This is an elec onic ep in o he o iginal a icle.
This ep in may di e om he o iginal in pagina ion and ypog aphic de ail.
Au ho (s):
Ti le:
Yea :
Ve sion:
Please ci e he o iginal e sion:
All ma e ial supplied ia JYX is p o ec ed by copy igh and o he in ellec ual p ope y igh s, and
duplica ion o sale o all o pa o any o he eposi o y collec ions is no pe mi ed, excep ha
ma e ial may be duplica ed by you o you esea ch use o educa ional pu poses in elec onic o
p in o m. You mus ob ain pe mission o any o he use. Elec onic o p in copies may no be
o e ed, whe he o sale o o he wise o anyone who is no an au ho ised use .
Comp ehensi e analysis o o y yeas mic oa ay da ase s e eals a no el subse o
genes (APha-RiB) consis en ly nega i ely associa ed wi h ibosome biogenesis
Abu-Jamous, Basel; Fa, Rui; Robe s, Da id J.; Nandi, Asoke
Abu-Jamous, B., Fa, R., Robe s, D. J., & Nandi, A. (2014). Comp ehensi e analysis o
o y yeas mic oa ay da ase s e eals a no el subse o genes (APha-RiB)
consis en ly nega i ely associa ed wi h ibosome biogenesis. BMC Bioin o ma ics, 15,
A icle 322. h ps://doi.o g/10.1186/1471-2105-15-322
2014
RESEARCH ARTICLE Open Access
Comp ehensi e analysis o o y yeas mic oa ay
da ase s e eals a no el subse o genes (APha-RiB)
consis en ly nega i ely associa ed wi h ibosome
biogenesis
Basel Abu-Jamous
1
, Rui Fa
1
, Da id J Robe s
2,3
and Asoke K Nandi
1,4*
Abs ac
Backg ound: The scale and complexi y o genomic da a lend hemsel es o analysis using sophis ica ed
ma hema ical echniques o yield in o ma ion ha can gene a e new hypo heses and so guide u he expe imen al
in es iga ions. An ensemble clus e ing me hod has he abili y o pe o m consensus clus e ing o e he same se o
genes om di e en mic oa ay da ase s by combining esul s om di e en clus e ing me hods in o a single
consensus esul .
Resul s: In his pape we ha e pe o med comp ehensi e analysis o o y yeas mic oa ay da ase s. One ecen ly
desc ibed Bi-CoPaM me hod can analyse exp essions o he same se o genes om a ious mic oa ay da ase s
while using di e en clus e ing me hods, and hen combine hese esul s in o a single consensus esul whose
clus e s’ igh ness is unable om igh , speci ic clus e s o wide, o e lapping clus e s. This has been adop ed in
a no el way o e genome-wide da a om o y yeas mic oa ay da ase s o disco e wo clus e s o genes ha
a e consis en ly co-exp essed o e all o hese da ase s om di e en biological con ex s and a ious expe imen al
condi ions. Mos s ikingly, a e age exp ession p o iles o hose clus e s a e consis en ly nega i ely co ela ed in all
o he o y da ase s while nei he p o ile leads o lags he o he .
Conclusions: The i s clus e is en iched wi h ibosomal biogenesis genes. The biological p ocesses o mos o he
genes in he second clus e a e ei he unknown o appa en ly un ela ed al hough hey show high connec i i y in
p o ein-p o ein and gene ic in e ac ion ne wo ks. The e o e, i is possible ha his mos ly uncha ac e ised clus e
and he ibosomal biogenesis clus e a e ansc ip ionally opposi ely egula ed by some common machine y.
Mo eo e , we an icipa e ha he genes included in his p e iously unknown clus e pa icipa e in gene ic, in
con as o speci ic, s ess esponse p ocesses. These no el indings illumina e coo dina ed gene exp ession in yeas
and sugges se e al hypo heses o u u e expe imen al unc ional wo k. Addi ionally, we ha e demons a ed he
use ulness o he Bi-CoPaM-based app oach, which may be help ul o he analysis o o he g oups o (mic oa ay)
da ase s om o he species and sys ems o he explo a ion o global gene ic co-exp ession.
Keywo ds: Ribosome biogenesis, S ess esponse, Co-exp ession, Co- egula ion, Genome-wide analysis, Budding
yeas , (Bina isa ion o consensus pa i ion ma ices) Bi-CoPaM
* Co espondence: [email p o ec ed]
1
Depa men o Elec onic and Compu e Enginee ing, B unel Uni e si y,
Uxb idge, Middlesex UB8 3PH, UK
4
Depa men o Ma hema ical In o ma ion Technology, Uni e si y o
Jy äskylä, Jy äskylä, Finland
Full lis o au ho in o ma ion is a ailable a he end o he a icle
© 2014 Abu-Jamous e al.; licensee BioMed Cen al L d. This is an Open Access a icle dis ibu ed unde he e ms o he
C ea i e Commons A ibu ion License (h p://c ea i ecommons.o g/licenses/by/4.0), which pe mi s un es ic ed use,
dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly c edi ed. The C ea i e Commons Public
Domain Dedica ion wai e (h p://c ea i ecommons.o g/publicdomain/ze o/1.0/) applies o he da a made a ailable in his
a icle, unless o he wise s a ed.
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322
h p://www.biomedcen al.com/1471-2105/15/322
Backg ound
Ad ances in mic oa ay echnology ha e enabled
measu emen s o exp ession o a as numbe o
genes simul aneously. Mos mic oa ay expe imen s
conside measu ing he exp ession alues o he en i e
genome o a speci ic o ganism o e mul iple ime-poin s,
se e al biological de elopmen al s ages, di e en ypes o
issues, o di e en condi ions [1]. Many di e en me hods
o mic oa ay analysis ha e been designed and applied in
o de o add ess such di e se ques ions. Some me hods
aim o iden i y genes ha a e di e en ially exp essed
be ween ce ain pheno ypes o condi ions, which would be
hen p edic ed o pa icipa e in causing such pheno ypes o
in he esponse o such condi ions [1-3]. O he me hods
ha e been p oposed o look o , and model he exp ession
o , genes ha ha e co-o dina ed exp ession o e cell o
me abolic cycles [4-7]. Mo eo e , a ious supe ised and
unsupe ised me hods ha e been designed o answe
ques ions ela ed o he co-exp ession o genes [8-12].
One class o supe ised me hods, which sea ch o
co-exp essed genes, is empla e-based mining. He e, he
mic oa ay da ase is mined o genes whose exp ession
p o iles a e simila (based on a simila i y c i e ion, e.g.
Euclidean dis ance) o an a p io i known empla e o
exp ession. Fo example, Nilsson and colleagues sea ched
in a la ge numbe o blood- ela ed human and mice
mic oa ay da ase s o genes ha a e consis en ly
co-exp essed wi h he a e age exp ession p o ile o eigh
well-known genes ha pa icipa e in haem biosyn hesis
[10]. Simila ly, Wade and colleagues mined ou budding
yeas da ase s o genes ha a e consis en ly co-exp essed
wi h he a e age exp ession p o ile o 65 p e iously
epo ed ibosomal biogenesis genes [9]. Al hough
hese empla e based me hods can con i m he consis ency
o co-exp ession o he genes ma ching he que y empla e
in mul iple da ase s, hey canno de e mine i he e a e any
o he clus e s o genes ha consis en ly ma ch di e en
empla es o exp ession.
Amongs he classes o unsupe ised me hods ha
mine o co-exp essed genes, gene clus e ing is he mos
commonly used. The objec i e o any o he a ious
me hods belonging o his class is o g oup genes in o
clus e s such ha genes included in a clus e a e simila o
each o he while being dissimila om he genes included
in he o he clus e s based on a speci ic c i e ion o
simila i y [2]. In his way, genes a e g ouped in o
subse s o co-exp essed genes. Examples o me hods
used o gene clus e ing a e k-means [11], hie a chical
clus e ing (HC) [8], sel -o ganizing maps (SOMs) [13,14]
and sel -o ganizing oscilla o ne wo ks (SOON) [15], as
well as ensemble me hods, e.g. elabeling and o ing [16],
co-associa ion ma ix [17], hype g aph me hods [18], and
he ecen ly p oposed bina isa ion o consensus pa i ion
ma ices (Bi-CoPaM) [19-21].
A majo d awback o mos clus e ing me hods is ha hey
impose he cons ain ha each gene mus be exclusi ely
assigned o one and only one clus e . Thus, eeding genome
wide da a o such clus e ing me hods always p oduces
clus e s ha include all o he genes in his genome;
he e o e, he size and complexi y o he da a a e no
dec eased signi ican ly. We ha e ackled his p oblem
by ou ecen ly published uncon en ional ensemble
clus e ing me hod (Bi-CoPaM), which p o ides a pla o m
ha allows o gene a ing con en ional complemen a y
clus e s in which each gene is exclusi ely assigned o a
single clus e , as well as uncon en ional clus e s such
as wide o e lapping clus e s in which genes can be
simul aneously assigned o mul iple clus e s, and igh
clus e s which lea e many genes unassigned o any clus e
[19,20]. P oducing such a ying o ms o uncon en ional
clus e s allows uning, such ha di e en gene disco e y
s udies can une he Bi-CoPaM o p oduce he pa icula
o m o clus e s ha helps in answe ing ha s udy’s
speci ic ques ions. Mo eo e , he uneable pa i ions
p oduced by he Bi-CoPaM a e based on he consis ency o
co-exp ession o a se o genes ac oss mul iple mic oa ay
da ase s and when clus e ed by a ious clus e ing me hods
[19,20]. The Bi-CoPaM me hod does no combine he
p o iles o he genes in mul iple da ase s in o de o
analyse hem collec i ely. I a he achie es his collec i e
analysis by examining each da ase independen ly and
hen combining hei esul s in o a single consensus
esul [19,20].
