BioMed Cen al
Page 1 o 13
(page numbe no o ci a ion pu poses)
BMC Bioin o ma ics
Open Access
P oceedings
On ology Design Pa e ns o bio-on ologies: a case s udy on he
Cell Cycle On ology
Mikel Egaña A angu en*1, E ick An ezana2,3, Ma in Kuipe 2,3 and
Robe S e ens1
Add ess: 1School o Compu e Science, Uni e si y o Manches e , Ox o d Road, M13 9PL Manches e , UK, 2Depa men o Plan Sys ems Biology,
VIB, Technologiepa k 927, 9052 Gen , Belgium and 3Depa men o Molecula Gene ics, Ghen Uni e si y, Technologiepa k 927, 9052 Gen ,
Belgium
Email: Mikel Egaña A angu en* - mike[email p o ec ed]; E ick An ezana - [email p o ec ed];
Ma in Kuipe - ma in.kui[email p o ec ed]; Robe S e ens - obe .s e[email p o ec ed]
* Co esponding au ho
Abs ac
Backg ound: Bio-on ologies a e key elemen s o knowledge managemen in bioin o ma ics. Rich
and igo ous bio-on ologies should ep esen biological knowledge wi h high ideli y and
obus ness. The ichness in bio-on ologies is a p io condi ion o di e se and e icien easoning,
and hence que ying and hypo hesis alida ion. Rigou allows a mo e consis en main enance.
Modelling such bio-on ologies is, howe e , a di icul ask o bio-on ologis s, because he necessa y
ichness and igou is di icul o achie e wi hou ex ensi e aining.
Resul s: Analogous o design pa e ns in so wa e enginee ing, On ology Design Pa e ns a e
solu ions o ypical modelling p oblems ha bio-on ologis s can use when building bio-on ologies.
They o e a means o c ea ing ich and igo ous bio-on ologies wi h educed e o . The concep
o On ology Design Pa e ns is desc ibed and documen a ion and applica ion me hodologies o
On ology Design Pa e ns a e p esen ed. Some eal-wo ld use cases o On ology Design Pa e ns
a e p o ided and es ed in he Cell Cycle On ology. On ology Design Pa e ns, including hose
es ed in he Cell Cycle On ology, can be explo ed in he On ology Design Pa e ns public
ca alogue ha has been c ea ed based on he documen a ion sys em p esen ed (h p://
odps.sou ce o ge.ne /).
Conclusions: On ology Design Pa e ns p o ide a me hod o ich and igo ous modelling in bio-
on ologies. They also o e ad an ages a di e en de elopmen le els (such as design,
implemen a ion and communica ion) enabling, i used, a mo e modula , well- ounded and iche
ep esen a ion o he biological knowledge. This ep esen a ion will p oduce a mo e e icien
knowledge managemen in he long e m.
om 10 h Bio-On ologies Special In e es G oup Wo kshop 2007. Ten yea s pas and looking o he u u e
Vienna, Aus ia. 20 July 2007
Published: 29 Ap il 2008
BMC Bioin o ma ics 2008, 9(Suppl 5):S1 doi:10.1186/1471-2105-9-S5-S1
<supplemen > < i le> <p>P oceedings o he 10<sup> h</sup> Bio-On ologies Special In e es G oup Wo kshop 2007. Ten yea s pas and looking o he u u e</p> </ i le> <edi o >Phillip Lo d, Robe S e ens, Susanna-Assun a Sansone</edi o > <no e>P oceedings</no e> </supplemen >
This a icle is a ailable om: h p://www.biomedcen al.com/1471-2105/9/S5/S1
© 2008 A angu en 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/2.0),
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 ci ed.
BMC Bioin o ma ics 2008, 9(Suppl 5):S1 h p://www.biomedcen al.com/1471-2105/9/S5/S1
Page 2 o 13
(page numbe no o ci a ion pu poses)
Backg ound
On ologies a e enginee ing a e ac s ha can o mally ep-
esen he concep s and hei ela ionships wi hin a gi en
knowledge domain. They can p o ide a compu a ionally
p ocessable concep ual ep esen a ion o ou cu en
unde s anding o eali y, as desc ibed wi hin he in o ma-
ion we hold. Bio-on ologies (on ologies ha ep esen
concep s om li e sciences and, in pa icula , om molec-
ula biology) a e becoming inc easingly impo an [1].
Bio-on ologies play a cen al ole in bioin o ma ics: hey
ac as knowledge bases, da abase in eg a o s, sha ed
ocabula ies, and mo e [1]. Many bio-on ologies a e
a ailable h ough he Open Biomedical On ologies
(OBO) p ojec [2], wi h he Gene On ology (GO) [3]
being he mos impo an example.
