describing biological protein interactions in terms of protein … · 2008-08-07 · prebind, a...

Describing Biological Protein Interactions inTerms of Protein States and State TransitionsTHE LiveDIP DATABASE*

Xiaoqun Joyce Duan, Ioannis Xenarios, and David Eisenberg‡

Biological protein-protein interactions differ from themore general class of physical interactions; in a biologicalinteraction, both proteins must be in their proper states(e.g. covalently modified state, conformational state, cel-lular location state, etc.). Also in every biological interac-tion, one or both interacting molecules undergo a transi-tion to a new state. This regulation of protein statesthrough protein-protein interactions underlies many dy-namic biological processes inside cells. Therefore, under-standing biological interactions requires information onprotein states. Toward this goal, DIP (the Database ofInteracting Proteins) has been expanded to LiveDIP,which describes protein interactions by protein states andstate transitions. This additional level of characterizationpermits a more complete picture of the protein-proteininteraction networks and is crucial to an integrated un-derstanding of genome-scale biology. The search toolsprovided by LiveDIP, Pathfinder, and Batch Search allowusers to assemble biological pathways from all the pro-tein-protein interactions collated from the scientific litera-ture in LiveDIP. Tools have also been developed to integratethe protein-protein interaction networks of LiveDIP withlarge scale genomic data such as microarray data. An ex-ample of these tools applied to analyzing the pheromoneresponse pathway in yeast suggests that the pathway func-tions in the context of a complex protein-protein interactionnetwork. Seven of the eleven proteins involved in signaltransduction are under negative or positive regulation of upto five other proteins through biological protein-protein in-teractions. During pheromone response, the mRNA expres-sion levels of these signaling proteins exhibit different timecourse profiles. There is no simple correlation betweenchanges in transcription levels and the signal intensity. Thispoints to the importance of proteomic studies to under-stand how cells modulate and integrate signals. Integratinglarge scale, yeast two-hybrid data with mRNA expressiondata suggests biological interactions that may participatein pheromone response. These examples illustrate howLiveDIP provides data and tools for biological pathwaydiscovery and pathway analysis. Molecular & CellularProteomics 1:104–116, 2002.

Understanding protein-protein interactions is crucial to in-tegrated biology. The availability of the genome sequences of100 species, ranging from bacteria to human, poses to biol-ogists two major challenges of interpreting this blueprint oflife, 1) understanding the function of each gene product, and2) understanding phenotypes through the biological interac-tions of the gene products. Large scale and high throughputexperimental techniques have been developed to addressthese questions by acquiring data on the whole genome,instead of just a few genes. One such example, an integratedgenomic and proteomic analysis of a systematically perturbedyeast galactose utilization pathway, suggested the impor-tance of analyzing both mRNAs and proteins for understand-ing biological systems (1). Protein-protein interactions are acrucial component of such integrated biology. The classicalview of protein function focuses on the action of a singleprotein molecule, its biochemical activity or molecular func-tion. An expanded view defines the function of a protein in thecontext of its network of interactions (2). Each protein inter-acts with several partners that also interact with other pro-teins. All these interactions connect proteins into an extensiveand complex web. Every protein functions in the context ofthis web of interacting molecules, and its interactions withother molecules define how its biochemical activities are uti-lized and regulated in the related biological processes.

Genome-scale analysis of biological systems requires easyaccess to information on a large number of protein-proteininteractions (1). Much of this information is embedded withinmany scientific articles and is hard to retrieve in an organizedfashion. These facts call for new ways of representing suchbiological knowledge, as also noted by other researchers (3,4). Several databases have been developed that provide web-accessible information on protein-protein interactions, assummarized in Table I.

A large proportion of known protein-protein interactions weredetected by genome-scale yeast two-hybrid assays (5). Theseinteractions indicate association of proteins for a certain periodof time and can be termed physical protein interactions. Phys-ical interactions may include nonspecific interactions, whichdo not have biological significance. A biologically importantinteraction requires the interacting partners to be in specificprotein states. It also causes transitions in the states of one orboth of the interacting partners (such as phosphorylationand/or activation). A biological interaction regulates the func-tion of the interacting proteins or transmits a signal from oneprotein to another and underlies all the dynamic biologicalprocesses in cells. Such interactions can be termed biologicalprotein interactions. DIP, the Database of Interacting Proteins

From the Howard Hughes Medical Institute, UCLA-DOE Laboratoryof Structural Biology and Molecular Medicine, University of California,Los Angeles, Los Angeles, California 90095-1570

Received, October 29, 2001Published, MCP Papers in Press, January 9, 2002, DOI

10.1074/mcp.M100026-MCP200

Research

© 2002 by The American Society for Biochemistry and Molecular Biology, Inc.104 Molecular & Cellular Proteomics 1.2This paper is available on line at http://www.mcponline.org

TAB

LEI

Ab

rief

sum

mar

yof

web

-acc

essi

ble

dat

abas

esco

ntai

ning

info

rmat

ion

onp

rote

in-p

rote

inin

tera

ctio

ns

Dat

abas

ean

dU

RL

Pro

tein

-Pro

tein

inte

ract

ion-

rela

ted

cont

ents

Uni

que

feat

ures

Com

men

tsD

ata

que

ries

Alli

ance

for

Cel

lula

rS

igna

ling

(AFC

S)

ww

w.c

ellu

lars

igna

ling.

org/

“Min

iMol

ecul

eP

ages

”co

ntai

nlit

erat

ure-

der

ived

info

rmat

ion

abou

tsi

gnal

ing

mol

ecul

esin

clud

ing

inte

ract

ions

.

Det

ails

abou

tsi

gnal

ing

pat

hway

sst

udie

db

yth

eal

lianc

e;re

crui

tm

emb

ers

from

scie

ntifi

cco

mm

unity

toau

thor

ind

ivid

ualA

CFC

mol

ecul

ep

age.

Par

tof

aco

ncer

ted

mul

tidis

cip

linar

yre

sear

chal

lianc

eto

und

erst

and

cell

sign

alin

gus

ing

two

sele

cted

mod

elsy

stem

s.O

nly

mol

ecul

esan

dp

athw

ays

ofin

tere

stto

the

Alli

ance

are

incl

uded

inth

ed

atab

ase.

Bro

wse

orse

arch

pro

tein

list;

bro

wse

pat

hway

map

s.

The

sign

alin

gm

aps

are

stat

ican

dd

raw

nin

div

idua

llyan

dm

ayno

tb

eco

mp

lete

.