Wade and colleagues iden i ied a subse o genes
consis en ly co-exp essed wi h a empla e o 65 ibosomal
biogenesis genes in ou di e en da ase s [9]. Tha subse
was ound o be en iched wi h RNA p ocessing and
ibosomal biogenesis genes (RRB), and was ound o
be up- egula ed when eleased om cell-cycle a es
while being down- egula ed unde s ess [9]. O he
s udies ha e iden i ied RRB-en iched subse s o genes
wi h p o iles ha a e consis en ly posi i ely co ela ed
wi h g ow h and nega i ely co ela ed wi h s ess [22,23].
On he o he hand, o he subse s o genes, mainly en iched
wi h s ess esponse genes, we e iden i ied as nega i ely
co ela ed wi h g ow h and posi i ely co ela ed wi h s ess
[22,23]. The egula ion o such subse s o da ase s has been
discussed by a ious s udies which lis ed di e en con-
i med o po en ial egula o s such as Tod6p, S b3p, and
S p1p o RRB genes, and Msn2/4p, Rg 1p, and Ad 1p o
s ess esponse genes [9,22-25]. Also, he ela ions be ween
g ow h a e and s ess esis ance, as well as be ween he
exp ession o RRB genes and o he egulons such as
ibosomal p o eins and cell cycle genes we e discussed
while conside ing signal ansduc ion pa hways (e.g. TOR1
and Ras/PKA pa hways) o ansc ip ion ac o s as
egula o y connec ions [9,23,24]. Each o hose s udies
conside ed one o ew da ase s o ob ain i s conclusions.
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 2 o 20
h p://www.biomedcen al.com/1471-2105/15/322
In his s udy, we adop a no el app oach o he Bi-CoPaM
me hod o analyse genome-wide da a om o y mic oa ay
yeas da ase s om a wide ange o biological condi ions
and con ex s in o de o iden i y he subse s o genes ha
a e consis en ly co-exp essed in Saccha omyces ce e isiae
budding yeas unde such a ious condi ions [19,20]. We
in es iga e i he RRB genes a e consis en ly co-exp essed in
a wide ange o condi ions han hose in es iga ed by
p e ious s udies [9,22,23]. Mo eo e , we explo e i he e a e
o he no el subse s o budding yeas genes ha a e consis -
en ly co-exp essed o e such wide ange o di e en mic o-
a ay da ase s and, i so, we in es iga e hei p e ious
cha ac e isa ions and known unc ion(s), and we d aw
hypo heses ega ding hei egula ion as well as he po en-
ial oles o hei poo ly unde s ood genes in cell biology.
Me hods
Bi-CoPaM
The Bi-CoPaM me hod consis s o ou main s eps
(Figu e 1) [19,20]:
1. Gene a ion o many pa i ions o he same se o
genes by applying a ious clus e ing me hods o e
he exp ession p o iles o hese genes om mul iple
mic oa ay da ase s.
2. Relabelling he gene a ed pa i ions such ha each
clus e om one pa i ion is ma ched wi h i s
co esponding clus e om e e y o he pa i ion.
3. Gene a ion o he uzzy consensus pa i ion ma ix
(CoPaM) by elemen -by-elemen a e aging o he
elabelled pa i ions.
4. Bina iza ion o he CoPaM by one o mo e o he six
unable bina iza ion echniques p oposed in [19].
To ampli y he a ia ion in clus e assignmen caused
by he di e ences in mic oa ay da ase s o e he one
caused by he di e ences amongs clus e ing me hods,
he pa i ions gene a ed by applying di e en clus e ing
me hods o e any single mic oa ay da ase a e i s
combined in o a single in e media e uzzy consensus
pa i ion ma ix (CoPaM) whose membe ship alues a e
p ocessed by pushing hem owa ds he bina y alues
o ze o and one (Figu e 1); his is ma hema ically
o mula ed as

ui;j¼
mjþmjsin πui;j−mj

2mj

;ui;j≤mj
mjþ1−mj

sin πui;j−mj

21−mj

!
;ui;j>mj
;
8
>
>
>
<
>
>
>
:
whe e u
i,j
and 
ui;ja e he uzzy membe ship alues o
he j
h
gene in he i
h
clus e be o e and a e p ocessing,
espec i ely, and m
j
is mean o he uzzy membe ship
alues o he j
h
gene in all o he clus e s in which i has
non-ze o alues. A e all CoPaM ma ices om all o
he mic oa ay da ase s a e gene a ed and p ocessed as
desc ibed he ein, hey a e combined o p oduce he inal
CoPaM, which is hen bina ised o p oduce he inal
bina y pa i ions.
The six bina iza ion echniques sc u inize he CoPaM in
di e en ways o p oduce bina y pa i ions wi h di e en
ea u es. Ou concen a ion in his s udy is on he di e ence
h eshold bina iza ion (DTB) echnique and i s wo ex eme
special cases maximum alue bina iza ion (MVB) and
in e sec ion bina iza ion (IB).
The MVB echnique assigns each gene o he clus e
in which i has i s maximum uzzy membe ship; his
gene a es con en ional complemen a y clus e s in which
each gene is exclusi ely assigned o one and only one
clus e . The DTB echnique imposes a s ic e policy; i
assigns his gene o ha maximum-membe ship clus e
Pa i ion 1,1
Se o
Genes
Indi idual pa i ions gene a ion Relabelling, CoPaM gene a ion, and
CoPaM p ocessing
Consensus
Fuzzy
CoPaM
Bina iza ion
Consensus
Bina y
Pa i ion
Pa i ion 1,C
Clus e ing
Me hod 1
Clus e ing
Me hod C
Da ase 1
Fuzzy
CoPaM
1
Pa i ion L,1
Pa i ion L,C
Clus e ing
Me hod 1
Clus e ing
Me hod C
Da ase L
Fuzzy
CoPaM
L
P ocessed
Fuzzy
CoPaM
1
P ocessed
Fuzzy
CoPaM
L
In e media e CoPaMs
Figu e 1 The pipeline o s eps in he Bi-CoPaM me hod.
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 3 o 20
h p://www.biomedcen al.com/1471-2105/15/322
only i he closes clus e compe ing on his gene has a
uzzy membe ship alue which is lowe han he maximum
by a leas he alue o he pa ame e (δ). O he wise, he
gene is no conside ed clea ly belonging o a speci ic clus e
and is unassigned om all o he clus e s acco dingly.
When his DTB pa ame e (δ) is ze o, i is equi alen o he
MVB echnique. Tigh e clus e s wi h mo e unassigned
genes a e ob ained when δis inc eased un il i eaches
one. When i s alue is one, only genes ha ha e been
consensually assigned o he same clus e s by all o
he single pa i ions a e p ese ed; all o he o he
genes a e le unassigned. This igh es case is equi alen
o he IB echnique.
Mean Squa ed E o (MSE) me ic
The mean squa ed e o (MSE) me ic has been used by
many s udies o e alua e he quali y o he gene a ed
clus e s so ha compa isons be ween di e en me hods
can be pe o med [26,27]. We adop he MSE me ic o
e alua ing he gene a ed clus e s.
Because he o al numbe o genes assigned o he
clus e s by Bi-CoPaM a any speci ic igh ness le el is
a iable, we use a no malized MSE measu e o be pe
gene.TheMSE
clus e
me ic which quan i ies he o al MSE
o he k
h
clus e is de ined as:
MSEclus e kðÞ¼1
N⋅MkX
xi∈Ck
xi−zk
kk
2;
Whe e Nis he numbe o dimensions ( ime-poin s) in
he da ase , M
k
is he numbe o genes in he k
h
clus e ,
C
k
is he se o ze o-cen ed uni y-s anda d-de ia ion
gene ic exp ession p o iles {x
i
} o he genes in he k
h
clus e , and z
k
is he mean exp ession p o ile o he
genes in he k
h
clus e .
I mul iple da ase s we e used o clus e ing, genes
p o iles and he clus e s cen oids will a y om one
da ase o ano he o he same pa i ion. In his case,
he MSE me ic can be calcula ed mul iple imes o
each da ase and hen a e aged o e hem.
Da ase s & expe imen al p ocedu es
In his s udy, we conside o y ecen Saccha omyces
ce e isiae mic oa ay da ase s which we e gene a ed by
using he A yme ix yeas genome 2.0 a ay in he las
six yea s, and include a leas ou di e en condi ions
o ime-poin s. Al hough choosing da ase s gene a ed by
using he same a ay is no a condi ion o Bi-CoPaM
analysis, i allows o mo e genes o be included in he
analysis as some genes migh no be ep esen ed by
p obes in all ypes o a ays, and he e o e ha e o be
disca ded om he analysis in such a case. Each o hese
da ase s measu es he gene ic exp ession o he en i e
yeas genome (5,667 genes) o e mul iple ime-poin s o
condi ions. The de ails o he da ase s a e lis ed in Table 1.