Bio-on ologies a e implemen ed in di e en Knowledge
Rep esen a ion (KR) languages, di e ing in p ope ies
ha can be desc ibed along he ollowing axes:
• Syn ax: wha cons i u es a well o med s a emen .
• Seman ics: wha well o med s a emen s mean, o en
de ined as he se o conc e e si ua ions (models) ha a e
consis en wi h a sen ence o se o sen ences.
• Exp essi eness: abili y o he language o dis inguish di -
e en kinds o conc e e si ua ions—some hing ha can
be called “p ecision”.
• Reasoning: answe ing some seman ic based que y, such
as de e mining i one s a emen ollows om ano he .
Reasoning is pe o med by a p og am called a “ easone ”.
The mos used KR languages in bioin o ma ics a e OBO
[4] and/o OWL [5]. OWL has h ee sub-languages,
depending on he exp essi i y: OWL-Li e, OWL-DL and
OWL-Full. OWL-Full is he mos exp essi e ype, and ea-
soning esul s a e no wa an ed. The exp essi eness o a
KR language can be exploi ed o p oduce ich bio-on olo-
gies, ha is, bio-on ologies ha ep esen he knowledge
mos accu a ely, p ecisely and comp ehensi ely, wi h he
highes possible esolu ion. Rich bio-on ologies a e ame-
nable o mo e di e se in e ac ions wi h biologis s, o
example when que ying. A ich bio-on ology can also
acili a e mo e in e es ing easoning, o example o
ob ain new hypo heses om biological knowledge. P es-
en ly, howe e , bio-on ologies mainly end o be lean, as
opposed o ich, due o he gap be ween he po en ial o
KR echniques and hei ac ual implemen a ion in bio-
on ologies. Mos bio-on ologies do no come close o
using all he exp essi eness o he selec ed KR language
[6], e en i ha language has limi a ions in i s abili y o
ully desc ibe he biological domain knowledge [7]. As a
esul , only a limi ed pa o he domain knowledge is cap-
u ed.
Ano he p oblem wi h cu en bio-on ologies is he lack
o igou (use o s ic , explici and well de ined seman-
ics). Rigou ensu es a sound s uc u e and hence a mo e
obus de elopmen and main enance o e ime. Despi e
e o s o imp o e he igou o some bio-on ologies [8-
10], igo ous modelling is no gene al p ac ice wi hin bio-
on ologies.
The modelling e o equi ed o ob aining a ich and ig-
o ous bio-on ology is usually oo demanding o many
bio-on ologis s, as hey a e usually biologis s wi h a lim-
i ed aining in ei he on ology de elopmen o he KR
language used o he on ology's ep esen a ion. I , how-
e e , we a e o imp o e he knowledge managemen in
bioin o ma ics and mo e om lean o ich bio-on olo-
gies, he bio-on ologies mus be buil by expe biologis s
who eally know he i al sub le ies o he knowledge
domain. This ension be ween modelling bes p ac ice
and modelling skills [11] is a undamen al ba ie o
p og ess in bio-on ologies, as he bio-on ologis s only
a ely use he whole powe o KR languages.
One way o help bio-on ologis s o model in a ich and
igo ous manne is o p o ide hem wi h “cookbook eci-
pes” named On ology Design Pa e ns (ODPs). ODPs a e
a de elopmen pa adigm analogous o So wa e Design
Pa e ns (SDPs) [12], widely used in OOP. A SDP is a
p o en solu ion o a known modelling p oblem ha
epea edly appea s when designing di e en so wa e sys-
ems. Mo eo e , SDPs o e an “o he shel ” solu ion o
he p og amme : o example, in he case o he Model-
View-Con olle SDP a me hod o implemen ing g aphi-
cal in e aces is p o ided. We p opose ha ODPs o e
simila ad an ages o he bio-on ologis s.
S uc u es simila o ODPs ha e al eady been used in
on ologies and appea in he li e a u e. The e a e, how-
e e , s ill open issues, such as documen a ion, ep esen a-
ion, applica ion me hods, de ec ion o applica ion
a ge s, e c. In addi ion, whole a eas o biological knowl-
edge lack ODPs. The wo k p esen ed he e begins o ackle
hose issues by p o iding a de ini ion and classi ica ion o
ODPs, me hodologies o spo ing applica ion a ge s o
ODPs in bio-on ologies, me hodologies o applying
ODPs, a documen a ion sys em and an ODPs public ca -
alogue [13]. Some examples o ODPs ha ha e been used
on he Cell Cycle On ology (CCO) a e p esen ed as use
cases (Sequence ODP [14] and Uppe Le el On ology
ODP [15]).