Am

aze

ww

w.e

bi.a

c.uk

/re

sear

ch/a

maz

e/

Clic

kab

lesi

gnal

ing

map

san

dle

gend

sp

rovi

de

info

rmat

ion

abou

tsi

gnal

ing

pat

hway

s.P

athw

aygr

aphs

.

Pro

toty

pe

pat

hway

anal

ysis

tool

s.A

dat

am

odel

tod

ealw

ithin

form

atio

non

met

abol

icp

athw

ays,

gene

regu

latio

n,su

b-

orsu

pra

cellu

lar

loca

tions

,an

dtr

ansp

ort.

Bro

wse

pat

hway

grap

hs;

pro

toty

pe

pat

hway

anal

ysis

tool

s.

Bio

mol

ecul

arIn

tera

ctio

nN

etw

ork

Dat

abas

e(B

IND

)w

ww

.bin

dd

b.o

rg

The

thre

em

ain

dat

aty

pes

are

inte

ract

ions

,m

olec

ular

com

ple

xes,

and

pat

hway

s.

BIN

DB

LAS

T;b

last

sear

chag

ains

tm

olec

ules

inB

IND

.P

reB

IND

,a

dat

am

inin

gto

ollo

catin

gp

aper

sth

atm

ayco

ntai

nin

form

atio

nab

out

inte

ract

ions

rela

ted

toa

que

ryp

rote

in.

Aco

mp

lete

dat

am

odel

des

igne

dto

stor

efu

lld

escr

iptio

nsof

inte

ract

ions

orch

emic

alre

actio

nsb

etw

een

any

two

mol

ecul

esof

any

typ

e.

Sea

rch

inte

ract

ions

,co

mp

lexe

s,or

pat

hway

sb

yp

rote

inna

mes

orte

xtq

uerie

s.

Vis

ualiz

e-in

tera

ctio

n,Ja

vaap

ple

td

raw

ing

sche

mat

icd

iagr

amin

tera

ctio

ns.

Cel

lSig

nalin

gN

etw

orks

Dat

abas

e(C

SN

DB

)ge

o.ni

hs.g

o.jp

/csn

db

Dat

abas

em

odel

sco

nsis

tof

cell

sign

alin

g,p

aper

,an

dsi

gnal

mol

ecul

e.

Pat

hway

find

er,

find

the

pat

hway

arou

nda

targ

etor

bet

wee

ntw

ota

rget

mol

ecul

es.

Com

pile

sth

ein

form

atio

non

bio

logi

cal

mol

ecul

es,

seq

uenc

es,

stru

ctur

es,

func

tions

,an

db

iolo

gica

lrea

ctio

nsth

attr

ansf

erth

ece

llula

rsi

gnal

sin

hum

ance

lls.

Que

ryb

yke

ywor

d;

bro

wse

by

mol

ecul

es.

Pat

hway

find

er.

Dat

abas

eof

Inte

ract

ing

Pro

tein

s(D

IPan

dLi

veD

IP)

dip

.doe

-mb

i.ucl

a.ed

u/

DIP

cont

ains

dat

aon

pro

tein

s,in

tera

ctio

ns,

and

exp

erim

enta

lmet

hod

s.Th

em

ajor

com

pon

ents

ofLi

veD

IPar

eP

rote

inS

tate

Pag

es,

Pro

tein

Sta

teTr

ansi

tion

Pag

es,

and

Live

Inte

ract

ion

Pag

es.

Pat

hfin

der

and

bat

chse

arch

,fin

dp

ath

conn

ectin

gq

uery

mol

ecul

es,

and

auto

mat

ical

lyd

raw

the

Inte

ract

ion

Map

.M

apex

pre

ssio

nd

ata,

dis

pla

yex

pre

ssio

nd

ata

sche

mat

ical

lyon

the

Inte

ract

ion

Map

.JD

IP,

ast

and

-alo

neJa

vaap

plic

atio

nth

atp

rovi

des

agr

aphi

cal,

bro

wse

r-in

dep

end

ent

inte

rfac

eto

the

DIP

dat

abas

e.

DIP

and

itsex

tens

ions

emp

hasi

zeco

llect

ing

exp

erim

enta

llyd

eter

min

edp

rote

in-p

rote

inin

tera

ctio

nsan

dp

rovi

din

gto

ols

toex

plo

rean

dvi

sual

ize

the

netw

ork

ofin

tera

ctio

ns,

the

sup

por

ting

exp

erim

enta

lev

iden

ce,

and

tool

sto

inte

grat

efu

nctio

nalg

enom

icd

ata

with

pro

tein

-pro

tein

inte

ract

ions

.

Bro

wse

pro

tein

stat

e;se

arch

inte

ract

ions

;P

athf

ind

er;

bat

chse

arch

.

Kyo

toE

ncyc

lop

edia

ofG

enes

and

Gen

omes

(KE

GG

)ke

gg.g

enom

e.ad

.jp/k

egg/

Gen

ean

nota

tion

and

pat

hway

info

rmat

ion.

Ort

holo

ggr

oup

tab

les

cont

ain

info

rmat

ion

abou

tco

nser

ved

sub

pat

hway

s(p

athw

aym

otifs

).

PA

THW

AY

dat

abas

eco

ntai

nsgr

aphi

calr

epre

sent

atio

nsof

cellu

lar

pro

cess

es,

such

asm

etab

olis

m,

mem

bra

netr

ansp

ort,

sign

altr

ansd

uctio

n,an

dce

llcy

cle.

Sea

rch

orb

row

se.

DIP Goes Live

Molecular & Cellular Proteomics 1.2 105

TAB

LEI—

cont

inue

d

Dat

abas

ean

dU

RL

Pro

tein

-Pro

tein

inte

ract

ion-

rela

ted

cont

ents

Uni

que

feat

ures

Com

men

tsD

ata

que

ries

Pro

teom

ew

ww

.pro

teom

e.co

mP

rote

inR

epor

tp

ages

cont

ain

det

aile

dan

nota

tion

onp

rote

ins

incl

udin

gp

rote

in-

pro

tein

inte

ract

ions

.

Ext

ensi

vean

dw

ell-

refe

renc

edin

form

atio

non

pro

tein

sco

llate

dfr

omsc

ient

ific

liter

atur

e.

Ext

ensi

vein

form

atio

non

pro

tein

s.S

earc

hb

yke

ywor

ds,

bro

wse

by

vario

usca

tego

ries.