The da ase s span a wide ange o biological condi ions
such as cell-cycle, s ess esponse, mu a ed s ains g ow h,
ea men wi h a ious ypes o agen s, and o he s. The
5,667 genes a e lis ed in Addi ional ile 1: Table S1.
These 5,667 genes we e clus e ed in o six een clus-
e s by k-means wi h Kau man ini ialisa ion (KA) [48],
sel -o ganising maps (SOMs) wi h bubble neighbou hood
and ou -by- ou g id [13], and hie a chical clus e ing
(HC) wi h Wa d’s linkage [8]. This was applied o hei
p o iles om all o he o y da ase s. The gene a ed
pa i ions we e combined in o a single consensus pa i ion
ma ix (CoPaM) as explained in Bi-CoPam whe e a
min-min app oach was adop ed o elabeling a he
CoPaM gene a ion s ep. The inal CoPaM was bina ised
by he di e ence h eshold bina iza ion (DTB) echnique
wi h δ alues anging om ze o o one and hen analysed
by he MSE me ic desc ibed in Mean Squa ed E o
(MSE) me ic. P io o clus e ing, he da ase s we e no -
malized by quan ile no maliza ion [49]. Then each gene’s
exp ession p o ile was shi ed and scaled o be ze o-mean
and uni y s anda d de ia ion. Also, when many eplica es
exis o he same ime-poin o condi ion, hey a e
summa ised by conside ing hei median alue.
Resul s
The numbe s o genes in he six een clus e s a all o he
a ying δ alues a e shown in Table 2. Clus e s we e
o de ed based on hei igh ness such ha hose clus e s
ha p ese e a leas se en genes up o highe alues o
δa e conside ed igh e . When many clus e s p ese e
a leas se en genes up o he same alue o δ, hey a e
o de ed based on he numbe o genes hey include a
ha le el. The numbe ‘se en’is jus used o o de ing
and is no a c i ical pa ame e ; i i had been se o ‘ en’
ins ead o example, no signi ican change in clus e
o de ing would ha e be obse ed. The comple e lis s o
genes included in each o hese clus e s a all o he δ
alues a e p o ided in Addi ional ile 1: Table S1.
MSE analysis
The MSE alues o each o he igh es six clus e s we e
calcula ed a all o he DTB δ alues as explained in
Mean Squa ed E o (MSE) me ic. Each o hese alues
was calcula ed based on he o y da ase s and hen
a e aged and plo ed in Figu e 2(A). Figu e 2(B)
shows he numbe s o genes included in each o hese
six clus e s a all o he δ alues. Missing poin s in
bo h plo s ep esen emp y clus e s.
We ha e conside ed he mean s anda d e o (MSE)
e alua ion me ic in andem wi h he numbe o genes
included in he clus e s o choose a ew clus e s o u he
analysis and disca d he o he ones. The objec i e he e is
o minimise he MSE alues while maximising he numbe
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 4 o 20
h p://www.biomedcen al.com/1471-2105/15/322

Table 1 Budding yeas mic oa ay da ase s
ID GEO accession Yea N Desc ip ion Re .
D01 GSE8799 2008 15 Two mi o ic cell-cycles (w/ ). [28]
D02 GSE8799 2008 15 Two mi o ic cell-cycles (mu a ed cyclins). [28]
D03 E-MTAB-643* 2011 15 Response o an impulse o glucose. [14]
D04 E-MTAB-643* 2011 15 Response o an impulse o ammonium. [14]
D05 GSE54951 2014 6 Response o dal80Δmu an yeas o oxida i e s ess induced by linoleic acid hyd ope oxide. -
D06 GSE25002 2014 9 Osmo ic s ess esponse and ea men o ans o man s exp essing he C. albicans Nik1 gene. -
D07 GSE36298 2013 6 Mu a ions o OPI1, INO2, and INO4 unde ca bon-limi ed g ow h condi ions. [29]
D08 GSE50728 2013 8 120-hou ime-cou se du ing e men a ion. -
D09 GSE36599 2013 5 S ess adap a ion and eco e y. [30]
D10 GSE47712 2013 6 Combina ions o he yeas media o complex’s ail subuni s mu a ions. [31]
D11 GSE21870 2013 4 Combina ions o mu a ions in DNUP60 and DADA2. -
D12 GSE38848 2013 6 Va ious s ains unde ae obic o anae obic g ow h. [32]
D13 GSE36954 2012 6 Response o myco oxic ype B icho hecenes. [33]
D14 GSE33276 2012 6 Response o hea s ess o h ee di e en s ains. -
D15 GSE40399 2012 7 Response o a ious pe u ba ions (hea , my iocin ea men , and lipid supplemen ). -
D16 GSE31176 2012 6 W/ , lm1Δ, and swi3Δcells wi h o wi hou Congo Red exposu e. [34]
D17 GSE26923 2012 5 Va ying le els o GCN5 F221A mu an exp ession. [35]
D18 GSE30054 2012 31 CEN.PK122 oscilla ing o wo hou s. -
D19 GSE30051 2012 32 CEN.PL113-7D oscilla ing o wo hou s. [36]
D20 GSE30052 2012 49 CEN.PL113-7D oscilla ing o ou hou s. [36]
D21 GSE32974 2012 15 Abou 5 hou s o cell-cycle (w/ ). [37]
D22 GSE32974 2012 15 Abou 4 hou s o cell-cycle (mu an lacking Cdk1 ac i i y). [37]
D23 GSE24888 2011 5 Un ea ed yeas e sus yeas s ea ed wi h E. a ense he bs om he USE, China, Eu ope, o India. -
D24 GSE19302 2011 6 Response o deg on induc ion o w/ and nab2- d mu an . [38]
D25 GSE33427 2011 5 Un ea ed w/ , and w / , yap1Δ,yap8Δ, and double mu an ea ed wi h AsV. [39]
D26 GSE17716 2011 7 E ec o o e exp ession and dele ion o MSS11 and FLO8. [40]
D27 GSE31366 2011 4 P esence and absence o mu li-inhibi o s o pa en al and ole an s ains. -
D28 GSE26171 2011 4 Response o pa ulin and/o asco bic acid. [41]
D29 GSE22270 2011 4 PY1 and Me 30 s ains in oom empe a u e o 35 C. -
D30 GSE29273 2011 4 Time-se ies du ing yeas second e men a ion. -
D31 GSE29353 2011 5 Di e en haploid s ains g owing in low glucose medium. [42]
D32 GSE21571 2011 8 Di e en combina ions o mu a ions in HTZ1, SWR1, SWC2, and SWC5. [43]
D33 GSE17364 2010 4 Un ea ed w/ and Sl 2-de icien yeas s, o ea ed wi h sodium a sena e o wo hou s. [44]
D34 GSE15352 2010 8 24-hou ime-cou se o yeas g own unde a low empe a u e (10 C). [45]
D35 GSE15352 2010 8 24-hou ime-cou se o yeas g own unde a no mal empe a u e (28 C). [45]
D36 GSE15352 2010 8 24-hou ime-cou se o yeas g own unde a high empe a u e (37 C). [45]
D37 GSE16799 2009 21 UC-V i adia ion o w/ , mig3Δ,SNF1Δ,RAD23Δ,RAD4Δ, and sn 1Δ ad23Δ.[46]
D38 GSE16346 2009 4 BY474 cells g own o mid-log unde p esence e sus absence o L-ca ni ine and/o H
2
O
2
.-
D39 GSE14227 2009 10 Two hou s o wild- ype yeas g ow h. [47]
D40 GSE14227 2009 9 Two hou s o sch9Δmu an yeas g ow h. [47]
The i s column shows he unique iden i ie which is used he eina e o e e o each o hese da ase s. The second o he six h columns espec i ely show he
Gene Exp ession Omnibus (GEO) accession numbe , he yea in which he da ase was published, numbe o ime-poin s o condi ions a e eplica e summa isa ion,
da ase desc ip ion, and e e ence.
*D03 and D04 ha e accession numbe s in he Eu opean Bioin o ma ics Ins i u e (EBI) eposi o y a he han GEO accession numbe s.
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 5 o 20
h p://www.biomedcen al.com/1471-2105/15/322
o genes included in he clus e s. This app oach o e comes
he dependency o MSE alues on he numbe s o genes
included in he clus e s. As can be seen in Figu e 2(A) and
(B), he clus e C1 shows signi ican ly lowe (be e ) alues
o MSE while including signi ican ly highe numbe s o
genes. The clus e C2 comes nex o C1 in e ms o ha ing
lowe MSE alues wi h mo e genes.
On he o he hand, while he clus e s C3 and C4 ha e
compa a i e MSE alues a δ= 0.2 wi h C2 (Figu e 2
(A)), hey ha e signi ican ly lowe numbe s o genes
(17 and 14 genes espec i ely o C3 and C4 in compa ison
wi h 47 in C2; see Table 2). Fu he mo e, he clus e s C5
and C6 a e signi ican ly wo se (highe MSE alues wi h
ewe genes) han he i s ou clus e s (Figu e 2). While
he a e age MSE alues o he se en h o he six een h
clus e s ha e no been included in ha Figu e, he numbe s
o genes included in hese clus e s a ela i ely lowe le els
o igh ness, as shown in Table 2, a e su icien o il e
hem ou . The e o e, we ha e conside ed he clus e s C1
and C2 o u he analysis in his s udy.
