1
SCIENTIFIC REPORTS | (2019) 9:179 | DOI:10.1038/s41598-018-37411-y
www.na u e.com/scien i ic epo s
INSIdE NANO: a sys ems biology
amewo k o con ex ualize he
mechanism-o -ac ion o enginee ed
nanoma e ials
Angela Se a1,2,3, I ica Le unic4, Vi o io Fo ino
2,3,5,6, Richa d D. Handy7, Beng Fadeel8,
Robe o Taglia e i1 & Da io G eco
2,3,5
Enginee ed nanoma e ials (ENMs) a e widely p esen in ou daily li es. Despi e he e o s o
cha ac e ize hei mechanism o ac ion in mul iple species, hei possible implica ions in human
pa hologies a e s ill no ully unde s ood. He e we pe o med an in eg a ed analysis o he e ec s o
ENMs on human heal h by con ex ualizing hei ansc ip ional mechanism-o -ac ion wi h espec o
d ugs, chemicals and diseases. We buil a ne wo k o in e ac ions o o e 3,000 biological en i ies and
de eloped a no el compu a ional ool, INSIdE NANO, o in e new knowledge abou ENM beha io .
We highligh s iking associa ion o me al and me al-oxide nanopa icles and majo neu odegene a i e
diso de s. Ou no el s a egy opens possibili ies o achie e as and accu a e ead-ac oss e alua ion o
ENMs and o he chemicals based on hei biosigna u es.
ENMs al eady pe ade ou e e yday li es, being p esen in nume ous consume p oduc s, and new nanoma-
e ials a e being p oduced a an e e -inc easing pace. Howe e , despi e conside able ad ances in he pas dec-
ade, we a e s ill a om a comp ehensi e unde s anding o he biological e ec s o he my iads o exis ing and
eme ging ENMs1,2. Global omics echnologies may aid in cha ac e izing he mechanism-o -ac ion (MOA) o
ENMs, opening new possibili ies o nex gene a ion sa e y assessmen based on sys ems biology app oaches3.
An eme ging s a egy in isk assessmen o chemicals is ead-ac oss analysis, unde he assump ion ha s uc-
u ally simila compounds exe compa able biological e ec s. To da e, only a ew ead-ac oss analyses ha e
been p oposed o ENMs due o he limi ed possibili y o compu a ionally de i e hei physical-chemical p op-
e ies, o hei molecula size and complexi y. Mo eo e , only ma ginal a emp s ha e been made o in eg a e
MOA signa u es in ead-ac oss, e en when e alua ing s uc u ally smalle and simple compounds. Howe e ,
he no ion ha any pheno ypic pe u ba ion p oduces a speci ic pa e n o molecula al e a ions ha can be used
as i s signa u e is well es ablished, o ins ance, in s udies o d ug eposi ioning4–6. Based on he hypo hesis ha
an e ec i e d ug should be able o coun e balance he pe u ba ions caused by a disease, co ela ions be ween
disease- and d ug-associa ed gene exp ession signa u es ha e been sough in a emp s o eposi ioning d ug mol-
ecules7. In e es ingly, he biological e ec s o chemicals ha e no ye been exploi ed in a sys ema ic ela ionship
wi h he molecula signa u es o human diseases, which in u n could add signi ican amoun o in o ma ion
o he ead-ac oss e alua ion. He e, we hypo hesized ha sys ema ic analysis o ansc ip ional mechanism o
ac ion ( MOA) signa u es could be used o con ex ualize o ‘posi ion’ ENMs wi h espec o human diseases, d ug
ea men s, and chemical exposu es. This s a egy could mi iga e he cu en limi a ion o in o ma ion a ailable
conce ning ENMs e ec s. Mo eo e , knowledge on he molecula e ec s o ENMs could be also used o iden i y
ad e se ou come pa hways ha may lead o pa hogenesis, o indeed MOA o ENMs ha acili a e hei applica-
ion as po en ial ea men s. To allow o sys ema ic con ex ualiza ion o he e ec s o ENMs, we de eloped he
1NeuRoNe Lab, DISA-MIS, Uni e si y o Sale no, Sale no, I aly. 2Facul y o Medicine and Li e Sciences, Uni e si y
o Tampe e, Tampe e, Finland. 3Ins i u e o Biosciences and Medical Technologies, Uni e si y o Tampe e, Tampe e,
Finland. 4BioBy e Solu ions GmbH, Heidelbe g, Ge many. 5Ins i u e o Bio echnology, Uni e si y o Helsinki,
Helsinki, Finland. 6Biomedicine Ins i u e, Uni e si y o Eas e n Finland, Kuopio, Finland. 7School o Biological
and Ma ine Sciences, Uni e si y o Plymou h, Plymou h, Uni ed Kingdom. 8Ins i u e o En i onmen al Medicine,
Ka olinska Ins i u e , S ockholm, Sweden. Co espondence and eques s o ma e ials should be add essed o D.G.