BMC Bioin o ma ics 2008, 9(Suppl 5):S1 h p://www.biomedcen al.com/1471-2105/9/S5/S1
Page 3 o 13
(page numbe no o ci a ion pu poses)
Resul s
De ini ion and classi ica ion o ODPs
ODPs a e solu ions o modelling p oblems ha help he
bio-on ologis o be e use he exp essi i y and igou o
he KR language o choice. ODPs a e examples o solu-
ions, a he han abs ac solu ions ha a e ins an ia ed
in di e en sys ems, unlike SDPs. Thus, he bio-on ologis
uses he ODP as a guide and is able o ec ea e he ODP in
he conc e e bio-on ology ha i is being buil .
ODPs a e used as samples o knowledge. Fo example, a
bio-on ologis may wan o model biological egula ion,
which can only be posi i e o nega i e. Wha cons uc s
does OWL, o example, o e o c ea e such a model o
egula ion and how can he bio-on ologis combine
hem? The answe is o use he Value Pa i ion ODP [16]
as a sample (Figu e 1; o he UML o OWL mapping used
in Figu es 1, 2, 6, 7, 8, 9 and 10, see Figu e 4). The Value
Pa i ion ODP consis s o a co e ing axiom and disjoin
axioms ha allow he alues a pa ame e may ake o be
cap u ed p ecisely. Fo example, a pe son can only be all
o sho , bu no bo h; he Value Pa i ion ODP also
makes he p ope y by which an objec ‘bea s’ he alue
unc ional—so an objec only has ha p ope y once.
Since egula ion can only be posi i e o nega i e (in his
iew o he wo ld), his Value Pa i ion ODP should be
used (Figu e 2).
ODPs a e in p inciple abs ac and implemen a ion inde-
penden . We ocus on OWL o p o ide a amewo k o
di ec implemen a ion, adequa e exp essi i y and ease o
sha ing. ODPs could be desc ibed in a mo e abs ac o -
malism (such as Fi s O de Logic) bu ha would
dec ease usabili y. ODPs can be classi ied acco ding o
hei complexi y, g anula i y, usabili y, popula i y, e c.
He e, we classi y hem acco ding o he way hey a e used:
• Ex ensional ODPs: ODPs ha p o ide ways o ex ending
he limi s o he chosen KR language. Some ODPs can be
used o o e come hose limi a ions and p esen a sui able
ep esen a ion o he knowledge domain ha needs o be
cap u ed. Fo example OWL canno be used o exp ess
excep ions [7,17] o n-a y ela ionships [7], and he e a e
ODPs o wo k a ound hose limi a ions (Excep ion ODP
[18] and N-a y Rela ionship ODP [19]).
• Good p ac ice ODPs: ODPs ha a e used o ensu e a
modelling good p ac ice. These ODPs a e used o p oduce
mo e modula , e icien and main ainable on ologies,
ackling al eady known pi alls o on ology enginee ing
such as ha d-coding o mul iple subsump ions [20].
Examples include No malisa ion ODP [21], Value Pa i-
ion ODP [16] and Uppe Le el On ology ODP [15].
• Domain Modelling ODPs: ODPs ha a e used o model
a conc e e pa o he knowledge domain. They can be
de ined as “signa u e” ODPs: each knowledge domain has
i s peculia i ies and hese ODPs a e used o model hose
peculia i ies. Biological knowledge some imes di e s
om o he domains because o con ingency, symme y,
di e en le els o complexi y in e ac ing wi h each o he ,
eme gen p ope ies, e c. Examples include Lis ODP [22],
Adap ed SEP ODP [23], Sequence ODP [14] and Species
ODP [24].
Ex ensional and Good P ac ice ODPs a e common o all
on ologies. Domain Modelling ODPs a e mo e speci ic o
he knowledge domain (in his case, biological knowl-
edge), bu hey can also be used in o he domains.
Applying ODPs
An impo an aspec o ODPs is unde s anding when i is
app op ia e o apply a pa icula ODP. Ideally, he si ua-
ion in which i is app op ia e o apply an ODP should be
appa en o a bio-on ologis ; he ODP should be sel
explana o y in e ms o documen a ion ( o example
when explo ing he ODPs public ca alogue p esen ed bel-
low). The bio-on ologis can, howe e , be guided using
compe ency ques ions such as he ones desc ibed in Table
1. These and o he ques ions will be used o o m a deci-
sion ee ha will guide a bio-on ologis owa ds an
app op ia e ODP. These ques ions can be supplemen ed
wi h ma e ial on he ypes o en i y ha can be in ol ed
and he kinds o ela ionships hey ha e wi h o he en i-
ies, e en ually being e ined down o he g anula i y o
seman ics used in he ODP i sel .
