliti\,~

null-acti\,ity

that niosI

a\-ailable tiata-

prmidc

o\.Cr-;iCti~;ilio11 of mo rqylamy

analvsis

yeast

case, the

match that of the

I>)- ;i \ arictv of nicthods:

replaccnicnt.

data

100 of these g:cncs

in an)- of die pilblicly

data that rhcse

the

this

ml>- ;I minor

in the gcnctic

for targets that do not sm

rcprcssion.

p:ith\v:iy.

in the gaps in padi\\a\s. Note

niicroarra~

two

prot~lciii \vhcn die proposed drug targcr is essential.

In these

of one ,qene is cm&

information

niicroarra~

i~nra\cI

the

that the expression

~\ould closcl>

I~nfortiinatcl?;

ribo/.ynies.

‘I’hesc

to assign f~inction to thcsc pre\ iouslv

izccl gcncs, and fill

In

gcncs that changed in response

translation

As ncarl! e\wy yeast gcnc is represcnrcd on these amy. in theory all p;ith\fys ;irc‘ reprcscnted. In practice. man! gcncs do nor ) et ha\ c ;I function assigned. and mm) pathuays ha\c holes in them. ‘I’his is echoed in die fact that ;‘i’i”-o~irnatcl~ halfofthc differentially regulated genes ha\.e

txixs.

that

3 tcmpcratilre-sciIsiti\.~

dates than those that ha\,c ~astl)

do no1 ha\ c a honiolo~

not

inhibitors.

kinasc,

kinasc-2.

‘I’hc\ espectcd

drug treatnicn~s. all-cycle nient

(UcZXp

essential,

1~as used.

For

challcngc rhc dogma that there is one key. rawlimiting

no nanic

of~encs

kinasc

\I ith cells differently.

of c!clin-depcndclit

pirh\\2\

patterns.

ncarl) e\.cry gene in the known glycolytic

c~ample.

ib the xx

similar

to niodcl the cffccts of a pcrfcct drug. (;ra) c~tc//. [.i7’] compmxi the cspression profile of their drugs in yeast

One hops with

two

arc interacting

of

\virh about

captured.

7.5% of the 6200 )‘cast jicnes

Imets.

rhc

file of the gcnctic niutanr

(,I N/. [lb’]

expression

In an cfhrr

gcnc

the gycarest contrit~i~tion in die array fictd.

Going genome

1:ult)

bcmeen

expression

die

conipo~~nds. I:or in cuprcssion was easy to

the chnge

\\‘hat is more intcrcsring

compounds

mutant I)cliisi

gcncs,

in common (knc

alter-ed

t,v more than tnwfold.

with knon n drugs and powntial

knou

I\ ill niakc

acri\,c conipo~~nds

‘I\\m of thcsc

647

process Braxton and Bedilion

oligonw

‘Ii) make this approxh robust. large-scale gene esprcssion datahscs for the interacrion of chemical compounds and cells arc essential. Ihtabascs are important hcciuse the\ fxiliute interprctarion of coniplcs patterns and relationships

in large-scale

the most profiles

gcnc expression

ob\Aous databases induced

and for all

I~ood

‘I’he expression

will

I>>- a large \ aricty and IIrug profile

expcrinienm.

‘Ike

of

bc the gene expression of knou n roxic

Administration

agents

appro\ ed drugs.

of ;I drug candidate

can then

ix

648

Pharmaceutical

compared

biotechnology

to expression

uncover

potential

databases

to measure

efficacy

and

expression

broadly

analysis

applicable

ment

process.

is the first new technology

for many

Expression

of the gene analysis

interaction

tissues, cacy

expression

will serve

essential

Global

for treating

patients

and toxicity),

benefit apply

expression

population clearly

and genetic

be unethical

(:onversely,

some diseases

is not

In practice,

as they

ments.

some

genetic

‘I’he

treatment ‘I’hc

patients

to choose

inaccessible mutation, phenotyping an entirely

the

genomic to global

IO.