(email: da io.g eco@s a .u a. i)
Recei ed: 14 June 2018
Accep ed: 30 No embe 2018
Published: xx xx xxxx
OPEN
www.na u e.com/scien i ic epo s/
2
SCIENTIFIC REPORTS | (2019) 9:179 | DOI:10.1038/s41598-018-37411-y
compu a ional ool INSIdE NANO (‘In eg a ed Ne wo k o Sys ems bIology E ec s o NANOma e ials’, a ailable
a h p://inano.bioby e.de, and b ie ly desc ibed in me hods sec ion). To his end, we de i ed, om he scien i ic
li e a u e o om he analysis o a ailable ansc ip omics da a, speci ic MOA signa u es o a la ge se o human
diseases ( he ull lis is epo ed in Da aS1), chemicals (Compa a i e Toxicogenomics Da abase - CTD8, he ull
lis is epo ed in Da aS2), FDA-app o ed d ugs (Connec i i y Map Da abase - Cmap9, he ull lis is epo ed in
Da aS3), and ENMs (NanoMine 10 - he ull lis is epo ed in Da aS4). Gene exp ession da a o ENMs exposu e
analyses we e e ie ed om NanoMine , a public ansc ip omics da abase encompassing in i o ansc ip om-
ics p o iles ob ained in human cells o cell lines o a panel o ENMs. See Supplemen a y Ma e ials and Fig.S1
o de ails on inpu da a and p ep ocessing. We hen compu ed he deg ee o simila i y be ween all he pai s o
biological en i ies p esen in his in eg a ed da a se based on he simila i y o hei MOA signa u es. In pa icu-
la : (i) he Jacca d index was used o compu e pai wise simila i y be ween gene se s; (ii) he Kendall Tau dis ance
was used o compu e simila i y be ween anked lis s o genes; (iii) he Gene Se En ichmen Analysis (GSEA) was
used o compu e simila i ies be ween anked lis o genes and gene se s. We hen used his in o ma ion o build
a la ge ne wo k o 3,516 nodes (pheno ypes) in e connec ed by 12,362,256 edges. The wo k- low o he analysis,
he da abase a chi ec u e and he da a in eg a ion s a egy a e schema ically shown in Figs1 and 2 and desc ibed
in de ails in he me hod sec ion.
Resul s
De ini ion o he INSIdE NANO pheno ypic ne wo k. We in eg a ed MOA signa u es o ou ypes
o pheno ypic en i ies (ENMs, d ugs, human diseases and chemical subs ances), ei he de i ed om de no o
ansc ip omics da a analysis o om scien i ic da abases. We s udied he pa e ns o simila i y o hese MOA
signa u es, and used hem o p edic he biological e ec s o ENMs. We de ined a lis o associa ed genes o each
pheno ypic en i y o be i s MOA signa u e. In ou analysis, MOA o ENMs and d ugs a e ep esen ed by o de ed
lis s o genes anked by hei di e en ial exp ession alues. Fu he mo e, MOA o chemicals and molecula
Figu e 1. INSIdE NANO wo k low. T ansc ip omics da a (ENMs (n = 28) and d ugs (n = 615)) and
p ecompiled lis s o associa ed genes (Human Diseases (n = 585) and Chemicals (n = 2288)) we e e ie ed
om mul iple sou ces (A). MOA signa u es we e de i ed o each pheno ypic en i y in o m o gene anks o
ENMs and D ugs exposu e and gene se s o human diseases and chemical exposu es (B). MOA based pa iwise
simila i y we e compu ed (C). Pai wise simila i ies we e used o in e a weigh ed ne wo k o pheno ypic en i ies
(D). Cliques and hei associa ed lis o genes unde lying he connec ions we e iden i ied (E). INSIdE NANO
achie es con ex ualiza ion o ENM MOA and o pe o m MOA-based ead-ac oss analysis (F).