Once chosen, he main me hod o applying an ODP is o
ec ea e comple ely o in pa he s uc u e o he example
ODP in he on ology, op ionally eusing (“impo ing” in
OWL pa lance) pa s o he example ODP. The use can be
guided in he p ocess wi h wiza ds, o example using he
wiza ds p o ided by he CO-ODE p ojec [25] o he P o-
égé on ology edi o [26].
Ano he me hod o applying ODPs is o use condi ion
ma ching. The On ology P ocessing Language (OPL) is a
syn ax ha allows condi ions o be de ined o ma ching
classes in an on ology w i en in OWL. The classes
ma ched ha e ans o ma ions applied on hem ha
change axioms o anno a ion alues. Thus OPL can be
used o c ea e ODPs in an on ology, by de ining and ODP
as he changes o be made when a ma ch happens.
The ma ching condi ion can be o wo ypes:
• Syn ac ic condi ion: he condi ion elies on a s ing
alue. Thus, he class name o any anno a ion alue, such
as label o commen , can be used o y o ma ch he con-
BMC Bioin o ma ics 2008, 9(Suppl 5):S1 h p://www.biomedcen al.com/1471-2105/9/S5/S1
Page 4 o 13
(page numbe no o ci a ion pu poses)
di ion. The condi ion can be a gi en alue (e.g. cell di e -
en ia ion) o a egula exp ession (e.g. (.+?)
(di e en ia ion)).
• Seman ic condi ion: he condi ion elies on he seman-
ic s uc u e o he on ology— he ODP is applied o any
classes ma ching a class exp ession. Fo example, a condi-
ion can be de ined so as o ma ch any class ha is sub-
sumed by he class exp ession loca ed_in all
(Ch omosome o (pa _o some Ch omosome)) (in
o he wo ds, any class ha has he exp ession loca ed_in
all (Ch omosome o (pa _o some Ch omosome)) as a
necessa y condi ion). A seman ic condi ion can be as
complex as he use wishes (using any o OWL's exp essi -
i y) as he easone [27,28] will p ocess he on ology and
e ie e any ma ched classes.
OPL is pa ially based on he Manches e OWL Syn ax
[29] and SPARQL [30], and i is a ailable as a s andalone
applica ion [31] o as a P o égé plugin [32]. The OPL
commands a e w i en in a la ile by he use and he
OPL p og am pa ses he ile, selec ing classes and apply-
ing he changes de ined, c ea ing a new on ology. ODPs
can be codi ied in he de ined changes, and human- ead-
able explana ions can be w i en in commen s. Thus, he
ODPs a e s o ed in a la ile o di ec applica ion,
oge he wi h any commen s bio-on ologis s migh ind o
in e es . The e o e, ODPs can be applied a any ime, o
any on ology, by unning he OPL p og am, and a e pe -
sis en ly s o ed (Figu e 3).
Documen ing ODPs
The documen a ion sys em p oposed in his esea ch is
inspi ed by he o iginal SDPs documen a ion sys em [12],
wi h some changes; he basic sys em is essen ially he
same, elying on some p ede e mined sec ions wi h which
each ODP mus be desc ibed; name, s uc u e, e c. In he
case o ODPs he sec ions a e di e en , and some o hem
a e op ional (Table 2). An implemen a ion o he docu-
men a ion sys em is a ailable as an ODPs public ca a-
logue [13]. The ca alogue is di ec ly implemen ed in
OWL: each ODP is desc ibed in an on ology, using anno-
a ion p ope ies o he sec ions ha desc ibe he ODP.
The seman ics o he ODP a e di ec ly exp essed in he
on ology, allowing o impo ing he ODPs and sha ing
he ODPs oge he wi h all he in o ma ion codi ied in he
anno a ion p ope ies. Each On ology is ansla ed o
HTML by an sc ip (OWL2HTML) and he URL o he
on ology is au oma ically gene a ed om he URI o he
on ology. The whole ca alogue can be downloaded [33]
and gene a ed locally om he OWL iles by unning he
OWL2HTML sc ip . The ca alogue is open o sugges ions
and co ec ions, and any use can p opose new ODPs o
be added using he mailing lis s and o ums p o ided by
Sou ce o ge [33].