A e age exp ession p o iles
The a e age exp ession p o iles o he clus e s C1 and
C2 a DTB wi h δ= 0.3 and 0.2 espec i ely, in each o
he o y da ase s a e plo ed in Figu e 3. Fo cla i y,
e o ba s ha e been supp essed as he in o ma ion,
which hey p o ide can be ob ained om he MSE analysis
in Figu e 2 and he plo s in Addi ional ile 2: Figu e S1,
which shows he exp ession p o iles o all o he genes in
hese wo clus e s a a ious δ alues.
De ailed sc u iny o Figu e 3 leads o he gene al
obse a ions ha he i s clus e , C1, is up- egula ed
when cells a e eleased om s ess condi ions such as
nu ien limi a ion; hey a e down- egula ed when s ess
condi ions a e e-imposed. Mos in e es ingly, he clus e
C2 shows opposi e a e age exp ession p o iles in almos
all o he o y da ase s o he a e age p o iles o clus e
C1 wi h no phase shi , i.e. wi h nei he p o ile leading o
lagging he o he ; i s genes a e up- egula ed unde s ess
condi ions and down- egula ed unde g ow h condi ions.
I is in e es ing, bu had no been an icipa ed a he ime
o expe imen al design be o e ob aining he esul s, ha
he wo mos consis en ly co-exp essed clus e s o genes
in budding yeas show such clea opposi e exp ession
p o iles ac oss la ge numbe o da ase s.
To assess ha obse ed opposi e co-exp ession quan-
i a i ely, we ha e calcula ed he Pea son’s co ela ion
Table 2 Numbe s o genes included in each o he 16 clus e s a all o he conside ed δ alues
Tigh ness δClus e
C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 C16
Complemen a y 0.0 1085 1457 610 655 592 268 303 175 175 154 143 92 51 49 29 10
0.1 516 394 84 105 79 12 9 3 1 2 2 0 0 0 0 0
0.2 344 47 17 14 2 0 0 0 0 0 0 0 0 0 0 0
0.3 257 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0
0.4 164 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0.5790 00000000 0 0 0 0 0 0
0.6220 00000000 0 0 0 0 0 0
0.70 0 00000000 0 0 0 0 0 0
0.80 0 00000000 0 0 0 0 0 0
0.90 0 00000000 0 0 0 0 0 0
Tigh es 1.0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0.1 0.2 0.3 0.4 0.5
0
0.2
0.4
0.6
0.8
A e age MSE
(A)
0 0.1 0.2 0.3 0.4 0.5
100
102
DTB δ alue
Numbe o genes in clus e
(B)
C1
C2
C3
C4
C5
C6
Figu e 2 A e age MSE alues and he numbe o genes
included in he igh es six clus e s a all o he adop ed δ
alues. (A) A e age MSE alues and (B) numbe o genes included.
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 6 o 20
h p://www.biomedcen al.com/1471-2105/15/322
alues be ween he a e age exp ession p o iles o C1 a
δ= 0.3 and C2 a δ= 0.3 om each o he o y da ase s.
A e y s ong nega i e co ela ion has been ound, ha
is lowe han he alue o −0.75 a 37 ou o 40 da ase s
and ne e exceeds he alue o −0.6 excep a a single
ou lie da ase , D35. This s ong nega i e co ela ion is
consis en e en when he δ alues a e a ied. Fo ins ance,
when conside ing C1 a he δ alues o 0.2 and 0.4, he
calcula ed co ela ion alues a e lowe han −0.75 a 38 and
36 ou o 40 da ase s, espec i ely. E en when conside ing
C2 a δ= 0.1, he case a which i s size is many olds la ge
han a δ=0.2(394genes e sus84),35ou o 40da ase s
show s ong nega i e co ela ion wi h alues lowe
han −0.75, and only couple o da ase s exceed he alue
o −0.7. The single ou lie da ase D35 has consis en ly
shown no ably weake nega i e co ela ion a all o he
a o emen ioned δ alues. These expe imen s demons a e
he obus ness o ou obse a ion ha C1 and C2 a e
consis en ly nega i ely co ela ed.
P omo e s en ichmen analysis
Because co-exp ession o e la ge numbe o di e en
mic oa ay da ase s s ongly indica es co- egula ion, we
ha e analysed he ups eam DNA sequences o he genes
D01D02D03D04D05
D06
D07D08D09
D10
D11D12D13
D14D15
D16D17
D18
D19D20
D21D22D23D24D25D26D27D28D29D30D31D32D33D34
D35
D36D37D38D39D40
C2 − δ = 0.2
(47 genes)
C1 − δ = 0.3
(257 genes)
condi ions / ime−poin s
condi ions / ime−poin s
Figu e 3 A e age exp ession p o iles o he clus e s C1 and C2 a DTB wi h he espec i e δ alues o 0.3 and 0.2, based on all o he
o y da ase s. Each column o plo s ep esen s a clus e and each ow ep esen s a da ase .
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 7 o 20
h p://www.biomedcen al.com/1471-2105/15/322
in he clus e s C1 and C2 o explo e po en ial common
ansc ip ion ac o s’binding si es. We ha e used he
MEME ool [50,51] o sea ch o he mos en iched DNA
sequence mo i s wi hin he 300 ups eam base-pai s o he
164 genes included in C1 a DTB wi h δ= 0.4. The h ee
disco e ed mo i s, which we label as C1-1, C1-2, and C1-3
espec i ely, we e hen ed o he TOMTOM ool [52,53]
o mine o p e iously known mo i s wi h high simila i y.
The i s mo i , wi h an E- alue o 3.3 × 10
−333
, was ound
o be he PAC mo i , which is he binding si e o he
wo pa alogous ansc ip ion ac o s Do 6p and Tod6p
wi h p- alues o 2.1 × 10
−5
and 1.4 × 10
−4
, espec i ely, and
i signi ican ly ma ches he binding si e o he ansc ip ion
ac o S l1p wi h a p- alue o 1.3 × 10
−4
(Figu e 4(A)). The
E- alue es ima es he expec ed numbe o mo i s wi h he
gi en p obabili y o highe , and wi h he same wid h and
si e coun , ha would be ound in a se o andom
sequences o a simila size. The second mo i , wi h
an E- alue o 2.2 × 10
−115
, was ound o be he RRPE
mo i , which is he binding si e o he ansc ip ion
ac o S b3p wi h a p- alue o 8.9 × 10
−7
(Figu e 4(B));
i also signi ican ly ma ches he binding si es o he
Figu e 4 Ups eam sequence analysis o he clus e C1. (A),(B), and (C) show he mo i s C1-1, C1-2, and C1-3 espec i ely and hei highly
ma ched known ansc ip ion ac o s’binding si es. (D) is a Venn diag am ha shows he numbe s o genes’ups eam sequences in C1 ha
con ain each o hese h ee mo i s.
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 8 o 20
h p://www.biomedcen al.com/1471-2105/15/322
o he signals hemsel es. Al hough his is a possible
p oposal, he ac ha he signals ha consis en ly
and synch onically egula e bo h g oups a e la gely
a ian , we ocus on he hypo hesis ha bo h g oups
a e egula ed by a common machine y, o ha hei
egula o y machine ies egula ed by a common egula o .
Indeed, he la e p oposal con o ms o he mo e gene al
s a emen o B aue and colleagues ha such consis en
posi i e o nega i e co ela ion e lec s sys em-le el
egula o y mechanisms [22].
Po en ial egula o s o APha-RiB and common egula o s
o RRB and APha-RiB
Gasch and Roy and hei collabo a o s commonly iden i ied
he Msn2p and i s pa alogue Msn4p as egula o s o he
subse s o genes which hey iden i ied as nega i ely
co ela ed wi h g ow h [23,60]. Gasch and colleagues
also iden i ied Yap1p as a egula o o hei g oup
[60] while Roy and colleagues iden i ied R g1p and
Ad 1p [23]. In e es ingly, ups eam analysis o ou
no el clus e APha-RiB (C2) has iden i ied Az 1p and he
pa alogous pai Msn2p and Msn4p as po en ial egula o s
(Figu e 5). I is wo h no icing ha he h ee s udies
mu ually iden i y Msn2p and Msn4p, which a e well
known o hei ole in s ess esponse egula ion h ough
binding o he STRE mo i (Figu e 10) [66,67].
Mo e in e es ingly, Az 1p has been iden i ied by ou
esul s as a po en ial egula o o in bo h clus e s RRB
(C1) and APha-RiB (C2) (Figu es 4, 5, and 10). Az 1p is a
zinc- inge ansc ip ion ac o , which has been p edic ed
o ha e ole in one o he pu a i e s ess esponse egula-
o y modules [68,69]. Mo eo e , i is exclusi ely localised
in he nucleus and i was ound o be syn hesised in highe
amoun s unde non- e men able g ow h condi ions [70].