Sig

nalin

gp

athw

ayd

atab

ase

(SP

AD

)w

ww

.grt

.kyu

shu-

u.ac

.jp

/sp

ad/

Clic

kab

lep

athw

aym

apan

dp

rote

inin

form

atio

np

age.

Bro

wse

by

extr

acel

lula

rsi

gnal

mol

ecul

es;

sign

alin

gp

athw

aym

ap.

SP

AD

isco

mp

iled

tod

escr

ibe

info

rmat

ion

onin

tera

ctio

nb

etw

een

pro

tein

and

pro

tein

and

pro

tein

and

DN

A,

asw

ella

sin

form

atio

non

seq

uenc

esof

DN

Aan

dp

rote

ins.

Bro

wse

by

extr

acel

lula

rsi

gnal

mol

ecul

es.

Sci

ence

’ssi

gnal

tran

sduc

tion

know

led

geen

viro

nmen

t(S

TKE

)st

ke.s

cien

cem

ag.o

rg

Con

nect

ion

map

s,co

nnec

tions

,an

dco

mp

onen

tp

ages

.

Can

onic

alp

athw

ays

doc

umen

tb

asic

pro

per

ties

ofsi

gnal

ing

mod

ules

,an

dsp

ecifi

cp

athw

ays

des

crib

esp

ecifi

cco

mp

onen

tsan

dre

latio

nsin

ap

artic

ular

orga

nism

,tis

sue,

orce

llty

pe.

Em

pha

size

sin

form

atio

nsu

pp

lied

by

auth

oriti

esin

the

field

,p

rovi

des

info

rmat

ion

man

agem

ent

tool

sto

orga

nize

allk

ind

sof

know

led

gere

late

dto

sign

alin

gtr

ansd

uctio

nin

clud

ing

pro

tein

-pro

tein

inte

ract

ions

.

Sea

rch

by

que

ryst

ring;

bro

wse

by

conn

ectio

nsm

ap.

Tran

spat

htr

ansp

ath.

gbf.d

e/in

dex

.ht

ml

Mol

ecul

e,re

actio

n,lit

erat

ure

refe

renc

ep

ages

.B

row

seab

lem

apof

sign

alin

gp

athw

ays.

An

exte

nsio

nm

odul

eto

the

TRA

NS

FAC

dat

ab

ase

ontr

ansc

riptio

nfa

ctor

san

dth

eir

bin

din

gsi

tes.

TRA

NS

PA

THfo

cuse

son

pat

hway

sin

volv

edin

the

regu

latio

nof

tran

scrip

tion

fact

ors

ind

iffer

ent

spec

ies,

mai

nly

hum

an,

mou

se,

and

rat.

Sea

rch

by

que

ryst

ring;

bro

wse

pat

hway

map

;ex

plo

recr

oss-

talk

s.

DIP Goes Live

106 Molecular & Cellular Proteomics 1.2

that documents experimentally determined protein-protein in-teractions (5), has now been expanded to LiveDIP to describebiological interactions in terms of protein states and statetransitions.

MATERIALS AND METHODS

LiveDIP is a relational database consisting of eleven tables. Pro-tein-protein interactions and their related transitions among the pro-tein states are described by the three main tables, the protein statetable, the state transition table, and the live interaction table. Theprotein state table stores the detailed information on protein states. Itincludes description of chemical modification, cellular localization,whether it is part of a complex, the presence of ligands, and whattriggers the protein state. Information on transitions between proteinstates is stored in the state transition table, which stores the keys forthe initial state and the final state, and the resultant changes. The liveinteraction table describes the two protein state transitions of aninteracting pair, factors that affect the interaction, and the articlereporting this interaction. This database schema defines the prereq-uisite in the protein states for a given interaction to occur and the

transitions in the protein states resulted from the interaction.All data in LiveDIP are collated from published literature. Medline

abstracts are rated for information on protein-protein interactions (6).Abstracts with high scores are read by curators, and data are enteredmanually to assure the quality of the data, through HTML forms.

RESULTS

The Scheme of LiveDIP and Its Relationship to DIP

DIP Describes Physical Protein-Protein Interactions—Pro-tein-protein interactions have different attributes. A protein-protein interaction can be described as a physical associationbetween two proteins, as represented by the black lines link-ing a pair of symbols in Fig. 1A. A protein may interact withmore than one protein. For example, protein B interacts withboth protein A and protein C; protein C also interacts withprotein D. These proteins may be different in properties suchas size or domain structure. For example, proteins A, B, andC are single-domain proteins, and protein D has multiple

FIG. 1. Biological (A) versus physical(B) interactions. Shown is a schematicillustration of the concepts of the DIP andLiveDIP databases and their relationshipto layered information about protein-pro-tein interactions. A, in DIP, physical inter-actions are modeled as binary relation-ships between pairs of proteins. B,LiveDIP is an extension of DIP in whichbiological interactions are modeled astransitions between protein states causedby the protein-protein interactions. Thesingle protein entity in DIP corresponds toa collection of states in the protein statespace of LiveDIP. For example, protein Ccan exist in any of the protein states C1,C2, or C3; B can exist in B1, B2, B3, orB4. A given protein state (e.g. A2) inter-acts only with a given state of its interact-ing partner (B4), represented by pairs ofprotein states shaded with the same pat-tern (A2 andB4; B3 and D1). The interac-tion (green arrow) causes a state transition(red arrow) in one or both of the interact-ing pairs (B to B3; C2 to C1), and theresulting new state either gains or loses itsability to interact with its other interactingpartners: protein B in state B3 interactswith protein C in state C2, and protein C instate C1 can no longer interact with pro-tein D. C, the three major types of data inLiveDIP and the interconnections amongthem. LiveDIP presents information aboutprotein states in the Protein State Page,information about the transition betweenprotein states in the Protein State Transi-tion Page, and information about protein-protein interactions in the Live InteractionPage. Each page is shown here enclosedin a rectangular box with the same coloras the corresponding symbols in the B.

DIP Goes Live


domains. These interactions can be termed physical interac-tions, and they may or may not have biological significance.DIP contains data on physical interactions including the iden-tities of the interacting proteins, their domain structures, theranges of amino acids or domains involved, the binding affin-ity, and the experimental techniques used to detect theseinteractions (5). DIP also provides tools to help assess thesignificance of these interactions.