www.na u e.com/scien i ic epo s/
3
SCIENTIFIC REPORTS | (2019) 9:179 | DOI:10.1038/s41598-018-37411-y
al e a ions o diseases a e ep esen ed by se s o associa ed genes e ie ed om he Compa a i e Toxicogenomics
Da abase (CTD) (Figs1A and 2a and me hod sec ion o a desc ip ion o he inpu da a). We hypo hesized
ha he ela edness o each pai o pe u ba ions (ENMs, d ugs, chemicals and diseases) can be quan i ied as
he deg ee o simila i y be ween hei speci ic MOA pa e ns. Following da a homogenisa ion (Fig.1B,C, and
me hod sec ion o a desc ip ion o he simila i y measu es), he in eg a ed pai wise simila i y ma ix was used
as an adjacency ma ix o cons uc a weigh ed undi ec ed in e ac ion ne wo k, which we called INSIdE NANO,
whe e he nodes a e he pheno ypes (ENMs, d ugs, diseases and chemicals) and he MOA simila i ies be ween
hem ep esen he edge weigh s. We also e ained he in o ma ion on he di ec ion o he simila i ies (posi i e o
nega i e), so ha he edges in he ne wo k ha e a sign a ibu e indica ing i he MOA signa u es o wo nodes a e
conco dan ( he genes a e al e ed in he same di ec ion by bo h he pe u ba ions) o disco dan ( he genes a e
al e ed in he opposi e di ec ion by he wo pe u ba ions). See Figs1D and 2a and me hod sec ion o a desc ip-
ion o he ne wo k in e ence p ocess.
MOA signa u es mi o chemically, biologically and clinically ele an pa e ns. One o he
ac o s p e en ing omics echnologies om being ully in eg a ed in egula o y assessmen o chemicals is he
“noisy” na u e o he MOA signa u es usually de i ed om hese high-con en assays. We hus es ed he hypo h-
esis ha ou compu a ional amewo k, in e ing simila i ies be ween pheno ypic en i ies om hei MOA sig-
na u es, can also highligh obus in o ma ion ha co esponds o ei he s uc u ally d i en (as implemen ed in
cu en ly es ablished ead-ac oss me hods) o clinically ele an pa e ns o simila i y. To his end, we sys em-
a ically compu ed pai wise simila i y ma ices be ween he se s o pheno ypic en i ies p esen in ou analysis
and independen da a se s conce ning o he ele an aspec s un ela ed om hei molecula e ec s (Figs1C
and 2a). Nex , we assessed he co ela ion be ween hese simila i y pa e ns and hose de i ed om ou in eg a-
i e MOA analysis (Table1). See sec ion me hod o mo e de ails. We indeed con i med ha ou MOA-based
simila i ies signi ican ly esembled hose compu ed by conside ing independen cha ac e is ics, such as he 2D
molecula s uc u e o he d ugs (Man el’s es P < 0.01) and chemicals (Man el’s es P < 1E − 05), espec i ely.