SDPs and ODPs a e desc ibed in a di e en manne in a
documen a ion sys em. SDPs a e desc ibed wi h UML in a
gene ic manne , and hen he ins ances o he SDP a e
applied in he p og amming language o choice. In con-
UML diag am o he applica ion o he Value Pa i ion ODPFigu e 2
UML diag am o he applica ion o he Value Pa i-
ion ODP. The Value Pa i ion is used o model biological
egula ion, which can only be ei he posi i e o nega i e, by
applying he ODP desc ibed in Figu e 1 o an ac ual bio-
on ology.
S uc u e o he Value Pa i ion ODP, in UMLFigu e 1
S uc u e o he Value Pa i ion ODP, in UML. The
co e ing axiom ( he class pa ame e is equi alen o he
union o classes alue_1 o alue_n) ensu es ha when a
new class is added, i is added as a subclass o he alues;
hus, no new alues can be added.
BMC Bioin o ma ics 2008, 9(Suppl 5):S1 h p://www.biomedcen al.com/1471-2105/9/S5/S1
Page 5 o 13
(page numbe no o ci a ion pu poses)
as , he e is no easy, g aphical, comple e and es ablished
language a la UML o desc ibing ODPs, because he e is
no such a language o KR languages. As a consequence,
ODPs ha e o be desc ibed using ins ances: he model,
a he han being a gene ic s uc u e like in SDPs, is an
ins ance ha implici ly desc ibes he gene ic s uc u e.
The absence o a language ha can desc ibe he same
s uc u e in di e en KR languages makes i e y di icul
o de elop a sui able g aphical ep esen a ion o on olo-
gies. In he case o he ODPs public ca alogue UML has
been chosen in he hope ha be e languages will be
de eloped. Despi e ha ing ce ain ad an ages (s anda d,
al eady widely used and wi h a ailable ooling) and ha -
ing been designed as a gene al pu pose modelling lan-
guage, UML lacks na i e s uc u es o a s aigh o wa d
ep esen a ion o OWL. The e o e, he UML ep esen a-
ion is no compac enough and i is oo complex. The
UML o OWL mapping used in he public ODPs public
ca alogue is he one p oposed in [34] (Figu e 4). OWLViz
[35] is also used o simple subsump ion hie a chies.
Use o ODPs on he Cell Cycle On ology
In he con ex o he FP6 p ojec DIAMONDS [36], an
on ology is being de eloped o ep esen he knowledge
abou he cell cycle [37]. This applica ion on ology, called
Cell Cycle On ology [38], comp ises da a om a numbe
o esou ces such as GO, Rela ions On ology (RO) [39],
In Ac [40], NCBI axonomy [41], UniP o [42] as well as
da a om DIAMONDS pa ne s. The esul ing CCO is
designed o p o ide a iche iew o he cell cycle egula-
o y p ocess, in pa icula by accommoda ing he in insic
dynamics o his p ocess. Fo ha pu pose, h ee majo
componen s a e conside ed: he (pe sis en ) en i y i sel ,
i s spa ial localiza ion, and i s empo al localiza ion. CCO
p o ides a es bed o he de elopmen o new
app oaches and ools necessa y o c ea e a ully- ledged
knowledge base. This knowledge base is expec ed o ena-
ble deploymen o ad anced easoning app oaches o
knowledge disco e y and hypo heses gene a ion. CCO
suppo s ou model o ganisms: human (Homo sapiens),
A abidopsis (A abidopsis haliana), Bake 's yeas (Saccha o-
myces ce e isiae) and Fission yeas (Schizosaccha omyces
pombe). The e is an on ology ile o each o he ou
model o ganisms and he ile is a ailable in se e al o -
ma s [43]: OBO [4], OWL-DL [5], XML, DOT [44] and
GML [45]. P esen ly, CCO holds mo e ha 20,000 con-
cep s (mo e han 1,000 bio-molecules and o e 9,000
in e ac ions) and mo e han 20 ypes o ela ionships. A
p esen , wo ODPs ha e been applied in CCO: he
Sequence ODP and he Uppe Le el On ology ODP.
The Sequence ODP [14] is used in CCO o model he cell
cycle (Figu e 5). The cell cycle is modelled as a sequence
o e en s, s a ing in he phase G1, ollowed by S, G2 and
inally M [46]. Fo he sake o simplici y, only he
desc ibed s eps o a s anda d cell cycle a e conside ed, no
Simple mapping o OWL o UMLFigu e 4
Simple mapping o OWL o UML. No all he possible
OWL axioms a e included. R: p ope y, C: class.