LiveDIP Describes Protein-Protein Interactions in Terms ofProtein States and State Transitions—Each protein can existin different protein states as shown in Fig. 1B. The collectionof all states for a given protein forms its protein state space(protein C can be in any of the protein states C1, C2, or C3; Bcan be in B1, B2, B3, or B4). A protein state is defined by oneor a combination of several attributes including post-transla-tional modification (protein state B3), presence or absence ofligand (protein state D3), oligomeric state (part of a complex,

protein states C3 and D4). Other attributes include cellular lo-calization, alternative splicing, proteolytic form, etc. Table II liststhe attributes currently included in LiveDIP. The state of a pro-tein determines its activity. For example, some proteins areactive when phosphorylated and inactive when dephosphoryl-ated. Post-translational modifications may also stabilize or de-stabilize the protein or target the protein to certain subcellularlocation. One example is Gpa1, the G-� subunit of the receptor-coupled G protein, which is targeted to the plasma membranethrough dual lipid modification (7). At a given time, the pool ofmolecules of a protein inside a cell may exist in one or several ofits protein states depending on the cellular context.

Another type of information related to protein states isprotein three-dimensional structures. Structural information isimportant for understanding the behaviors of different proteinstates; e.g. why one state is active and another is not. Forexample, the phosphorylated state of Erk2, a homologue of

TABLE IIDescription of protein states, state transitions, and protein-protein interactions in LiveDIP

Attributes that define protein states. Characterization of a protein state in LiveDIP includes one or several of these attributes, which describethe difference between the protein state and the base state of the protein.

Type of attribute List of descriptions

Post-translational modification Autophosphorylation, unphosphorylated, phosphorylation, ubiquitination, multiubiquitination,monoubiquitination, N-acetylation, acetylation, farnesylation, O-glycosylation,N-glycosylation, glycosylation, unglycosylated, methylation, carboxymethylation,N-myristoylation, myristoylation, palmitoylation, unknown

Presence of ligand �Ca, �glucose, �GTP, �GDP, �heme, �ATP, �ADPOligomeric state Complex, homodimer, homo-oligomerCellular localization Nucleus, cytosol, ER, Schmoo tip, plasma membrane, budProteolytic form N-terminal cleavage, C-terminal cleavage, cleavageProtein structure Open form, close form, conformation 1, conformation 2

Type of changes and their descriptions associated with protein state transitions as categorized in LiveDIP. These changes describe thedifferences between two protein states.

Type of change List of descriptions

Molecular activity Activation, inhibition, stimulation, suppression, regulationChemical modification Autophosphorylation, unphosphorylated, phosphorylation, ubiquitination, multiubiquitination,

monoubiquitination, N-acetylation, acetylation, farnesylation, O-glycosylation,N-glycosylation, glycosylation, unglycosylated, methylation, carboxymethylation,N-myristoylation, myristoylation, palmitoylation, unknown

Protein stability Stabilization, destabilizationSubcellular localization Localization, transport, translocalizationOligomeric state Complex formation, complex dissociationStructure Conformational changeProteolytic form CleavageOthers Targeting

Example description of biological interactions causing transitions in the protein states as represented in LiveDIP. A protein state transitionmay be characterized by one or several types of changes listed above.

Example interaction Transition in the protein states Description of change

Ste7 phosphorylates and activates Fus3. Unphosphorylated, inactive Fus3 ➔ phosphorylated, active Fus3 Activation, phosphorylationCln1 forms complex with Cdc28. Cln1 ➔ Cln1 in complex with Cdc28 Complex formationMsn5 transports Far1 across the nuclear

membrane.Far1 in the nucleus ➔ Far1 in cytoplasm Transport

Ste4 stimulates activity of Ste5 byinducing conformational change.

Ste5 active, conformation 1 ➔ Ste5 more active, conformation 2:association of NC termini

Conformational change,stimulation

DIP Goes Live


Fus3, causes refolding of the activation lip in the protein andmakes the active site accessible. It also causes conformationchanges in regions outside the activation lip, through whichthe phosphorylation state can be sensed by other proteins (8).This type of information about the relationship between pro-tein structures and protein states, including alignments andbrief annotations on the structures, is stored in LiveDIP. Otherrelated information in LiveDIP includes what triggers a proteinstate and the article reporting this state.

Proteins inside a cell are not static; they undergo statetransitions (green arrow in Fig. 1B) in response to environmen-tal signals or change in cellular context. For example, Ste20 isa kinase of the pheromone response pathway in yeast. Inresponse to pheromone treatment, yeast cells arrest as un-budded, G1 phase cells. Ste20 from cells of pheromone-arrested cultures is post-translationally modified by phospho-rylation. Removal of pheromone results in the appearance ofthe unphosphorylated form within 30 min (9).

Transition between protein states regulates biological pro-cesses including cell signaling. One example is the use ofcellular localization as a common mechanism to regulate ac-tivity of signaling proteins (10). Proper signaling requires co-localization at the right time of the proteins involved in suc-ceeding steps of a signaling pathway. It was reported thatSte5 shuttles through the nucleus (10). In the presence ofpheromone, Ste5 undergoes enhanced export from the nu-cleus and is recruited to plasma membrane by the G� subunitof the G protein receptor and triggers activation of down-stream kinase Fus3. This function of the nucleus to sequesterproteins destined for the plasma membrane may preventactivation of downstream targets in the absence of signal.Thus, cells can regulate biological processes by modifyingprotein states. A protein state transition in LiveDIP is charac-terized by one or several of the changes listed in Table II.

When is a physical interaction (black lines in Fig. 1A) abiological interaction (green arrows in Fig. 1B)? This requiresthe interacting partners to be in certain protein states for theinteraction to occur and that the interaction causes a transi-tion in the state of one or both of the interacting partners asshown schematically in Fig. 1B. The interaction (green arrow)between two proteins can occur only when each protein is inits specific state, represented, respectively, by the two statesshaded with the same pattern (A2 interacts with B4 and B3with D1). Each protein state can participate only in a subset ofall the interactions of the protein. The interaction (green arrow)causes a state transition (red arrow) in one or both membersof the interacting pair. The transition can be formation ordissociation of a protein complex (the complex of proteins Cand D), changing the states of both proteins (C2 into C3, D1into D4). Alternatively, one protein can modify its interactingpartner and convert its partner into a new state by chemicallymodifying it (A phosphorylates B and changes its state fromB4 into B3) or by other possibly unknown modification (Bmodifies C and changes its state from C2 to C1). The protein

in its new state may gain or lose its ability to interact with itsother interacting partners, e.g. A2 interacts with B4 andchanges it into B3, which then modifies C1; B3 modifies C2into C1, which can no longer form a complex with protein D.Through coupling of chemical events and mechanical actions,protein-protein interactions cause transitions in the proteinstates, regulate the function of the proteins, or transmit asignal from one protein to another. For example, pheromonetreatment activates Fus3 by changing it into its phosphoryl-ated state, which subsequently activates transcription factorSte12 and induces the pheromone response. Msg5, on theother hand, can dephosphorylate Fus3 and change it into theinactive state and is suggested to function in attenuation ofthe mating signal (11). A Live Interaction entry in LiveDIPdescribes the state transitions of the interacting partners.Other information includes references, factors that affect theinteraction, the cell stage or subcellular localization whereinthe interaction was detected, and cellular function of theinteraction such as what signaling pathway it belongs to.