In addi ion, ou MOA-de i ed ela edness o d ugs could also success ully ecapi ula e hei analogy based on
known molecula a ge s (Man el’s es P < 1E − 05). In e es ingly, also s uc u al simila i ies be ween d ugs and
chemicals we e signi ican ly simila o hose compu ed om MOA signa u es, al hough de i ed om di e en
Figu e 2. INSIdE NANO da a and a chi ec u e. The pheno ypic en i ies in he disco e y da a se s we e
in eg a ed o pe o m ENMs con ex ualiza ion. The INSIdE NANO ne wo k con ains 28 ENMs, 615 d ugs, 585
human diseases and 2288 chemicals connec ed by 12,362,256 edges. The weigh on he edges a e p opo ional
o he s eng h o simila i y be ween he en i ies. This simila i y was compu ed by means o di e en me ics:
he Kendall Tau dis ance was used o compu e simila i ies be ween he anked lis o genes associa e o he
ENMs and d ugs; he Jacca d Index was used o compu e simila i ies be ween he se s o genes associa ed o
Chemicals and Diseases; he Gene Se s En ichmen Analysis (GSEA) was used o compu e simila i ies be ween
he anked lis o genes associa ed o he ENMs and D ugs and he se s o genes associa ed o chemicals and
diseases (a). Da a se s used o alida e he connec ions in e ed in he INSIdE NANO ne wo k. The simila i y
be ween he en i ies based on he molecula al e a ion p o iles we e alida ed by compa ing i wi h al eady
compu ed simila i y measu es un ela ed om he molecula al e a ions. D ugs simila i ies we e compa ed wi h
smiles and a ge based simila i ies. Diseases simila i ies based on symp om we e compu ed, while chemicals
simila i ies a e compu ed using smiles. D ugs and diseases simila i ies we e compu ed based on p esc ip ion
in o ma ion downloaded om he MEDI da abase. D ugs chemicals similila i ies we e based on smiles and
disease chemicals simila i ies we e download om he CTD da abase (b).
www.na u e.com/scien i ic epo s/
4
SCIENTIFIC REPORTS | (2019) 9:179 | DOI:10.1038/s41598-018-37411-y
da a sou ces (Man el’s es P < 1E − 04). Simila ly, we obse ed subs an ial consis ency o ou disease-disease
simila i ies based on pa e ns o molecula al e a ion wi h hose calcula ed by aking in o accoun he o e lap-
ping clinical symp oms (Man el’s es P < 1E − 05). Fu he mo e, ou in e ence was subs an ially cohe en o he
known d ug- o-disease ela ionships based on he use o speci ic d ugs o ea ce ain diseases in clinical p ac ice
(Man el’s es P < 1E − 04) as well as known chemical- o-disease connec ions based on epidemiological causal
e idence o he pa hogenic e ec s o exposu es (Man el’s es P < 1E − 05). Taken oge he , hese esul s s ongly
suppo ha ou s a egy o da a in eg a ion and homogeniza ion is obus and allows highligh ing meaning ul
ela ionships be ween pheno ypic en i ies o di e en ypes.
Ex apola ion o pheno ypic cliques e eals connec ions be ween ENMs and espi a o y and
de mal diseases. G aphs (o ne wo ks) can e icien ly ep esen complex phenomena and hey can be ap-
idly analyzed wi h ad hoc algo i hms ha conside he pa e ns o ela edness o hei cons i uen s. We hypo h-
esized ha deg ees o MOA-de i ed simila i y be ween se s o pheno ypes could be used as an indica ion o
biological associa ion. Speci ically, we scanned INSIdE NANO in sea ch o ‘clique’ subne wo ks, i.e., quad uple
s uc u es o he e ogeneous nodes (a disease, a d ug, a chemical and an ENM) comple ely in e connec ed by
s ong pa e ns o simila i y o an i-simila i y (Fig.1E). Mo e de ails on he sea ch algo i hm a e epo ed in
he me hod sec ion and Fig.S2. We could alida e ou p edic ions ela ed o he ela i e p oximi y and con-
nec i i y o pheno ypic en i ies in ou ne wo k agains a se o known associa ions be ween diseases and d ugs
(Kolmogo o -Smi no es , P < 0.002), based on d ug use in clinical p ac ice11,12, and be ween diseases and
chemicals (Kolmogo o -Smi no es , P < 0.001), based on li e a u e analysis. Chemical-disease in e ac ion da a
we e e ie ed om he CTD. Fu he , he lis o he e ogeneous cliques o size h ee and ou was anked o