By moni o ing di e en ially exp essed genes when AZF1
was knocked down, Sla e y and colleagues showed ha
his gene’s p oduc pa icipa es in he ansc ip ion o
wo non-o e lapping subse s o genes unde wo di e en
C1
RRB
G ow h
condi ions
Gene ic
s ess
condi ions
C2
APha-RiB
S b3
Do 6/
Tod6
S p1
S p1 S b3
Do 6/
Tod6
Sch9
TORC1
Sch9 P
PKA
Tpk1-3 Bcy1
Tpk1-3 Bcy1
Msn2/4
C 1
C 1
Yak1
Az 1
Me abolic
in e ac ion /
ansloca ion
Rep ession ( om
li e a u e)
Rep ession ( om
his s udy)
Ac i a ion ( om
li e a u e)
Ac i a ion ( om
his s udy)
Link also de ec ed
by he li e a u e
Link hypo hesised
in his s udy
Nucleus
Cy oplasm
Figu e 10 Regula ion o he RRB clus e (C1) and he APha-RiB clus e (C2). Ticked dashed links ha e been de ec ed in his s udy and we e
also p e iously iden i ied in he li e a u e while dashed links wi h ques ion ma ks ha e been only de ec ed in his s udy. Howe e , mos o he
p e ious s udies conside one o ew s ess condi ions in con as o “gene ic s ess condi ions”. No ice ha he clus e “C2 APha-RiB”is no el and
ha he links om he li e a u e ha poin a i a e based on he assump ion ha i is a s ess esponse module.
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 15 o 20
h p://www.biomedcen al.com/1471-2105/15/322

condi ions. The common aspec be ween hese non-
o e lapping subse s o genes is ha ing he mo i
AAAAGAAA in hei p omo e s [71]. Al hough ou
C2 genes a δ= 0.2 a e no included in any o hese wo
subse s, he exis ence o he AZF1 binding si e in hei
p omo e s indica es ha AZF1 may egula e exp ession o
genes in his clus e unde o he condi ions.
Ano he candida e common egula o is S b3p
(Figu e 10), which binds o he consensus mo i
TGAAAAA [61,62,72]. This mo i la gely o e laps wi h
he RRPE mo i ound in he ups eam sequences o he
RRB genes in ou esul s, as iden i ied by he TOMTOM
ool (Figu e 4(B)). Al hough no iden i ied by he
TOMTOM ool as a po en ial binding ansc ip ion
ac o , i s binding mo i TGAAAA la gely o e laps
wi h he pa o he mo i C2-1 (Figu e 5). Mo eo e ,
S b3p o e exp ession was shown o inc ease esis ance o
oxida i e s ess [63] and o esul in down- egula ion o
ibosome biogenesis genes [61,62,72], and Liko and
colleagues also p edic ed ha S b3p would be expec ed o
egula e ansc ip ion o o he unknown se s o genes
posi i ely [61,62].
The e idence o Az 1p o S b3p ac ing as a ansc ip ion
ac i a o and/o ep esso wi h ela ion o bo h g oups o
genes –RRB genes (C1), and APha-RiB genes (C2) is
unclea . Ne e heless, he e a e enough obse a ions
o specula e ha one o hem o bo h o hem may
play a ole in he mu ual ansc ip ional egula ion o
bo h RRB and APha-RiB. The molecula mechanism(s)
and signi icance o hose ansc ip ion ac o s in his
con ex emain o be es ablished.
A subse o eigh genes in he APha-RiB clus e a e highly
connec ed ac oss a ious gene ne wo ks
S ikingly, a no el subse o eigh ou o he 47 genes in
APha-RiB (C2 a DTB wi h δ= 0.2) ha e shown high
connec i i y in co-exp ession, p o ein-p o ein physical
in e ac ions, and gene ic in e ac ions (highligh ed in
Figu e 7 and Figu e 8). The genes YIR016W, AIM19,
and OM14 ha e unknown biological p ocesses. The la e
wo localise in he mi ochond ia while he localisa ion o
YIR016W is unknown. UGA2 is an oxida i e s ess
esponse gene which localises in he cy oplasm. PMP3 is a
plasma memb ane gene ha pa icipa es in esponse o
d ugs and egula ion o memb ane po en ial. YOR228C’s
p oduc is a mi ochond ial p o ein which is in ol ed in
lipid homeos asis bu wi h an unknown unc ion. NCR1’s
p oduc is a acuola memb ane p o ein which pa ici-
pa es in acuola p o ein so ing pa hway. Finally, YSA1’s
p oduc pa icipa es in ibose phospha e me abolism and
was ound localising in he mi ochond ion, cy oplasm,
and he nucleus. Clea ly, hose genes, gene ally, ha e
unknown o appa en ly un ela ed unc ions despi e his
high connec i i y.
One ocal gene wi h p e iously unknown unc ion is
YIR016W. La ge scale o e exp ession sc eening in yeas
e ealed ha his gene’s o e exp ession causes cell-cycle
o be a es ed by accumula ing cells a he G2/M s age
[73], which is consis en wi h i s down- egula ion du ing
he cell-cycle as shown in ou esul s (e.g. da ase s D01,
D02, D21, and D22 in Figu e 3). A es ing he cell-cycle
unde s ess is one o he known mechanisms o s ess
esponse [9]. I s co-exp ession, p o ein-p o ein physical
in e ac ion, and nega i e gene ic in e ac ion wi h he
s ess esponse gene UGA2 s eng hens he hypo hesis
ha his gene may pa icipa e in s ess esponse, and i s
connec ions wi h UGA2, YSA1, and AIM19 p o ide a
concise pla o m o u u e unc ional s udies. This
gene’s ole in cell-cycle a es /delay and in any o he
mechanisms o s ess esponse has o be e ealed in
u u e unc ional s udies.
The highly consis en up- egula ion unde s ess,
down- egula ion unde g ow h, and high ne wo k
connec i i y o his no el and concise subse o
genes ac oss such wide ange o condi ions in o y
di e en da ase s indeed indica e ha hey ha e oles
in some o he mechanisms ela ed o gene ic s ess
esponse. A e sc u inising his sub-ne wo k, as well as
he APha-RiB clus e in gene al, i becomes clea ha
many de ails ega ding gene ic s ess esponse mechanisms
and hei membe genes a e ye o be elucida ed.
The Bi-CoPaM me hod is use ul o genome-wide
consis en ly co-exp essed genes disco e y
Ou esul s ha e demons a ed he use ulness o ou no el
app oach o using he Bi-CoPaM me hod o explo e
genome-wide exp ession da a, om a ious mic oa ay
da ase s om di e en biological con ex s and condi ions.
Mo e speci ically, we ha e de ined subse s o genes ha
a e consis en ly co-exp essed ac oss a ious mic oa ay
da ase s using a unable me hod and wi hou he need o
a p io i knowledge-based il e ing.
In con as o o he clus e ing and ensemble clus e ing
me hods, con igu ing he Bi-CoPaM me hod o gene a e
six een clus e s does no imply ha he inal objec i e is
o ge six een in o ma i e clus e s; he inal objec i e is
a he o mine o he ew subse s o genes which a e
consis en ly co-exp essed in all o mos o he conside ed
da ase s [18]. A la ge numbe o clus e s han expec ed
o be in o ma i e is equi ed o accoun o he la ge
a ia ion in he genome-wide exp ession. In Bi-CoPaM,
he genes ha a e consis en ly co-exp essed in all o mos
o he da ase s when conside ed by a ious clus e ing
me hods will cons an ly appea in he same clus e . The
majo i y o genes, which will show low consis ency in
co-exp ession, a e allowed a wide space o six een
clus e s o be assigned o. These genes will appea in
di e en clus e s when hei exp ession p o iles indi e en
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 16 o 20
h p://www.biomedcen al.com/1471-2105/15/322
da ase s a e conside ed om by he a ious clus e ing
me hods. The e o e, inconsis ency in co-exp ession is
e lec ed by inconsis ency in clus e assignmen .
Ou esul s ha e demons a ed ha a wide ange o
numbe s o clus e s (K) will esul in he same igh
clus e s. The Bi-CoPaM’s di e ence h eshold bina isa ion
(DTB) echnique unably igh ens he clus e s o include
he mos consis en ly co-exp essed genes while lea ing he
la ge bulk o he poo ly co-exp essed genes unassigned
om all o he clus e s [19,20]. This unable igh ening is
con olled by he pa ame e δ, which inc eases he
igh ness as i is inc eased. As can be seen in Table 2,
mos o he six een clus e s lose all o hei genes a
ela i ely low δ alue, which is o be expec ed as
mos o he genes will no be co-exp essed in mos o
he s udies and da ase s conside ed.
In hei 2001 s udy, Wade and colleagues clus e ed
abou hal o he genome in o 24 clus e s om h ee
di e en da ase s by a single clus e ing me hod, which
is he pa i ioning a ound medoids (PAM) me hod [74].
They hen pe o med s a is ical analysis o iden i y
o e lapping clus e s om di e en esul s, which led
o inding ha only one clus e om one da ase has
signi ican o e lap o ano he clus e om one o he
da ase . The in e sec ion be ween bo h clus e s had
65 genes, which we e ound o be la gely pa icipa ing
in ibosome biogenesis [74].
When compa ing he wo app oaches, h ee majo
di e ences a e he mos impo an . Fi s , ou app oach is
mo e sui able when la ge numbe s o da ase s a e consid-
e ed because o he sys ema ic way o using he esul s
in o a single consensus esul ha e lec s all in o ma ion.