Marshall A, Hodgson J: DNA chips: an array of possibilities. Nat Biotechnol 1998, 16:27-31,

11.

Schena M, Heller RA, Theriault TP, Konrad K, Lachenmeler E, Davis RW: Microarrays: biotechnology’s discovery platform for functional genomics. Trends Biotechnol 1998, 16:301-306.

12.

Ramsay G: DNA chips: state-of-the 16:40-44.

13.

Wodlcka L, Dong H, Mittmann M, Ho MH, Lockhart DJ: Genome-wide expression monitoring in Saccharomyces cerevisiae. Nat Biotechnol 1997, 15:1359-i 367.

14.

Lockhart DJ, Dong H, Byrne MC, Follettle MT, Gallo MV, Chee MS, MIttmann M, Wang C, Kobayashl M, Horton H, Brown EL: Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat Biotechnol 1996, 14:1675-l 680.

15.

DeRisi J, Penland L, Brown PO, Blttner ML, Meltzer PS, Ray M, Chen ‘f, Su YA, Trent JM: Use of a cDNA microarray to analyse gene expression patterns in human cancer. Nat Genet 1996, 14:457-460.

systematic

of time,

paradigm

they

and effccti\,e

based

exploitation

a

on

tests measure-

is the ability from

the thcra-

diagnostic

on more objective

expression,

are

to restore cycles.

patients of

to treat

of previousI) gcnomic

in effect

point

extending

Icvel. L%‘eare on the verge of

for treating

References and recommended

human

massivley parallel genomics. Soence

gene

art. Nat Rote&no/

1998,

over

it could take years

segregated

information: gent:

drug treat-

one therap);

As we understand

makeup

of microarrays

to the molecular ntw

Johnston M: Gene chips: array of hope for understanding regulation. Curr B/o/ 1998, 8:R171 -RI 74.

on

to one drug, and if it

period

invention

to be based

promise

using

9.

microarrays.

gain

the cost sav-

have many potential

we have always

phenotype.

allowed

Wallraff G, Labadie J, Brock P, DlPietro R, Nguyen T, Huynh T, Hinsberg W, McGall G: DNA sequencing on a chip. Chemtech 1997, February: 22-32.

with Qene expression

the it will

is sure to put pressure

drug. In some casts,

bet\vccn

the

to segregate

py, wc will have ;I much better opportunity patient to health without iterative treatment

some

8.

how to

non-responders,

to hit upon the right drug treatment. relationship

Fodor SP: DNA sequencing 1997, 277:393-395.

risks (side

we understand

non-responders,

reason

within

to another

be

bc used only when

are usuc~lly assigned

effcctivc

switched

7.

effi-

uill

tests.

with no objective Patients

for

about

certain

information

only responders

appropriate

another.

Schena M: Genome analysis Bioessays 1996, 18:427-443.

and cells,

but incur all the risk. In addition,

ings for treating

6.

effectively.

versus

to treat

of dis-

information

the risks. When

developing

mcnts,

should

the

marker

information

more

into responders

no benefit

Shalon D, Smith SJ, Brown PO: A DNA microarray system for analyzing complex DNA samples using two-color fluorescent probe hybridization. Genome Res 1996, 6:639-645.

new drug targets.

involves

drugs

outweighs

allow

compounds

genetic

As the use of pharmaceuticals effwts

5.

to bc

develop-

component

as a surrogate

chemical

and will yield

toxicity.

drug will

of promising

between

or humans and

in the

microarrays

ease and the identification the

steps

High-density

measurement

Schena M, Shalon D, Heller R, Chal A, Brown PO, Davis RW: Parallel human genome analysis: microarray-based expression monitoring of 1000 genes. Proc Nat/ Acad SC/ USA 1996, 93:10614-10619.

toxic interactions.

Conclusion Gene

4.

disease.

reading

Papers of particular Interest, published wIthIn the annual period of review, have been hlghlighted as: l of special interest . . of outstanding interest 1.