iden i y he mos obus ones. Fi s ly, since lowe h esholds in he clique sea ch algo i hm deno es highe con-
nec i i y s eng h be ween he nodes, only he cliques iden i ied wi h a h eshold lowe o equal han 0.4 we e
selec ed. We hen ocused ou analysis on he cliques including a leas one known connec ion. A pe mu a ion
es was execu ed (as desc ibed in he me hods sec ion) o asses he signi icance o he subse o cliques. Only
he cliques wi h high connec ion s eng h, a leas one known connec ion, and signi ican p alue (p alue < 0.05)
whe e inally selec ed. We hen ocused on he possible in ol emen o ENMs in he mos obus iden i ied cliques
and in e ed connec ions be ween speci ic ENMs and se e al human diseases, including, o ins ance, condi ions
a ec ing he espi a o y sys em and skin (FigsS3–S7). The la e obse a ions a e s ongly co obo a ed by he
well es ablished no ion in li e a u e abou he pulmona y and de mal e ec s o ce ain ENMs.
Associa ion o me al and me al oxide nanopa icles wi h neu odegene a i e diso de s. Ou
sys ema ic sea ch o cliques highligh ed a subse o in iguing MOA simila i y pa e ns ela ed o h ee impo -
an neu odegene a i e diso de s, i.e., Pa kinson’s disease (PD, Figs3A, S8, Da aS5), Alzheime ’s disease (AD,
Figs3B, S9, Da aS5), and amyo ophic la e al scle osis (ALS, Figs3C, S10, Da aS5). We ocused on he mos
signi ican cliques whe e disease-d ug and disease-chemical associa ions we e al eady known and in es iga ed
he po en ial connec ions o ENMs in his con ex . Ou analysis clea ly poin ed o an associa ion be ween me al
and me al oxide nanopa icles (NP), including ungs en ca bide cobal (WCCo), i anium dioxide (TiO2), zinc
oxide (ZnO), and gold (Au), and neu odegene a i e diso de s (Fig.4A). The neu o oxici y o me als, such as lead,
me cu y, aluminium, cadmium, and a senic, is well known13–15. The e is also some e idence o a ela ionship
be ween inhaled pa icles, e.g., ul a ine pa icle exposu es in ambien ai o a he wo kplace (e.g., me al umes)
and neu o oxici y in humans16–18. We ound WCCo NP o be s ongly associa ed wi h PD (Figs3A and S8),
oge he wi h he neu o oxin 1-me hyl-4-phenylpy idinium (MPP+), which is known o cause PD by des oy-
ing dopamine gic neu ons in he b ain and i s p od ug 1-me hyl-4-phenyl-1,2,3,6- e ahyd opy idine (MPTP).
Mo eo e , he an i-PD d ugs le odopa, dopamine and b omoc ip ine comple ed he PD- ela ed cliques (Figs3A
and S8). WCCo NP a e known o be cy o oxic and geno oxic, and as ocy es cul u ed in i o we e ound o be
he mos sensi i e in a s udy in ol ing a ange o mammalian cell models19. To he bes o ou knowledge, he e
a e no in i o s udies on WCCo e ec s on he CNS. Howe e , u he in es iga ion should add ess he possibili y
ha WCCo NP may be especially ha m ul o he b ain. The po en ial neu o oxici y o TiO2 NP has al eady been
in es iga ed bo h in i o and in i o20,21. TiO2 NP a e easily ansloca ed in o he b ain o exposed mice, ei he ia
he blood-b ain ba ie o he nose-b ain pa h, bu hei elimina ion a e is limi ed, hus esul ing in hei accu-
mula ion and consequen damage o neu ons and glial cells22. I is o in e es o no e ha di e en TiO2 NP a e
INSIdE NANO (MOA) Simila i y By Man el’s Tes P
D ugs - D ugs chemical s uc u es 1E − 02
Chemicals - Chemical chemical s uc u es 1E − 05
D ugs - Chemicals chemical s uc u es 1E − 04
D ugs - D ugs molecula a ge s 1E − 05
Diseases - Diseases symp oms 1E − 05
D ugs - Disease use in clinical p ac ice 1E − 05
Chemicals - Diseases pa hogenic exposu es 1E − 04
Table 1. INSIdE NANO associa ions based on MOA simila i ies. The co ela ions (simila i ies) be ween
ce ain ypes o biological en i ies (in ows) compu ed based on he ansc ip ional mechanism-o -ac ion
( MOA) simila i y we e sys ema ically compa ed o hose calcula ed conside ing independen biochemical
aspec s. Man el’s es P is epo ed, unde he null hypo hesis ha wo compa ed ma ices a e di e en .