Ex ac o an OPL la ile, o be p ocessed by he OPL p o-g amFigu e 3
Ex ac o an OPL la ile, o be p ocessed by he
OPL p og am. The p og am eads he la ile and pe -
o ms he ac ions in he on ology. The s a emen s end wi h ;
and he commen s (s a ing wi h #) a e no p ocessed, ?x is
equi alen o “any class”. The s a emen s o be p ocessed in
his example a e a SELECT s a emen ollowed by wo
ADD s a emen s. When pa sing, he p og am will selec any
class ha has he alue egula ion in i s label anno a ion
p ope y. The ADD s a emen s a e applied o any ma ching
classes ob ained om he SELECT s a emen . I will add
wo axioms o any ma ching class: he i s axiom se s he
ma ching class o be equi alen o he union o he (al eady
exis ing) classes posi i e and nega i e. The second s a e-
men makes hose classes disjoin . The esul ing s uc u e is
he ec ea ion o he Value Pa i ion ODP.
BMC Bioin o ma ics 2008, 9(Suppl 5):S1 h p://www.biomedcen al.com/1471-2105/9/S5/S1
Page 6 o 13
(page numbe no o ci a ion pu poses)
conside ing he o he s eps ( ha also migh play impo -
an oles) o a ia ions such as endo eduplica ion.
The e a e o he sequences o e en s ha in p inciple can
be modelled in he same manne , such as me abolic pa h-
ways. This ODP is a “ immed down” e sion o ano he
ODP, he Lis ODP [22]. The Lis ODP is a much mo e
complex s uc u e in which he exac o de o i ems is e y
impo an , whe eas in he Sequence ODP he only aspec
modelled is wha happens a e o be o e a gi en e en .
Fo example, he Sequence ODP canno be used o com-
pa e di e en sequences o e en s. The sequence ODP
(Figu e 6) makes use o he ela ionships p ecedes and
p eceded_by om RO, bo h being ansi i e. I also uses
wo ela ionships no p esen in RO, namely
immedia ely_p ecedes (subp ope y o p ecedes) and
immedia ely_p eceded_by (subp ope y o
p eceded_by), bo h being unc ional.
When he ODP is applied o CCO (Figu e 7), each phase
o he cell cycle is immedia ely_p eceded_by a phase and
immedia ely_p ecedes ano he phase, only one in bo h
cases, due o he ac ha immedia ely_p eceded_by and
immedia ely_p ecedes a e unc ional. Any phase ha is
immedia ely_p eceded_by one phase is also assumed o
be p eceded_by he same phase, because p eceded_by is
a supe p ope y o immedia ely_p eceded_by. The same
applies o immedia ely_p ecedes and p ecedes.
The use o he Sequence ODP allows o do lexible que ies
agains he on ology. Fo example, i a gi en in e ac ion
occu s a M, and a que y is de ined o e ie e any hing
ha happens a e S (Figu e 8), a easone will e ie e he
in e ac ion (and any in e ac ion occu ing a G2, as bo h
G2 and M a e p eceded by S). This is due o he ansi i i y
o p eceded_by, which is assumed o ela e he pe inen
phases by he easone (e en i i is no explici ly s a ed in
he on ology) because i is he supe p ope y o he ac ual
p ope y ha has been used o asse he ela ionship in
he model, immedia ely-p eceded_by. Howe e , i he
use is only in e es ed in some hing happening jus a e S
(G2 bu no M), immedia ely_p eceded_by should be
used ins ead.
The Uppe Le el On ology ODP [15] (Figu e 9) can be
used o acili a e modelling h ough i s basic on ological
dis inc ions. A p inciple applica ion o uppe le el on ol-
ogies is o in eg a e di e en on ologies. This can be done
because an uppe le el on ology makes dis inc ions
be ween classes, independen o any pa icula domain:
he classes in i ep esen ypes o concep s, such as phys-
ical en i y, p ocess, e c. Fo example, i an on ology ha
desc ibes p ocesses needs o be in eg a ed, i can be done
so unde he class p ocess. The classes o he uppe le el
on ology a e gene ally c ea ed acco ding o philosophical
c i e ia such as con inuan s s. occu en s. The e o e, he
use o an uppe le el on ology is con o e sial, because
he e a e many la ou s o philosophical app oach and
he bio-on ologis may ollow a pa icula iew o he
wo ld ha will highly in luence he s uc u e o he bio-
on ology. In he case o CCO an uppe le el on ology has
been c ea ed (Figu e 10) o include classes om o he
on ologies such as he whole Cell Cycle subon ology
om GO. The use o an uppe le el on ology also helps o
ensu e a good modelling p ac ice, as di e en kinds o
classes (p ocesses, molecules) a e c ea ed in sepa a e dis-
join sub ees, esul ing in a cleane model.