Chee M, Yang R, Hubbell E, Berno A, Huang XC, Stern D, Wlnkler J, Lockhart DJ, Morris MS, Fodor SP: Accessing genetic information with high-density DNA arrays. Science 1996, 274:61 O-61 4.

2.

Fodor SP, Read JL, Plrrung MC, Stryer L, Lu AT, Solas D: Light-directed, spatially addressable parallel chemical synthesis. Science 1991, 251:767-773.

3.

Schena M, Shalon D, Davis RW, Brown PO: Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 1995, 270:467-470.

16. ..

DeRlsl JL, lyer VR, Brown PO: Exploring the metabolic and genetic control of gene expression on a genomic scale. Soence 1997, 278:680-686. This is the first example of genome-wide gene expressjon monitoring. The authors followed the temporal response of yeast undergoing a shift from fermentation to resplratlon. About half of the genes that change expressIon level have no name or currently recognized function. Several key enzymes were altered to force the redIrectIon of metabolltes Into the TCA cycle. Furthermore, nearly every TCA/glyoxylate cycle and carbohydrate storage gene was Induced during this diauxlc shift. In addltlon, the gene expression pattern was measured when TUPI, a transcnptlonal co-repressor was deleted, or YAPI, a transcripttonal activator was overexpressed. 17. .

Winzeler EA, Richards DR, Conway AR, Goldstein AL, Kalman S, McCullough MJ, Mccusker JH, Stevens DA, Wodicka L, Lockhart DJ, Davis RW: Direct allelic variation scanning of the yeast genome. Science 1998, 281 :l 194-I 197. Nearly four thousand biallellc markers were Identified by competttlve hybridization of two different strams of yeast to a high-density ollgonucleotide array. This was accompllshed without prior knowledge of sequence or the speclftc nature of the variations between the two strains and did not require the design and synthesis of speclflc PCR pnmers, nor the creation of new strains or constructs. These markers were used to map several loci with high resolutfon. 18.

Kozal MJ, Shah N, Shen N, Yang R, Fucinl R, Merlgan TC, Richman DD, Moms D, Hubbell E, Chee M, Gingeras TR: Extensive polymorphisms observed in HIV-1 clade B protease gene using high-density oligonucleotide arrays. Nat Med 1996, 2:753-759.

19.

Cronln MT, Fuclnl RV, Kim SM, Maslno RS, Wespl RM, Mlyada CG: Cystic fibrosis mutation detection by hybridization to light-generated DNA probe arrays. Hum Il/lutar 1996, 7:244-255.

20.

Yershov G, Barsky V, Belgovskiy A, Kirlllov E, Kreindlin E, lvanov I, Parlnov S, Guschin D, Drobishev A, Dublley S, Mirzabekov A: DNA analysis and diagnostics on oligonucleotide microchips. froc Nat/ Acad Sci USA 1996, 93:4913-4918.

The integration

21.

of microarray

information

Hacia JG. Edgemon K, Sun B, Stern D, Fodor SPA, Collins FS: Two color hybridization analysis using high density oiigonucleotide arrays and energy transfer dyes. Nuclefc Acids Res 1998, 2638653866.

23.

Nelson SF, McCusker JH, Sander MA, Kee Y, Modrlch P, Brown PO: Genomic mismatch scanning: a new approach to genetic linkage mapping. Nat Genet 1993, 4:l l-1 8.

24.

Nelson SF: Genomic mismatch scanning: current progress potential applications. Nectrophoresis 1995, 16:279-285.

25.

Cheung VG, Nelson SF: Genomic mismatch scanning identifies human genomic DNA shared identical by descent. Genomics 1998, 47:1-6.

26.

Mlrzayans F, Mears AJ, Guo SW, Pearce WG, Walter MA: Identification of the human chromosomal region containing iridogoniodysgenesis anomaly locus by genomic-mismatch scanning. Am J Hum Genet 1997, 61 :l 1 l-1 19.

and

the

27.

Felngold E, Brown PO, Siegmund D: Gaussian models for genetic linkage analysis using complete high-resolution maps of identity by descent. Am J Hum Genet 1993,53:234-251.