www.na u e.com/scien i ic epo s/
5
SCIENTIFIC REPORTS | (2019) 9:179 | DOI:10.1038/s41598-018-37411-y
iden i ied in di e en cliques, sugges ing ha di e ences in ma e ial p ope ies a e associa ed wi h dis inc dis-
o de s (Figs3 and S8–S10). Fo ins ance, he TiO2 nanobel s (NB)23 we e ound o be associa ed wi h ALS, while
sphe ical TiO2 NP (o di e en p ima y pa icle sizes) we e associa ed wi h AD and PD. Die hylene-glycol coa ed
ZnO NP, bu no o he ypes o ZnO NP in eg a ed in INSIdE NANO, we e signi ican ly associa ed wi h bo h PD
(Figs3A and S8) and AD (Figs3B and S9). In a p e ious in i o s udy, a panel o nine ZnO NP we e es ed o
hei cy o oxici y po en ial using he Ju ka leukemic cell line24. Die hylene-glycol-ZnO was ound o be he mos
cy o oxic o all he ZnO nanopa icles es ed and also elici ed he s onges ansc ip omic esponse among he
sc eened nanopa icles21,24,25. In e es ingly, Xie e al. epo ed ha epea ed adminis a ion o ZnO NP elici ed
beha io al and elec ophysiological imp o emen s in a a model o dep ession26. We also obse ed a signi ican
associa ion o Au NP wi h PD (Figs3A and S8) and ALS (Figs3C and S10), a de as a ing neu ological disease
cha ac e ized by he dea h o mo o neu ons. In e es ingly, Au NP ha e been shown o induce oxida i e s ess and
o educe he ac i i y o an ioxidan enzymes in a b ain27. Mo eo e , exposu e o Au NP dec eased he le els o
he neu o ansmi e s dopamine and se o onin. I is pe inen o no e ha gold is widely used o he ea men
o heuma oid a h i is (RA) and ha neu o oxici y has been documen ed in pa ien s wi h RA ecei ing o al o
injec able gold28,29. Whe he o no Au NP also elici simila e ec s is unknown. O he elemen s e ie ed in he
con ex o he Au NP-ALS connec ions using he INSIdE NANO ool included quinidine and py e h in (Figs3C
and S10). Quinidine, in combina ion wi h dex ome ho phan, is used o ea a ec i e diso de s in pa ien s wi h
ALS30. Py e h in, on he o he hand, has insec icidal ac i i y by a ge ing he ne ous sys em o insec s31. Taken
oge he , hese esul s sugges ha INSIdE NANO does no indisc imina ely g oup ENMs based on hei co e
chemis y, and p o ides e idence o he impo ance o o he physicochemical p ope ies, including, in he case
o ZnO NP, he su ace coa ing and a endan a e o pa icle dissolu ion, and, in he case o TiO2 NP, he shape o
aspec - a io o he pa icles, as discussed abo e.
Figu e 3. Signi ican associa ion be ween ENM and neu odegene a i e diseases. Rele an op-10 cliques
including associa ions be ween ENM, chemicals and d ugs MOA wi h Pa kinson’s disease (A), Alzheime ’s
disease (B), amyo ophic la e al scle osis (C). Cliques including a leas one known connec ion be ween disease-
d ug and disease-chemical we e selec ed.