In summary, DIP and LiveDIP provide layered descriptionsof protein-protein interactions. DIP documents physical inter-actions as binary relationships, and LiveDIP describes biolog-ical interactions as transitions in the protein states of theinteracting pairs caused by the interactions. A single proteinentity (protein C) in DIP may correspond to several proteinstates (protein states C1, C2, or C3) in LiveDIP (Fig. 1A). Eachinteraction in DIP may correspond to several LiveDIP interac-tions, because different aspects of the same binary interac-tion may be reported by different research articles.

Presentation of Interactions in LiveDIPInformation on biological interactions collected by LiveDIP

is presented by three major web pages, the Protein StatePages, the Protein State Transition Pages, and the Live Inter-action Pages, as shown in Fig. 1B. The Protein State Pagedescribes the different attributes that define a protein state,including chemical modification, presence/absence of ligand,and subcellular location, etc. Other information includes re-lated three-dimensional structures and factors triggering theprotein state. A link to the reference in PubMed is provided.Each Protein State Transition Page offers a short descriptionof the beginning and ending protein states and the resultantchanges. All protein states in the state transition are linked totheir related Protein State Pages, respectively. The Live Inter-action Page uses schematic drawings and tables to describethe interaction. The Live Interaction Page is linked to therelated Protein State Transition Pages, Protein State Pages,and references in PubMed. Links are also provided to a re-lated entry in DIP, which presents the experimental evidence,interacting domains, and binding affinity. Our aim is to presentinformation about biologically important protein-protein inter-actions in a form familiar to biologists.

DIP Goes Live


Database StatisticsThe September 2001 release of LiveDIP contains 304

unique proteins, 35 types of chemical modifications, and 408interactions. All these data were collected from 341 papers.

Data QueriesUsers can enter the database by browsing protein states by

the types of post-translational modifications or ligands bound.Alternatively, one may query the protein states with a proteinname (Fig. 2). The Search Results page lists all the proteinstates found in a table, each row displaying a short descrip-tion of one protein state. The View Live Interaction buttonleads to the Interaction Map. It shows interactions related tothe selected protein state, with rectangular boxes represent-ing protein states, cyan lines ending on circles representinginhibition, and magenta lines with arrows representing activa-tion. All the entries in these pages are linked to the relevantProtein State Pages, State Transition pages, and Live Inter-action Pages as denoted by the boxes following the colorscheme of Fig. 1. Advanced search tools have been devel-oped to facilitate pathway analysis and pathway discovery(Fig. 3). The Pathfinder feature searches for paths going frommolecule A to molecule B via molecule C with molecules A, B,and C supplied by the user. Users can leave one or two ofthese molecules unspecified. For example, a query with mol-ecule A unspecified will search for all the paths going from anymolecule through molecule B and ending in molecule C. A query

where only molecule B is specified will find all the paths goingthrough molecule B. Also available is the Batch Search featurethat searches for all interactions connecting a group of proteinswithin a given number of steps. Results can be explored in thefollowing several ways as shown in the Result Display Menu: alist of all interactions with links to Live Interaction Pages, a list ofall the linear paths, or an Interaction Map automatically drawnand schematically showing all the interactions. Tools are alsoprovided to integrate mRNA gene expression data with protein-protein interaction networks. Users can choose to map datafrom a single experiment onto the Interaction Map found byadvanced search tools or onto a specific path extracted fromthe search results (see example below). Data from a set ofexpression experiments such as time course can also beplotted as a line chart for each component of a selected pathto analyze temporal regulation of gene expression.

Example Applications

Pathways Function within the Network—The advancedsearch features of LiveDIP can help researchers analyze sig-naling pathways in the context of protein interaction networksas illustrated by the example shown in Fig. 4. Using proteinsinvolved in the pheromone signaling transduction pathway asquery molecules (Ste2, Ste18, Ste4, Ste20, Ste5, Ste11, Ste7,Fus3, Dig1, Dig2, Ste12), results of the Batch Search for

FIG. 2. Basic Search and Browse inLiveDIP. Users can find informationabout protein-protein interactions eitherby browsing the protein states (BrowseMenu) or by performing a query on pro-tein states with a protein name (SearchMenu). The Search Results page lists allthe protein states found for the queryprotein in a table, each row providing ashort description of one protein state.There are also buttons linking to relevantinteracting proteins and protein statetransitions. The schematic InteractionMap shows all proteins interacting with aselected protein state, with rectangularboxes representing protein states, cyanlines ending on circles representing inhi-bition, and magenta lines with arrowsrepresenting activation. All the entries inthese pages are linked to the relevantProtein State pages, Protein State Tran-sition pages, and Live Interaction pagesas denoted by the boxes following thecolor scheme of Fig. 1.

DIP Goes Live


interactions within two steps away from the query moleculesare presented by the Interaction Map of Fig. 4. The smallrectangular boxes represent proteins and complexes. Thelines connecting boxes represent distinct kinds of interac-tions; arrowheads indicate activating interactions, and emptycircles indicate inhibiting interactions. The pheromone re-sponse pathway is enclosed in a shaded rectangular box. Oneobservation from this map is that the pheromone signalingpathway, like so many other pathways in higher organisms,does not work in isolation (12). Seven of the eleven compo-nents of the signaling transduction pathway are regulated byother proteins in the current version of LiveDIP, with oneinteraction each for Dig1 and Dig2 and up to five interactionsfor Ste20. These interactions change the states of the signal-ing proteins mainly through phosphorylation or dephospho-rylation and modulate their activity, and consequently, inten-

sity of the signal. Phosphatase Msg5, for example, causestransition in the protein state of Fus3 by dephosphorylatingand inactivating it and is suggested to be involved in adapta-tion and recovery of cells from pheromone treatment (11).