Second, Bi-CoPaM allows o a ious c isp and/o uzzy
clus e ing me hods o be applied o e each single da ase ,
which adds ano he le el o di e si y. Thi d, and mos
impo an ly, ou Bi-CoPaM-based app oach is unable
and is no me ely limi ed o he in e sec ions o clus e s; i
di ec in e sec ion wo ked well o wo da ase s in Wade’s
app oach, i would esul in emp y clus e s in he case o
o y da ase s wi h a ious clus e ing me hods. This can
be di ec ly and clea ly seen in ou esul s as in e sec ion is
a special case o Bi-CoPaM’s esul s, and is ob ained
by DTB wi h δ= 1.0, he case a which all o he six een
clus e s ha e been ound comple ely emp y (Table 2). On
he o he hand, conside ing con en ional complemen a y
clus e s, which is again a special case o he Bi-CoPaM
esul s (DTB wi h δ= 0.0), is imp ac ical as i does
no educe he complexi y o he da ase s, and he
clus e s a his le el a e gene ally loose han accep able
(see Figu e 2 and Table 2). The e o e, he mos ui -
ul analysis, as demons a ed by ou s udy, is when
clus e s a e igh ened while main aining signi ican
numbe s o genes and he e he Bi-CoPaM app oach
allows obse a ion o clus e s’beha iou when δ
pa ame e is uned o p oduce igh clus e s in Table 2
and Figu e 2.
Taken oge he , ou app oach can analyse la ge amoun s
o high- h oughpu da ase s o p oduce ela i ely ocused
and comp ehensible esul s ha cap u e he mos consis en
aspec s o he aw da a. The me hod can he e o e disco e
hose subse s o genes mos consis en ly co-exp essed unde
a ious condi ions.
Conclusions
We ha e applied he Bi-CoPaM me hod o e genome-wide
da a om o y mic oa ay da ase s wi h wide ange o
di e en biological con ex s and expe imen al condi ions in
o de o iden i y he subse s o budding yeas genes ha a e
mos consis en ly co-exp essed. We ound wo clus e s o
genes ha ha e signi ican consis ency o co-exp essions,
which we ha e labelled as RRB (C1) and APha-RiB (C2).
These wo clus e s p ese ed hei s a us as he igh es
wo clus e s a a ying alues o K, which shows hei
impo ance as well as he obus ness o he p oposed
Bi-CoPaM app oach. By GO analysis, C1 has been ound
o be highly en iched wi h ibosome biogenesis and RNA
p ocessing (RRB) genes. On he o he hand, mos o
he genes included in C2 ha e unknown o appa en ly
un ela ed unc ions.
Finding RRB genes (C1) in he igh es clus e by his
comple ely unsupe ised app oach, con i ms no only
ha hese genes a e consis en ly co-exp essed unde
a ious condi ions [9], bu also ha hey a e he
mos consis en ly co-exp essed genes ac oss he
whole genome. Addi ionally, ou C1 clus e includes
ew genes wi h unknown p ocesses ha may be wo hy o
biological in es iga ion.
The mos in e es ing clus e o genes in ou esul s a-
ppea s o be C2, and his is o h ee main easons – i s ,
hese genes a e mos ly unknown o appa en ly un ela ed
o each o he , despi e he ac ha hey a e he second
mos consis en ly co-exp essed subse o genes in budding
yeas ; second, hei a e age exp ession p o iles show
consis en ly an i-phase (opposi e) exp ession o he
a e age exp ession p o iles o RRB genes (C1) ac oss
all o he o y da ase s; and hi d, signi ican gene ic
and p o ein-p o ein physical in e ac ions ha e been
epo ed be ween hem by high- h oughpu s udies in
he li e a u e. These obse a ions lead us o label C2
as he subse o genes in an i-phase wi h ibosome
biogenesis (APha-RiB), o sugges ha many o he
unknown genes in APha-RiB (C2), such as YIR016W, may
pa icipa e in di e en gene ic, in con as o speci ic,
s ess esponse mechanisms, and o sugges ha RRB
genes (C1) and he APha-RiB genes (C2) may be
ansc ip ionally egula ed by common machine y o
ha hei egula ion machine ies may be con olled by
common pos - ansla ional egula o s. We ha e iden i ied
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 17 o 20
h p://www.biomedcen al.com/1471-2105/15/322
po en ial ac o s ha migh be in ol ed in such ecip ocal
egula ion, o example Az 1p and S b3p.
This s udy has yielded globally consis en co-exp ession
in budding yeas and p oduced new, ocused insigh s o
u u e wo k o elucida e and con i m he componen s o
he common egula o y machine y o RRB and APha-RiB,
and o de ine he unc ion o poo ly cha ac e ised genes in
bo h clus e s. The esul s om he applica ion o he
Bi-CoPaM me hod o yeas da ase s s ongly sugges s
ha i may be help ul o he analysis o o he g oups
o mic oa ay da ase s om o he species and sys ems
o he explo a ion o global gene ic co-exp ession.
Addi ional iles
Addi ional ile 1: Table S1. A lis o he 5667 genes included in his
s udy and he con en s o all o he six een clus e s a all o he adop ed
δ alues.
Addi ional ile 2: Figu e S1. P o ides he p o iles o he genes
included in he clus e s C1 and C2 a he igh ness le els o DTB wi h
δ= 0.3 and 0.2 espec i ely. The p o iles a e p o ided om all o he
o y conside ed da ase s.
Addi ional ile 3: Table S2. GO Te m analysis esul s o he p ocesses
en iched in he clus e s C1 and C2.
Addi ional ile 4: Table S3. GO Slim analysis esul s o he p ocesses
en iched in he clus e s C1 and C2.
Addi ional ile 5: Table S4. GO Te m and GO Slim analysis esul s o
he cellula componen s en iched in he clus e C2 a δ= 0.2.
Addi ional ile 6: Figu e S2. C1 o e lap wi h simila clus e s in he
li e a u e, and he a ios o included genes associa ed wi h he “ ibosome
biogenesis”GO e m.
Compe ing in e es s
The au ho s decla e ha hey ha e no compe ing in e es s.
Au ho s’con ibu ions
BAJ, RF, and AKN concei ed he s udy, designed and pe o med he
expe imen s, analysed he da a, and w o e he pape . DJR concei ed he
s udy and w o e he pape . All au ho s ead and app o ed he inal
manusc ip .
Acknowledgemen s
This a icle summa ises independen esea ch unded by he Na ional
Ins i u e o Heal h Resea ch (NIHR) unde i s P og amme G an s o Applied
Resea ch P og amme (G an Re e ence Numbe RP-PG-0310-1004). The iews
exp essed a e hose o he au ho (s) and no necessa ily hose o he NHS,
he NIHR o he Depa men o Heal h. We hank D Taco Kooij o cons uc i e
commen s and ca e ul eading o he manusc ip . A. K. Nandi would like o
hank TEKES o hei awa d o he Finland Dis inguished P o esso ship.
Au ho de ails
1
Depa men o Elec onic and Compu e Enginee ing, B unel Uni e si y,
Uxb idge, Middlesex UB8 3PH, UK.
2
Na ional Heal h Se ice Blood and
T ansplan , Ox o d, UK.
3
Radcli e Depa men o Medicine, Uni e si y o
Ox o d, John Radcli e Hospi al, Ox o d, UK.
4
Depa men o Ma hema ical
In o ma ion Technology, Uni e si y o Jy äskylä, Jy äskylä, Finland.
Recei ed: 10 June 2014 Accep ed: 22 Sep embe 2014
Published: 29 Sep embe 2014
Re e ences
1. Robe s PC: Gene exp ession mic oa ay da a analysis demys i ied.
Bio echnol Annu Re 2008, 14:29–61.
2. Bou os PC, Okey AB: Unsupe ised pa e n ecogni ion: an in oduc ion
o he whys and whe e o es o clus e ing mic oa ay da a. B ie Bioin o m
2005, 6:331–343.
3. Kim J, Ogden RT, Kim H: A me hod o iden i y di e en ial exp ession p o iles
o ime-cou se gene da a wi h Fou ie ans o ma ion. BMC Bioin o ma ics
2013, 14.
4. Spellman PT, She lock G, Zhang MQ, Iye VR, Ande s K, Eisen MB, B own PO,
Bo s ein D, Fu che B: Comp ehensi e iden i ica ion o cell cycle– egula ed
genes o he yeas saccha omyces ce e isiae by mic oa ay hyb idiza ion.
Mol Biol Cell 1998, 9:3273–3297.
5. Liu D, Umbach DM, Peddada SD, Li L, C ocke PW, Weinbe g CR: A
andom-pe iods model o exp ession o cell-cycle genes. PNAS 2004,
11:7240–7245.
6. Tu BP, Kudlicki A, Rowicka M, McKnigh SL: Logic o he yeas me abolic
cycle: empo al compa men aliza ion o cellula p ocesses. Science 2005,
310:1152–1158.
7. P amila T, Wu W, Miles S, Noble WS, B eeden LL: The Fo khead
ansc ip ion ac o Hcm1 egula es ch omosome seg ega ion genes and
ills he S-phasegap in he ansc ip ional ci cui y o he cell cycle.
Genes De 2006, 20:2266–2278.
8. Eisen MB, Spellman PT, B own PO, Bo s ein D: Clus e analysis and display
o genome-wide exp ession pa e ns.InP oc. Na l. Acad. Sci;
1998:14863–14868.
9. Wade CH, Umba ge MA, McAlea MA: The budding yeas RNA and
ibosome biosyn hesis (RRB) egulon con ains o e 200 genes.