Discussion
Figu e 11 shows a simpli ied o e iew o p io a emp s o
p o ide solu ions simila o ODPs; o he sake o b e i y,
he “W3C Bes P ac ices” a e assumed o be equi alen o
he pa e ns desc ibed in [47-57] (see bellow).
E en i ODPs ha e al eady been documen ed in he li e -
a u e, hey ha e no been explici ly men ioned as such
un il ecen ly [58-60]. In [60] hey a e men ioned as pa
o some on ology building me hodologies, wi hou u -
he analysis such as documen a ion and applica ion. The
idea o CODePs (Concep ual Design Pa e ns) [58,59] is
close o ODPs, bu hey di e in he le el o g anula i y o
Simpli ied model o cell cycleFigu e 5
Simpli ied model o cell cycle. This simpli ied iew o he
cell cycle is assumed when modelling i using he Sequence
ODP. The model, howe e , su ices o ep esen many ac s
abou he cell cycle.
BMC Bioin o ma ics 2008, 9(Suppl 5):S1 h p://www.biomedcen al.com/1471-2105/9/S5/S1
Page 7 o 13
(page numbe no o ci a ion pu poses)
he p oposed solu ion; CODePs a e necessa ily less ine
g ained han ODPs, as hey ep esen concep ual and gen-
e al solu ions, whe eas ODPs o e solu ions in a gi en KR
language wi h ull seman ic co e age. We p opose ha
CODePs and ODPs a e complemen a y: a CODeP will
inca na e i sel in an ODP ha will show he bio-on olo-
gis how o implemen he CODeP in a conc e e KR lan-
guage, much as happens wi h he CODeP Desc ip ion-
Rei ica ion and he N-a y Rela ionship ODP [19] in OWL.
In summa y, he applica ion p ocedu e, documen a ion
sys em and ep esen a ion o ODPs and CODePs a e di -
e en . Knowledge Pa e ns [61] a e concep ual gene al
pa e ns ha a e “mo phed” in o a gi en knowledge base
by a se o mapping axioms. Thus, he knowledge pa e n
can no be applied di ec ly by he bio-on ologis : his is a
d awback since he applica ion o he pa e n needs o be
as in ui i e as possible. The same a gumen applies o he
Seman ic Pa e ns [62]. The ODPs p esen ed he ein a e
eal solu ions o biological knowledge modelling p ob-
lems, a he han heo e ical p oposi ions o gene al pa -
e ns; he alue o hese ODPs is ha hey a e eady o be
used by bio-on ologis s, wi hou any mo phing axiom.
ODPs a e p esen ed in OWL o make ull use o he lan-
guage's seman ics. Those seman ics can be mapped o
o he languages o in e ope abili y ( o example is ela-
i ely easy o map om OWL o OBO [63-66]), bu he
opposi e does no o en happen: i is di icul o bio-
on ologis s, gi en a pa e n in an abs ac o malism, o
apply ha pa e n o an ac ual bio-on ology wi h a con-
c e e KR language.
Some a emp s ha e been made o p o ide bes p ac ice
guidelines in on ology enginee ing and KR, which in
some cases a e seman ically equi alen o ODPs. Some o
hose e o s ha e been collec ed (albei no as a sys ema-
ized collec ion) in he W3C Seman ic Web Bes P ac ices
and Deploymen Wo king G oup web [67]. O he e o s
ha e been published as sel -con ained pa e ns in single
publica ions, ega ding pa onomy [47,48], ansi i e
p opaga ion [49-52,68], on ology le el [53,54] and mul i-
on ology le el bes p ac ices [55,56], and g anula i y [57],
o men ion some ep esen a i e examples. In all cases,
documen a ion, g aphical ep esen a ion and applica ion
me hodologies as such ha e no been add essed in de ail;
a bes , hey we e only implici ly and pa ially used. Some
o hose ODPs a e collec ed in he ODPs public ca alogue
[13].