The Interaction Map also provides visualization of thecross-talk between different pathways. Signaling pathwaysrarely operate in isolation; instead the action of an individualsignaling pathway often affects and/or is integrated with otherpathways to control cell response to outside stimuli. One suchexample is the cross-talk between the pheromone responsepathway and the filamentous growth pathway. Previous stud-ies showed that low concentrations of mating pheromonesincreases agar-invasive growth of haploid yeast cells and thatfilamentation reporter genes are activated in fus3 mutant hap-loid cells (13). The molecular basis of this cross-talk is visu-alized in the Interaction Map in Fig. 4 where MAPKK Ste7

FIG. 3. Tools for analyzing pathwaysby integrating microarray data withprotein-protein interactions. Two Ad-vanced Search features are provided.The Pathfinder feature searches for pro-tein interactions going from molecule Ato molecule B via molecule C (top leftbox), and the Batch Search feature looksfor interactions that connect a group ofproteins within a given number of steps(top right box). Users can choose differ-ent ways to explore the interactionsfound through the Result Display Menu(center box, middle row). Results can berepresented by listing all the linear paths(middle right box) or by a schematic, au-tomatically drawn Interaction Map, withrectangular boxes representing proteinsand lines between them representing in-teractions (middle left box). The bottomof the figure illustrates how mRNA ex-pression data of a selected experimentcan be displayed in the Interaction Map.Boxes representing the correspondingproteins are colored to indicate changesin mRNA expression level, red for in-crease and green for decrease. Datafrom a set of experiments, such as timecourse data, can also be plotted for eachcomponent of a selected linear path.

DIP Goes Live


activates both Fus3 in the pheromone response pathway andKss1 in the filamentous growth pathway. During pheromoneresponse, scaffold protein Ste5 is required for activation ofSte7 by Ste11 and activation of Fus3 by Ste7; interactionsamong Ste11, Ste7, and Fus3 require these proteins to be inthe protein states where they are part of a complex with Ste5;activation of Kss1, on the other hand, does not require Ste5(14). Changes in the protein states of Ste11, Ste7, and Fus3,because of either low dose of pheromone or by mutation ofFus3, may cause misactivation of Kss1 and the cross-talk.LiveDIP provides a tool to map an individual pathway in theprotein-protein interaction network. This visualization helpsunderstanding of how signals are modulated by regulating theprotein states of the pathway components.

Understanding Transcriptional Regulation of the SignalingNetwork—LiveDIP also provides tools to integrate gene expres-sion data with interaction networks as an aid to understandinghow a signaling pathway is regulated to generate the properamount of signal intensity at the right time in response to anoutside stimulus. As an example, we analyzed data on the timecourse of gene expression upon pheromone treatment; thesedata represent changes in mRNA levels between wild-type cellstreated with � factor at different time points from 0 to 120 minversus untreated cells (13). The number of genes induced sig-nificantly is about 30 at 0 min, going up to 112 after 30 min,remaining approximately constant for a while, and then jumpingup to 658 after 120 min. Fig. 5 displays the changes in mRNAexpression level of the proteins in the Interaction Map generatedin Example I (confidence level higher than 90%, with p value lessthan 0.1) at three representative time points (initial, 0 min; mid-point, 45 min; and late point, 120 min) after pheromone treat-ment. The proteins in the interaction map are colored accordingto changes in their expression level, with red representing in-crease and green representing decrease as shown by the colorscale. Uncolored proteins do not have high confidence dataavailable (p � 0.1).

There is no apparent correlation among changes in tran-scription levels of proteins involved in pheromone signaling.Although participating in the same signaling process, compo-nents of the pheromone pathway (proteins enclosed in thecyan box in Fig. 5) exhibit different time course profiles. Someshow increases in their mRNA levels immediately (Ste12,Fus3) or 30 min after (Ste4, Ste2, Gpa1); half of them do notshow significant change throughout the time course. The onlyevident observation is the lack of down-regulation. This seemsto be in contrast with the consistent up-regulation of the glycerolbiosynthetic pathway during salinity stress, which was morestrong as time progressed (15). Assuming a general correlationbetween mRNA level and protein level, the above analysis sug-gests that cells use different mechanisms to regulate metabolicpathways and signaling pathways. Metabolic pathways are reg-ulated more on the transcription level, and signaling pathwaysare regulated by changing the signaling proteins from one stateto another rather than by changing the amounts of these pro-teins. All these observations point to the importance of futureproteomic studies for understanding cell signaling.

Roberts et al. (13) showed that the entire transcriptionalresponse to pheromone is derived from pathway-dependentactivation of the transcription factor Ste12. In LiveDIP, thereare 31 proteins in the protein network around the pheromonesignaling pathway that affect activity of Ste12, directly orindirectly through protein-protein interactions, 19 by activa-tion and 12 by inhibition. More interactions may be discoveredas the database grows. Among the regulators that have highquality expression data during the time course, most activa-tors (7 of 9) reach their maximum change before 60 min whereasall (7) inhibitors reach their maximum change after 45 min. Theactivator Fus3 shows the largest increase of about 12-fold after45 min. The increased mRNA level of these activators mayfunction as positive feedback, leading to stronger signals at theearly stage of signaling. Dig2, which inhibits transcription factorSte12, shows an increase of 2.5-fold only after 90 min. Because

FIG. 4. Pathway analysis using theAdvance Search feature of LiveDIP,applied to the pheromone signalingpathway of yeast. The interactions weregenerated by the Batch Search featureof LiveDIP, using proteins known to beinvolved in pheromone signaling trans-duction (enclosed in the shaded rectan-gular box) as query molecules (Ste2,Ste18, Ste4, Ste20, Ste5, Ste11, Ste7,Fus3, Dig1, Dig2, and Ste12). The resultsare represented by the automaticallydrawn, schematic Interaction Map. Thesmall rectangle boxes depict proteins andcomplexes, and the lines connecting pairsof boxed proteins represent differentkinds of interactions; arrowheads indicateactivating interactions, and empty circlesindicate inhibiting interactions.

DIP Goes Live


FIG. 5. Pathway discovery by com-bining interaction and microarraydata. Mapping changes in mRNA ex-pression level (13) onto the interactionnetwork at three representative timepoints after pheromone treatment (initial,0 min; middle, 45 min; late, 120 min) areshown. The interaction network is thesame as that in Fig. 4. The gene expres-sion data represent changes in mRNAlevel of cells treated with � factor versusuntreated cells at various time pointsfrom 0 to 120 min (13). Changes inmRNA expression level with confidencelevels higher than 90% (p value less than0.1) are represented by red for increaseand green for decrease. Proteins that arenot colored correspond to those with ex-pression data of lower quality (p � 0.1).