28.

Dupuis J, Brown PO, Siegmund D: Statistical methods for linkage analysis of complex traits from high-resolution maps of identity by descent. GenetIcs 1995, 140:843-856.

29.

Cheung VG, Gregg JP, Gogolin-Ewens KJ, Bandong J, Stanley CA, Baker L, Higgins MJ, Nowak NJ, Shows TB, Ewens WJ et al.: Linkage-disequilibrium mapping without genotyping. Nat Genef 1998, 18:225-230.

30

Xiong M, Guo SW: Fine-scale genetic mapping based on linkage disequilibrium: theory and applications. Am J Hum Genef 1997, 60:1513-1531.

development

process

Braxton

and Bedillon

649

31.

Guo SW: Linkage disequilibrium measures for fine-scale mapping: a comparison. Hum Hered 1997,47:301-314.

32.

Kruglyak L: The use of a genetic map of biallelic linkage studies. Nat Genet 1997,17:21-24.

33.

Zhao LP, Aragaki C, Hsu L, Quiaoit F: Mapping of complex traits by single-nucleotide polymorphisms. Am J Hum Genet 1998, 63:225-240.

34.

Devlln B, Risch N: A comparison of linkage disequilibrium measures for fine-scale mapping. Genomics 1995, 29:31 l-322.

35.

Heller RA, Schena M, Chai A, Shalon D, Bedilion T, Gllmore J, Woollev DE. Davis RW: Discoverv and analvsis of inflammatorv disease-related genes using cDNA microarrays. Proc Nat/ Acab SC; USA 1997, 94:2150-2155.

22. .

Wang DG, Fan JB, Siao CJ, Berno A, Young P, Sapolsky R, Ghandour G, Perkins N, Winchester E, Spencer J et a/.: Large-scale identification, mapping, and genotyping of single-nucleotide polymorphisms in the human genome. Science 1998,280:1077-l 082. 2.3 megabases of DNA were sequenced by gel-based systems, and resequenced from seven lndivlduals using about 150 Independent oligonucleotlde array designs. 2748 candidate SNPs were identified, slightly more than one per kilobase. The average spacing between markers was about 2cM, with an average heterozygosity of 34%. This IS a first step toward a human genome-wide SNP map, but covers less than 0.1% of the genome.

in the drug

markers

in

36. ..

Weinstein JN, Myers TG, O’Connor PM, Friend SH, Fornace AJJ, Kohn KW, Fojo T, Bates SE, Rubinstein LV, Anderson NL et al.: An information-intensive approach to the molecular pharmacology of cancer. Science 1997, 275:343-349. -.. The NU has screened more than 60,000 compounds against 60 well-characterized tumor cell lines. This paper tries to make sense of the vast amount of data, and various Interesting patterns emerge. Tumor cells with similar molecular lesions respond to similar drugs. Clustering information IS useful for determination of the mechanism of actlon of a new compound. Multidlmenslonal structure-activity relationship (SAR) informatlon from this set of data will be applied to large chemical Ilbraries, and better anti-cancer drugs are more likely to be found. 37. .

Gray NS, Wodicka L, Thunnissen AM, Norman TC, Kwon S, Esplnoza FH, Morgan DO, Barnes G, Leclerc S, Meijer L et al.: Exploiting chemical libraries, structure, and genomics in the search for kinase inhibitors. Science 1998, 281:533-538. The authors measured the gene expression changes in yeast induced by several inhibitors of the kinase Cdc28p. Although several genes changed In common for these treatments, more than half of the mRNA changes were distinct. Gene expression monitoring suggests that these compounds inhibit the target to different degrees, or that other targets are also affected. Comparison of the gene expression pattern of drug treatment to yeast with a temperaturesensitive mutation In Cdc28p should have yielded similar patterns, but did not on the whole. These results were confounded by incomplete gene inactivation and the bevy of changes associated with heat shock and cell cycle arrest. The approach outlined here should be more practical when a drug interacts with only one target, and the target can be specifically ablated.