The use o ODPs will mos likely gi e se e al ad an ages
o bio-on ologis s when c ea ing and main aining bio-
on ologies. The ollowing ad an ages ha e no been ho -
oughly es ed, and he e o e he e is no expe imen al e i-
UML diag am o he Sequence ODPFigu e 6
UML diag am o he Sequence ODP
BMC Bioin o ma ics 2008, 9(Suppl 5):S1 h p://www.biomedcen al.com/1471-2105/9/S5/S1
Page 8 o 13
(page numbe no o ci a ion pu poses)
dence o hem, bu hey a e easonable assump ions
based in he au ho s' expe ience in bio-on ology enginee -
ing. The ad an ages a e di ided in h ee a eas:
1. Design (seman ics, modelling):
• Rich and g anula modelling. ODPs should acili a e he
p oduc ion o mo e ichly axioma ised on ologies by
allowing a mo e ine-g ained modelling o he knowledge
domain. They should help in making he implici knowl-
edge ound, o example, in e m names, explici , encod-
ing i in he seman ics o he on ology. Addi ionally, bio-
on ologies a e deepening he knowledge hey model, and
ODPs o ep esen ha deepe knowledge wi h he sui a-
ble g anula i y a e needed.
• Seman ic encapsula ion. ODPs p o ide an easy way o
dealing wi h he complexi y o seman ics in concep ual
modelling by encapsula ing i in he ODP.
• Robus ness and modula i y. Some ODPs help in c ea -
ing mo e obus and modula on ologies.
• Reasoning. The iche axioms needed o e icien and
p oduc i e easoning should be eached mo e easily using
ODPs. The e o e, as mo e axioms a e placed in he on ol-
ogy mo e sophis ica ed in e ences can be unde aken.
• Alignmen . Mo e and mo e on ologies a e being de el-
oped and e icien ways o compa ing/aligning hem a e
UML diag am o a que y ha demons a es he u ili y o he Sequence ODPFigu e 8
UML diag am o a que y ha demons a es he u il-
i y o he Sequence ODP. The OWL-DL que y
occu s_a some (p eceded_by some S) e u ns any
in e ac ion ha occu s a e S (G2 and M). Howe e , i a
use is only in e es ed in any hing occu ing immedia ely
a e S (G2 bu no M) immedia ely_p eceded_by should
be used: occu s_a some (immedia ely_p eceded_by
some S).
UML diag am o he applica ion o he Sequence ODP o CCOFigu e 7
UML diag am o he applica ion o he Sequence ODP o CCO. The cell cycle is de ined as a sequence o phases ha
happen one a e he o he , using he ela ionships immedia ely_p eceded_by and immedia ely_p ecedes.
BMC Bioin o ma ics 2008, 9(Suppl 5):S1 h p://www.biomedcen al.com/1471-2105/9/S5/S1
Page 9 o 13
(page numbe no o ci a ion pu poses)
consequen ly necessa y. The consis ency o modelling
inhe en in he use o ODPs should suppo seman ic
ma ching be ween di e en on ologies.
2. Implemen a ion (ac ual de elopmen o he on ology):
• Focused de elopmen . Ha ing an ODP as an enginee -
ing a e ac should educe he de elopmen ime, so ha
he domain expe can be ocused on he modelling
de ails o he speci ic a ea ha is being modelled.
• Tooling. ODPs can be codi ied p og amma ically, p o-
iding ools ha can au oma ically build sec o s o an
on ology ha a e complex o egula . The same ools
could also guide he on ologis in he p ocess o building
on ologies.
• Rapid p o o yping. ODPs a e ideal o apidly de elop-
ing p o o ypes. Ha ing p o o ypes should allow de elop-
e s o discuss comple e models o on ologies in ea ly
s ages and hence make mo e sound on ologies. I should
also allow as e de elopmen .
• Re-enginee ing. ODPs could be applied in he beginning
o an on ology de elopmen p ocess as well as du ing he
li e cycle o i , p o iding, o ins ance, aluable insigh s o
e ac o ing some componen s which may hold an incon-
sis ency o which may iola e design p inciples.
3. Communica ion:
• Good communica ion. The use o ODPs should
imp o e communica ion be ween on ology de elope s.
The de elope s could easily ecognize he di e en ea-
u es o he on ology p oduced by he ODP, as i ep e-
sen s a well known and ho oughly documen ed
abs ac ion.
• Documen ed modeling. When c ea ing on ologies he
p ocess should be mo e p ecisely documen ed by simply
men ioning which ODPs we e used. As a esul , he design
decisions would become explici .
• Comp ehension o ad ances in KR. KR languages a e
e ol ing as ( o example OWL 1.1 [69]) and i is usually
di icul o unde s and he new ea u es o he languages:
by p o iding ODPs i should become much easie , as
ODPs a e examples o how o use he new ea u es.
UML diag am o he Uppe Le el On ology ODP, as applied in CCOFigu e 10
UML diag am o he Uppe Le el On ology ODP, as applied in CCO. Disjoin axioms no shown.
UML diag am o he Uppe Le el On ology ODPFigu e 9
UML diag am o he Uppe Le el On ology ODP