DIP Goes Live


prolonged activation of a signaling pathway may lead to desen-sitization, Dig2 may provide negative feedback to down-regu-late the signal. These observations may indicate that activatorsof the signal are induced in the early stage to maximize theintensity whereas inhibitors are generally induced later todampen the signal and promote recovery.

Integrating Yeast Two Hybrid Data with Functional GenomicData for Pathway Discovery—It is possible to integrate mRNAexpression with large scale protein-protein interaction data.Among proteins shown to interact with one or more compo-nents of the yeast pheromone signaling pathway in the largescale yeast two-hybrid experiments, SphI interacts with Ste7and Ste11, and Spa2 interacts with Ste7, Ste11, and Bni1.Both show more than 3-fold decreases in their mRNA levelduring the time course (13), suggesting these interactions maybe relevant to cell response to pheromone treatment, consist-ent with previous observations that both proteins are involvedin schmoo formation (16). Integration of mRNA expressiondata with large scale protein-protein interaction data is oneway to validate yeast two-hybrid data and discover new pro-teins that are involved in signaling processes. It is also pos-sible to use other data, such as protein expression levels andpredicted phosphorylation sites, to help elucidate the rele-vance of protein-protein interactions.

DISCUSSION

Protein-Protein Interactions and Biological PathwaysAs illustrated by our analysis of the pheromone response

pathway in yeast, biological pathways are not isolated; theyexist in the context of complex protein-protein interactionnetworks. Consequently, LiveDIP does not feature staticpathways. Furthermore, our knowledge of biological path-ways is growing rapidly. In many cases it is still not wellunderstood whether and/or when a particular interaction ispart of a given pathway. Instead of storing static pathways,we provide query tools such as Pathfinder and Batch Searchto assemble pathways from currently available knowledge ofprotein interactions in LiveDIP. With sufficient annotation,such as when and where an interaction occurs, and by ap-plying different types of filters, we may be able to reconstructwhich subset of interactions take place inside the cell underspecific conditions. The goal of LiveDIP is to reflect objec-tively what has been reported on the subject of a particularsignaling pathway in the scientific literature.

Application of DIP and LiveDIPInformation about protein-protein interactions provided by

DIP and LiveDIP can be used in three ways by biologists.Based on the functions of its interacting partners, possiblegeneral functions can be assigned to an unannotated protein,or new functions can be discovered for a previously charac-terized protein (17). The large amounts of binary interactionsgenerated by two-hybrid studies are useful for this purpose.On the other hand, annotations on protein-protein interac-

tions, as provided by LiveDIP, including the effect of interac-tions on protein states and state transitions, provide the mo-lecular details on how these proteins carry out their functions.This type of information is important for understanding themechanisms of biological processes and how misregulationleads to disease. Lastly, easy access to large quantities ofexperimental observations of protein-protein interactions fa-cilitates computational manipulation for validating and as-sessing the quality of these data. It provides benchmarks fordeveloping computational methods to predict protein func-tions or protein-protein interactions.

From DIP to LiveDIP: from Physical Interactions toBiological Interactions

LiveDIP contains the subset of interactions in DIP aboutwhich we have detailed information on related protein states.These interactions in LiveDIP (green arrows in Fig. 1B) can betermed biological interactions to distinguish them from phys-ical interactions (black lines in Fig. 1A) in DIP. Some physicalinteractions are not biological interactions, because the inter-acting proteins are not in the right states. For example Bni1regulates polarized growth within the bud by perhaps estab-lishing an anchoring site for Kar9p-Bim1p complex, which isinvolved in capturing microtubules at the bud cortex (16). Thisinteraction can contribute to bud formation only when bothproteins are in the protein states defined by localization to thebud. Other physical interactions do not belong to biologicalinteractions, because they are nonspecific interactions and donot change the protein states. Data from large scale two-hybrid experiments are physical interactions; their biologicalrelevance remains to be tested.

All the biological protein-protein interactions in a given or-ganism may be considered to form the universe of protein-protein interactions as shown in Fig. 6A. The actual boundaryof the universe of interactions is unknown and is representedby the oval in a dashed line. Because of false positive inter-actions, some fraction of DIP and LiveDIP may lie outside ofthe universe of protein interactions. Advances in large scaleexperimental techniques, such as yeast two-hybrid assay andmass spectrometry, lead to accumulation of data in DIP.Another source for protein-protein interactions is throughcomputational methods, and these inferred interactions in-clude both physical interactions and functional interactionssuch as participation in the same pathway (18). It is not clearhow much these interactions from different sources overlapwith the interaction universe.

Many new interactions detected by large scale methods arephysical interactions between two protein entities with un-known biological relevance. They belong only to DIP and notLiveDIP. It is clear that to further our understanding aboutprotein-protein interactions and thus to increase the overlapbetween DIP and the biological interaction universe, and theoverlap between LiveDIP and the interaction universe, we arefaced with the following two tasks: 1) To evaluate the quality

DIP Goes Live


of the large scale interaction data in DIP. It is important tocompare and/or integrate interaction data from differentsources to estimate the confidence for all the interactingprotein pairs. DIP annotates each interaction with all the ex-periments used and can be used as a standard benchmark forlarge scale interaction maps. 2) To understand the functionsand dynamics of these interactions. Information about pro-tein-protein interactions, such as the protein states requiredfor the interactions to occur and the transition in the proteinstates of the interacting pairs caused by the interactions, mustbe specified, such as in LiveDIP. Functional genomics data,such as mRNA expression array and computational methods,can be used to supplement large scale interaction data and tohelp elucidate their biological meaning (Fig. 6B).

Unique Features of LiveDIPLiveDIP has several unique features. It aims to store exper-

imentally determined biological protein-protein interactionsinstead of all kinds of biological relations. The design of Live-DIP captures the essence of biological interactions, their pre-requisite on the states of the interacting partners and theresultant transition in the states of these proteins. One of themajor goals of LiveDIP is to provide the biological community

with convenient access to information about protein-proteininteractions. We emphasize providing easy data accessthrough such tools as Browse and Basic Search and onpresenting the information in a format and language familiar tobiologists. The Batch Search and Pathfinder features can beused to assemble biological pathways based on knowledgeabout protein-protein interactions on the fly, and the resultscan be explored in various ways including automaticallydrawn schematic Interaction Maps. We provide tools to inte-grate large scale functional genomics data with protein-pro-tein interaction networks. These features of LiveDIP render ita useful tool for biological pathway analysis and pathwaydiscovery.

Acknowledgments—We thank Lukasz Salwinski, Parag Mallick,and Michael Thompson for suggestions.