Yeas 2006, 23:293–306.
10. Nilsson R, Schul z IJ, Pie ce EL, Sol is KA, Na anun a a A, Wa d DM,
Baughman JM, Pa adka PN, Kingsley PD, Culo a VC, Kaplan J, Palis J,
Paw BH, Moo ha VK: Disco e y o genes essen ial o heme biosyn hesis
h ough la ge-scale gene exp ession analysis. Cell Me ab 2009,
10:119–130.
11. Cooke EJ, Sa age RS, Ki k PD, Da kins R, Wild DL: Bayesian hie a chical
clus e ing o mic oa ay ime se ies da a wi h eplica es and ou lie
measu emen s. BMC Bioin o ma ics 2011, 12:399.
12. Ayadi W, Elloumi M, Hao JK: Pa e n-d i en neighbo hood sea ch o
biclus e ing o mic oa ay da a. BMC Bioin o ma ics 2012, 13(Suppl 7):S11.
13. Xiao X, Dow ER, Ebe ha R, Miled ZB, Oppel RJ: Gene clus e ing using
sel -o ganizing maps and pa icle swa m op imiza ion. In IEEE Pa allel and
Dis ibu ed P ocessing Symposium P oceedings: Indianapolis; 2003:154–163.
14. Dikicioglu D, Ka abekmez E, Rash B, Pi P, Ki da B, Oli e SG: How yeas
e-p og ammes i s ansc ip ional p o ile in esponse o di e en
nu ien impulses. BMC Sys Biol 2011, 5:148–163.
15. Salem SA, Jack LB, Nandi AK: In es iga ion o sel -o ganizing oscilla o
ne wo ks o use in clus e ing mic oa ay da a. IEEE T ans Nanobioscience
2008, 7:65–79.
16. Ayad HG, Kamel MS: On o ing-based consensus o clus e ensembles.
Pa e n Recogn 2010, 43:1943–1953.
17. F ed A, Jain AK: Da a clus e ing using e idence accumula ion. In
P oceedings o he Six een h In e na ional Con e ence on Pa e n Recogni ion
(ICPR); 2002:276–280.
18. Yu Z, Wong HS, Wang H: G aph-based consensus clus e ing o class
disco e y om gene exp ession da a. Bioin o ma ics 2007, 23:2888–2896.
19. Abu-Jamous B, Fa R, Robe s DJ, Nandi AK: Pa adigm o unable clus e ing
using bina iza ion o consensus pa i ion ma ices (Bi-CoPaM) o gene
disco e y. PLoS One 2013, 8(2):e56432.
20. Abu-Jamous B, Fa R, Robe s DJ, Nandi AK: Yeas gene CMR1/YDL156W is
consis en ly co-exp essed wi h genes pa icipa ing in DNA-me abolic
p ocesses in a a ie y o s ingen clus e ing expe imen s. J R Soc
In e ace 2013, 10(81):20120990.
21. Abu-Jamous B, Fa R, Robe s DJ, Nandi AK: Iden i ica ion o genes consis en ly
co-exp essed in mul iple mic oa ays by a genome-wide app oach. In The
P oceedings o he 38 h In e na ional Con e ence on Acous ics, Speech, and Signal
P ocessing (ICASSP); Vancou e , Canada; 2013:1172–1176.
22. B aue MJ, Hu enhowe C, Ai oldi EM, Rosens ein R, Ma ese JC, G esham D,
Boe VM, T oyanskaya OG, Bo s ein D: Coo dina ion o g ow h a e, cell
cycle, s ess esponse, and me abolic ac i i y in yeas . Mol Biol Cell 2008,
19:352–367.
23. Roy S, Wapinski I, P i ne J, F ench C, Socha A, Konieczka J, Habib N, Kellis M,
Thompson D, Rege A: A bo e um: econs uc ion and analysis o he
e olu iona y his o y o condi ion-speci ic ansc ip ional modules.
Genome Res 2013, 23:1039–1050.
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 18 o 20
h p://www.biomedcen al.com/1471-2105/15/322
24. Con ad M, Scho ho s J, Kankipa i HN, Zeeb oeck GV, Rubio-Texei a M,
The elein JM: Nu ien sensing and signaling in he yeas Saccha omyces
ce e isiae. FEMS Mic obiol Re 2014, 38:254–299.
25. Hube A, F ench SL, Teko e H, Ye likaya S, S ahl M, Pe epelkina MP, Tye s M,
Rougemon J, Beye AL, Loewi h R: Sch9 egula es ibosome biogenesis
ia S b3, Do 6 and Tod6 and he his one deace ylase complex RPD3L.
EMBO 2011, 30:3052–3064.
26. Lam YK, Tsang PWM: eXplo a o y K-Means: a new simple and e icien
algo i hm o gene clus e ing. Appl So Compu 2012, 12:1149–1157.
27. Zhu Z, Liu W, He S, Ji Z: Meme ic clus e ing based on pa icle swa m
op imize and k-means.In2012 IEEE Cong ess on E olu iona y Compu a ion
(CEC); B isbane, Aus alia; 2012.
28. O lando DA, Lin CY, Be na d A, Wang JY, Socola JES, I e sen ES, Ha emink AJ,
Haase SB: Global con ol o cell-cycle ansc ip ion by coupled CDK and
ne wo k oscilla o s. Na u e 2008, 453:944–947.
29. Chumnanpuen P, Nookaew I, Nielsen J: In eg a ed analysis,
ansc ip ome-lipidome, e eals he e ec s o INO-le el (INO2 and
INO4) on lipid me abolism in yeas . BMC Sys Biol 2013, 7:S7.
30. Xue-F anzén Y, Hen iksson J, Bü glin TR, W igh AP: Dis inc oles o he
Gcn5 his one ace yl ans e ase e ealed du ing ansien s ess-induced
ep og amming o he genome. BMC Genomics 2013, 14:479.
31. La sson M, U ell H, Sands öm J, Rydén P, Sel h LA, Bjö klund S: Func ional
s udies o he yeas med5, med15 and med16 media o ail subuni s.
PLoS One 2013, 8:e73137.
32. Liu Z, Ös e lund T, Hou J, Pe ano ic D, Nielsen J: Anae obic α-amylase
p oduc ion and sec e ion wi h uma a e as he inal elec on accep o in
Saccha omyces ce e isiae. Appl En i on Mic obiol 2013, 79:2962–2967.
33. Suzuki T, Iwahashi Y: Comp ehensi e gene exp ession analysis o ype B
icho hecenes. J Ag ic Food Chem 2012, 60:9519–9527.
34. Sanz AB, Ga cía R, Rod íguez-Peña JM, Díez-Muñiz S, Nombela C, Pe e son CL,
A oyo J: Ch oma in emodeling by he SWI/SNF complex is essen ial o
ansc ip ion media ed by he yeas cell wall in eg i y MAPK pa hway.
Mol Biol Cell 2012, 23:2805–2817.
35. Lanza AM, Blazeck JJ, C ook NC, Alpe HS: Linking yeas Gcn5p ca aly ic
unc ion and gene egula ion using a quan i a i e, g aded dominan
mu an app oach. PLoS One 2012, 7:e36193.
36. Chin SL, Ma cus IM, Kle ecz RR, Li CM: Dynamics o oscilla o y pheno ypes in
Saccha omyces ce e isiae e eal a ne wo k o genome-wide ansc ip ional
oscilla o s. FEBS J 2012, 279:1119–1130.
37. Ko acs LAS, Mayhew MB, O lando DA, Jin Y, Li Q, Huang C, Reed SI,
Mukhe jee S, Haase SB: Cyclin-dependen kinases a e egula o s and
e ec o s o oscilla ions d i en by a ansc ip ion ac o ne wo k. Mol Cell
2012, 45:669–679.
38. González-Aguile a C, Tous C, Babiano R, C uz JDL, Luna R, Aguile a A: Nab2
unc ions in he me abolism o RNA d i en by polyme ases II and III.
Mol Biol Cell 2011, 22:2729–2740.
39. Fe ei a RT, Sil a ARC, Pimen el C, Ba is a-Nascimen o L, Rod igues-Pousada C,
Menezes RA: A senic s ess elici s cy osolic Ca(2+) bu s s and C z1 ac i a ion
in Saccha omyces ce e isiae. Mic obiology 2012, 158:2293–2302.
40. Bes e MC, Jacobson D, Baue FF: Many saccha omyces ce e isiae cell wall
p o ein encoding genes a e co egula ed by Mss11, bu cellula adhesion
pheno ypes appea only lo p o ein dependen . G3 (Be hesda) 2012,
2:131–141.
41. Suzuki T, Iwahashi Y: Gene exp ession p o iles o yeas Saccha omyces
ce e isiae sod1 caused by pa ulin oxici y and e alua ion o eco e y
po en ial o asco bic acid. J Ag ic Food Chem 2011, 59:7145–7154.
42. Pa ei as LS, Kohn LM, Ande son JB: Cellula e ec s and epis asis among
h ee de e minan s o adap a ion in expe imen al popula ions o
Saccha omyces ce e isiae. Euka yo Cell 2011, 10:1348–1356.
43. Mo illo-Huesca M, Clemen e-Ruiz M, Andúja E, P ado F: The SWR1
his one eplacemen complex causes gene ic ins abili y and
genome-wide ansc ip ion mis egula ion in he absence o H2A.Z.