Figu e 4. The cliques including a leas one known connec ion be ween he disease-d ug and disease-chemical
we e selec ed. The numbe o signi ican in e ac ions be ween Pa kinson disease (da k g een), Alzheime
disease ( ed), and amyo ophic la e al scle osis (ligh g een) and each ENM (X-axis) a e depic ed as ba plo (A).
The d ugs included in he signi ican cliques, ca ego ized by he i s le el o hei ATC code, a e shown as ba
plo (B).

www.na u e.com/scien i ic epo s/
6
SCIENTIFIC REPORTS | (2019) 9:179 | DOI:10.1038/s41598-018-37411-y
Toxic e ec s o me al and me al oxide nanopa icles on he cen al ne ous sys em in
i o. Clinical case s udies ha demons a e an associa ion be ween exposu e o ENMs and neu odegene -
a i e diseases in humans a e cu en ly missing. Ine i ably, gi en he la ency o hese diseases i will be some
decades be o e occupa ional heal h da a becomes a ailable om exposu e moni o ing in he wo kplace, o om
adi ional epidemiology in public heal h. Howe e , he undamen al e en s in chemical oxicology ha may
lead o b ain inju y a e known. Figu eS11 shows he key e en s in he ad e se ou come pa hway (AOP) leading
o human disease. The in ol emen o ENMs has been demons a ed in i o o key e en s in he AOP a he
molecula /biochemical, physiological and pa hophysiological le els. The e iology o b ain inju y om ENMs
includes oxida i e s ess, iono egula o y dis u bances, b ain pa hology, and changes in ish beha io , ha can
only be explained by neu ological de ici . The in i o s udies (TableS1) mapped on o he AOP ha e been ca e ully
selec ed o be b ain-speci ic and no caused by seconda y sys emic hypoxia (e.g., om espi a o y dis ess) han
can indi ec ly lead o b ain inju y. Figu eS11 shows ENM in ol emen in mos o he s eps o he AOP, sugges ing
ha he INSIdE NANO p edic ed associa ions be ween me al nanopa icles and neu odegene a i e diso de s a e
ecapi ula ed in an in i o model. Ou analysis also highligh ed key genes, whose exp ession is al e ed by speci ic
me al and me al oxide NP, po en ially in ol ed in media ing he pi o al s eps in he pa hogenesis o Pa kinson’s
disease (TableS2), Alzheime ’s disease (TableS3), and amyo ophic la e al scle osis (TableS4). Taken oge he ,
ou esul s no only a e able o acili a e apid p edic ion o possible implica ions o ENMs exposu e in human
pa hogenesis, bu p o ide also s ong e idence o possible key molecula e en s media ing he ENMs e ec s.
Po en ial applica ion o INSIDE NANO o d ug ( e)posi ioning. D ug-d ug and d ug-disease
MOA-based simila i y pa e ns in e ed in INSIdE NANO signi ican ly mi o ed hose de i ed om chemical
and clinical e idence (Table1), hus sugges ing ha INSIdE NANO could also se e as a disco e y ool o d ug
posi ioning. Along his line, we obse ed ha he d ugs in he signi ican cliques in ol ing neu odegene a i e
diso de s a e known o a ge he ne ous sys em and senso y o gans (Fig.4B). In addi ion, an i-in lamma o y
molecules and d ugs known o exe hei he apeu ic e ec on he ca dio ascula sys em we e also e ie ed in
connec ion o neu odegene a i e diso de s (Fig.4B). We ecen ly desc ibed compu a ional eposi ioning o many
compounds ac ing on he ca dio ascula sys em as neu oac i e d ugs, p obably due o simila molecula s uc-
u e and MOA, which o en a ec s he s abili y o he memb ane po en ial6. Based on hese esul s, i is possible o
a gue ha posi ioning o ENMs o biomedical applica ions is also concei able, using INSIdE NANO.