* This work was supported by Department of Energy and NationalInstitutes of Health. The costs of publication of this article weredefrayed in part by the payment of page charges. This article musttherefore be hereby marked “advertisement” in accordance with 18U.S.C. Section 1734 solely to indicate this fact.

‡ To whom correspondence should be addressed: HowardHughes Medical Inst., UCLA-DOE Laboratory of Structural Biologyand Molecular Medicine, University of California, Los Angeles, P. O.Box 951570, Los Angeles, CA 90095-1570. Tel.: 310-825-3754; Fax:310-206-3914; E-mail: [email protected].

REFERENCES

1. Ideker, T., Thorsson, V., Ranish, J. A., Christmas, R., Buhler, J., Eng, J. K.,Bumgarner, R., Goodlett, D. R., Aebersold, R., and Hood, L. (2001)Integrated genomic and proteomic analyses of a systematically per-turbed metabolic network. Science 292, 929–934

2. Eisenberg, D., Marcotte, E. M., Xenarios, I., and Yeates, T. O. (2000) Proteinfunction in the post-genomic era. Nature 405, 823–826

3. Bader, G. D., and Hogue, C. W. (2000) BIND-a data specification for storingand describing biomolecular interactions, molecular complexes andpathways. Bioinformatics 16, 465–477

4. van Helden, J., Naim, A., Mancuso, R., Eldridge, M., Wernisch, L., Gilbert,D., and Wodak, S. J. (2000) Representing and analysing molecular andcellular function using the computer. Biol. Chem. 381, 921–935

5. Xenarios, I., Fernandez, E., Salwinski, L., Duan, X. J., Thompson, M. J.,Marcotte, E. M., and Eisenberg, D. (2001) DIP: the database of interact-ing proteins: 2001 update. Nucleic Acids Res. 29, 239–241

6. Marcotte, E. M., Xenarios, I., and Eisenberg, D. (2001) Mining literature forprotein-protein interactions. Bioinformatics 17, 359–363

7. Manahan, C. L., Patnana, M., Blumer, K. J., and Linder, M. E. (2000) Duallipid modification motifs in G(alpha) and G(gamma) subunits are requiredfor full activity of the pheromone response pathway in Saccharomycescerevisiae. Mol. Biol. Cell 11, 957–968

8. Canagarajah, B. J., Khokhlatchev, A., Cobb, M. H., and Goldsmith, E. J.(1997) Activation mechanism of the MAP kinase ERK2 by dual phospho-rylation. Cell 90, 859–869

9. Wu, C., Leeuw, T., Leberer, E., Thomas, D. Y., and Whiteway, M. (1998) Cellcycle- and Cln 2p-Cdc28p-dependent phosphorylation of the yeastSte20p protein kinase. J. Biol. Chem. 273, 28107–28115

10. Elion, E. A. (2000) Pheromone response, mating and cell biology. Curr.Opin. Microbiol. 3, 573–581

11. Zhan, X. L., Deschenes, R. J., and Guan, K. L. (1997) Differential regulationof FUS3 MAP kinase by tyrosine-specific phosphatases PT, 2/PTP3 anddual-specificity phosphatase MSG5 in Saccharomyces cerevisiae.Genes Dev. 11, 1690–1702

12. Chambers, D. (2001) Scientists signal the way forward. Trends Genet. 17,309–310

13. Roberts, C. J., Nelson, B., Marton, M. J., Stoughton, R., Meyer, M. R.,Bennett, H. A., He, Y. D., Dai, H., Walker, W. L., Hughes, T. R., Tyers, M.,

FIG. 6. Relationship of DIP and LiveDIP to the universe of pro-tein-protein interactions. A, all protein-protein interactions impor-tant for biological processes in a given organism form the universe ofprotein-protein interactions. The actual boundary of the universe ofprotein-protein interactions is unknown and is represented by thearea enclosed by the oval in a dashed line. Advances in large scaleexperimental techniques such as the yeast two-hybrid assay (19) andmass spectrometry (20) lead to the accumulation of protein-proteininteraction data in DIP. LiveDIP contains the subset of protein-proteininteractions in DIP, about which we have detailed information relatedto the protein states of the interacting pairs. Because of false positiveinteractions, some fraction of DIP and LiveDIP may lie outside of theuniverse of protein interactions. Computationally inferred interactionsinclude both physical interactions and functional interactions such asparticipation in the same pathway (18). B, understanding protein-protein interactions. To understand the network of protein interac-tions documented in DIP requires us to make these interactions goLive; that is, the details of protein states and state transitions andother biologically relevant information need to be specified as inLiveDIP. Deepened understanding can come by supplementing theinteraction data in DIP with data from functional genomics and com-putational methods.

DIP Goes Live


Boone, C., and Friend, S. H. (2000) Signaling and circuitry of multipleMAPK pathways revealed by a matrix of global gene expression profiles.Science 287, 873–880

14. Gustin, M. C., Albertyn, J., Alexander, M., and Davenport, K. (1998) MAPkinase pathways in the yeast Saccharomyces cerevisiae. Microbiol. Mol.Biol. Rev. 62, 1264–1300

15. Posas, F., Chambers, J. R., Heyman, J. A., Hoeffler, J. P., de Nadal, E., andArino, J. (2000) The transcriptional response of yeast to saline stress.J. Biol. Chem. 275, 17249–17255

16. Arkowitz, R. A., and Lowe, N. (1997) A small conserved domain in the yeastSpa 2p is necessary and sufficient for its polarized localization. J. Cell

Biol. 138, 17–3617. Schwikowski, B., Uetz, P., and Fields, S. (2000) A network of protein-

protein interactions in yeast. Nat Biotechnol. 18, 1257–126118. Marcotte, E. M., Pellegrini, M., Thompson, M. J., Yeates, T. O., and

Eisenberg, D. (1999) A combined algorithm for genome-wide predictionof protein function. Nature 402, 83–86

19. Ito, T., Chiba, T., Ozawa, R., Yoshida, M., Hattori, M., and Sakaki, Y. (2001)A comprehensive two-hybrid analysis to explore the yeast protein inter-actome. Proc. Natl. Acad. Sci. U. S. A. 98, 4569–4574

20. Gygi, S. P., Rist, B., and Aebersold, R. (2000) Measuring gene expressionby quantitative proteome analysis. Curr. Opin. Biotechnol. 11, 396–401

DIP Goes Live


describing biological protein interactions in terms of protein … · 2008-08-07 · prebind, a...

Documents