PLoS One 2010, 5:e12143.
44. Ma ia-González AM, Rod íguez-Gab iel MA: Sl 2 MAPK pa hway is essen ial
o cell in eg i y in he p esence o a sena e. Yeas 2011, 28:9–17.
45. S assbu g K, Wal he D, Takahashi H, Kanaya S, Kopka J: Dynamic
ansc ip ional and me abolic esponses in yeas adap ing o
empe a u e s ess. OMICS 2010, 14:249–259.
46. Wade SL, Poo ey K, Beki ano S, Auble DT: The Sn 1 kinase and
p o easome-associa ed Rad23 egula e UV- esponsi e gene exp ession.
EMBO J 2009, 28:2919–2931.
47. Ge H, Wei M, Fab izio P, Hu J, Cheng C, Longo VD, Li LM: Compa a i e
analyses o ime-cou se gene exp ession p o iles o he long-li ed
sch9Del a mu an . Nucleic Acids Res 2010, 38:143–158.
48. Pena JM, Lozano JA, La anaga P: An empi ical compa ison o ou
ini ializa ion me hods o he K-Means algo i hm. Pa e n Recogn Le
1999, 20:1027–1040.
49. Bols ad B, I iza y R, As and M, Speed T: A compa ison o no maliza ion
me hods o high densi y oligonucleo ide a ay da a based on a iance
and bias. Bioin o ma ics 2003, 19:185–193.
50. Bailey TL, Elkan C: Fi ing a mix u e model by expec a ion maximiza ion
o disco e mo i s in biopolyme s. In P oceedings o he Second
In e na ional Con e ence on In elligen Sys ems o Molecula Biology; Menlo
Pa k, Cali o nia; 1994:28–36.
51. MEME: Mul iple Em o Mo i Elici a ion (MEME) ool. [Online]; 2014. A ailable
om: HYPERLINK h p://meme.nbc .ne /meme/cgi-bin/meme.cgi.
52. Gup a S, S ama oyannopolous JA, Bailey T, Noble WS: Quan i ying simila i y
be ween mo i s. Genome Biol 2007, 8:R24.
53. TOMTOM: TOMTOM Mo i Simila i y Tool. [Online]; 2014. A ailable om:
HYPERLINK h p://meme.nbc .ne /meme/cgi-bin/ om om.cgi.
54. Peng J, Chen J, Wang Y: Iden i ying c oss-ca ego y ela ions in gene
on ology and cons uc ing genome-speci ic e m associa ion ne wo ks.
BMC Bioin o ma ics 2013, 14(Suppl 2):S15.
55. SGD: Saccha omyces Genome Da abase GO Te m Finde . [Online]; 2014. [ci ed
2012 Sep embe . A ailable om: HYPERLINK h p://www.yeas genome.o g/
cgi-bin/GO/goTe mFinde .pl.
56. SGD: Saccha omyces Genome Da abase GO Slim Mappe . [Online]; 2014.
[ci ed 2012 No embe . A ailable om: HYPERLINK h p://www.
yeas genome.o g/cgi-bin/GO/goSlimMappe .pl.
57. GeneMANIA: Gene Mul iple Associa ion Ne wo k In eg a ion Algo i hm.
[Online]; 2014. A ailable om: HYPERLINK h p://www.genemania.o g/.
58. Cos anzo M, Ba yshniko a A, Bellay J, Kim Y, Spea ED, Se ie CS, Ding H,
Koh JLY, Tou ighi K, Mos a a i S, P inz J, Onge RPS, Vande Sluis B,
Makhne ych T, Vizeacouma FJ, Alizadeh S, Bah S, B os RL, Chen Y, Cokol
M, Deshpande R, Li Z: The gene ic landscape o a cell. Science 2010,
327:425–431.
59. Tsanko AM, Thompson DA, Socha A, Rege A, Rando OJ: The ole o
nucleosome posi ioning in he e olu ion o gene egula ion. PLoS Biol
2010, 8:e1000414.
60. Gasch AP, Spellman PT, Kao CM, Ca mel-Ha el O, Eisen MB, S o z G, Bo s ein D,
B own PO: Genomic exp ession p og ams in he esponse o yeas cells o
en i onmen al changes. Mol Biol Cell 2000, 11:4241–4257.
61. Liko D, Sla e y MG, Heideman W: S b3 binds o ibosomal RNA
p ocessing elemen mo i s ha con ol ansc ip ional esponses o
g ow h in Saccha omyces ce e isiae. J Biol Chem 2007, 282:26623–26628.
62. Liko D, Conway MK, G unwald DS, Heideman W: S b3 plays a ole in he
glucose-induced ansi ion om quiescence o g ow h in Saccha omyces
ce e isiae. Gene ics 2010, 185:797–810.
63. D obna E, Gazdag Z, Culako a H, Dzugaso a V, Gbelska Y, Pes i M, Subik J:
O e exp ession o he YAP1, PDE2, and STB3 genes enhances he
ole ance o yeas o oxida i e s ess induced by 7-chlo o e azolo[5,1-c]
benzo[1,2,4] iazine. FEMS Yeas Res 2012, 12:958–968.
64. Gasch AP, Huang M, Me zne S, Bo s ein D, Elledge SJ, B own PO: Genomic
exp ession esponses o DNA-damaging agen s and he egula o y ole
o he yeas ATR homolog Mec1p. Mol Biol Cell 2001, 12:2987–3003.
65. Tkach JM, Yimi A, Lee AY, Ri le M, Cos anzo M, Jaschob D, Hend y JA, Ou J,
Mo a J, Boone C, Da is TN, Nislow C, B own GW: Dissec ing DNA damage
esponse pa hways by analysing p o ein localiza ion and abundance
changes du ing DNA eplica ion s ess. Na Cell Biol 2012, 14:966–976.
66. Ma ínez-Pas o MT, Ma chle G, Schülle C, Ma chle -Baue A, Ruis H, Es uch
F: The Saccha omyces ce e isiae zinc inge p o eins Msn2p and Msn4p
a e equi ed o ansc ip ional induc ion h ough he s ess esponse
elemen (STRE). EMBO 1996, 15:2227–2235.
67. Schmi AP, McEn ee K: Msn2p, a zinc inge DNA-binding p o ein, is he
ansc ip ional ac i a o o he mul is ess esponse in Saccha omyces
ce e isiae. PNAS 1996, 93:5777–5782.
68. Segal E, Shapi a M, Rege A, Pe'e D, Bo s ein D, Kolle D, F iedman N:
Module ne wo ks: iden i ying egula o y modules and hei
condi ion-speci ic egula o s om gene exp ession da a. Na Gene
2003, 34:166–176.
69. SGD: Saccha omyces Genome Da abase. [Online]; 2014. [ci ed 2012 Ap il 15.
A ailable om: HYPERLINK h p://www.yeas genome.o g/.
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 19 o 20
h p://www.biomedcen al.com/1471-2105/15/322
70. S ein T, K icke J, Beche D, Lisowsky T: Az 1p is a nuclea -localized
zinc- inge p o ein ha is p e e en ially exp essed unde non- e men a i e
g ow h condi ions in Saccha omyces ce e isiae. Cu Gene 1998,
34:287–296.
71. Sla e y MG, Liko D, Heideman W: The unc ion and p ope ies o he
Az 1 ansc ip ional egula o change wi h g ow h condi ions in
Saccha omyces ce e isiae. Euka yo ic Cell 2006, 5:313–320.
72. Zhu C, Bye s KJRP, McCo d RP, Shi Z, Be ge MF, Newbu ge DE, Saul ie a K,
Smi h Z, Shah MV, Radhak ishnan M, Philippakis AA, Hu Y, Masi FD, Pacek M,
Rol s A, Mu hy T, LaBae J, Bulyk ML: High- esolu ion DNA binding
speci ici y analysis o yeas ansc ip ion ac o s. Genome Res 2009,
19:556–566.
73. Niu W, Li Z, Zhan W, Iye VR, Ma co e EM: Mechanisms o cell cycle
con ol e ealed by a sys ema ic and quan i a i e o e exp ession sc een
in S. ce e isiae. PLoS Gene 2008, 4:e1000120.
74. Wade C, Shea KA, Jensen RV, McAlea MA: EBP2 is a membe o he yeas
RRB egulon, a ansc ip ionally co egula ed se o genes ha a e equi ed
o ibosome and RNA biosyn hesis. Mol Cell Biol 2001, 21:8638–8650.
doi:10.1186/1471-2105-15-322
Ci e his a icle as: Abu-Jamous e al.:Comp ehensi e analysis o o y
yeas mic oa ay da ase s e eals a no el subse o genes (APha-RiB)
consis en ly nega i ely associa ed wi h ibosome biogenesis. BMC
Bioin o ma ics 2014 15:322.
Submi you nex manusc ip o BioMed Cen al
and ake ull ad an age o :
• Con enien online submission
• Tho ough pee e iew
• No space cons ain s o colo figu e cha ges
• Immedia e publica ion on accep ance
• Inclusion in PubMed, CAS, Scopus and Google Schola
• Resea ch which is eely a ailable o edis ibu ion
Submi you manusc ip a
www.biomedcen al.com/submi
Abu-Jamous e al. BMC Bioin o ma ics 2014, 15:322 Page 20 o 20
h p://www.biomedcen al.com/1471-2105/15/322