Discussion
In he pos -genomic e a, omics s udies ha e been ou inely used o add ess a ple ho a o biomedical ques ions
and, consequen ly, eno mous amoun o omics da a and omics-de i ed in o ma ion a e accumula ing. Al hough
he alue o omics sc eenings has been ecognized also in he ield o chemical sa e y, o da e he use o hese ech-
nologies is mainly limi ed o he measu emen s o he p ima y molecula esponses o d ugs o chemicals. This
in o ma ion, in u n, is used o cha ac e ize he MOA du ing exposu es and de ining pa hways o oxici y (PoT)
ha could se e as biological signa u es. Gi en he inc easing amoun o da a ega ding he MOA o d ugs and
chemicals, he nex challenge appea s o be he sys ema ic in eg a ion o hese exposu e-speci ic biological signa-
u es wi h he pa e ns o molecula al e a ion o human diseases. This could g ea ly help he posi ioning o chem-
icals and d ugs as oxican s o he apeu ics o a speci ic disease, and hence p o ide a aluable indica ion in e ms
o haza d assessmen as well as d ug de elopmen . Howe e , lack o s anda diza ion in he compu a ional s a e-
gies and algo i hms used o de i ing and compa ing MOA signa u es has, un il now, p e en ed omics da a om
being ully exploi ed in sa e y assessmen . In his s udy, we assumed ha compa isons o MOA signa u es could
be used o ind obus and meaning ul ela ionships be ween di e en ypes o exposu es and human diseases.
O e all, ou esul s demons a e ha his is indeed possible by in eg a ing di e en ypes o da a, including omics.
Mo eo e , he igo ous alida ions o ou no el da a analysis and in eg a ion me hods sugges ha ou compu a-
ional amewo k could pa e he way o a comple e in eg a ion o omics echnologies in o egula o y ead-ac oss
analysis. Read-ac oss is apidly becoming a s a egic ins umen o mee he inc easing need o pe o m apid
assessmen and labelling o many compounds, including ENMs32,33. This knowledge gap- illing s a egy adi-
ionally consis s o de ining g oups o molecules wi h high s uc u al simila i y, unde he assump ion ha hey
will also exe simila biological e ec s. Cu en ly, ead-ac oss sys ems p esen se e al limi a ions. Fi s , al hough
o he wise en isaged, hey a e usually es ic ed o pa ial chemical spaces consis ing o se s o compounds wi h
ela i ely homogeneous applica ions/e ec s, limi ing hei applicabili y domains. In his con ex , he analysis
o ENMs is hampe ed by he di icul ies o compu a ionally de i e s uc u al desc ip o s o be implemen ed in
ead-ac oss sys ems, and hence only ew s udies limi ed o speci ic classes o ENMs ha e been p oposed hus
a 34–36. Second, excep o a ew aluable a emp s37, ead-ac oss mos ly elies on g ouping ENMs o chemicals
based only on he simila i y o hei molecula s uc u e, neglec ing hei MOA. Thi d, ead-ac oss sys ems so
a wo k on speci ic endpoin s o s ic oxicological in e es ; and do no s i e o he possibili y o di ec ly in e
exposu e-disease ela ionships which could also be used o posi ion an exposu e as d ug38. The wo k p esen ed
he e signi ican ly add esses each o hese limi a ions. In ac , we could success ully analyze, in he same p ope y
space, di e en ypes o d ugs and chemicals, and ENMs. INSIdE NANO b oadens he classical e alua ion o
chemical exposu es, based on he s uc u al p ope ies o he compounds, o hei p ima y MOA. Doing so, we
e ie ed ele an in o ma ion abou ENMs and hei e ec s by con ex ualizing hei molecula beha io wi h
espec o mul iple pheno ypic en i ies (ENMs s chemicals, d ugs and diseases). To he bes o ou knowledge,
his is he i s a emp o analyze he molecula e ec s o ENMs in he con ex o a la ge space including o he
chemicals, d ugs, and human diseases. Mo eo e , we demons a ed ha , when accu a ely de i ed and in e p e ed,
simila i y pa e ns o omics-de i ed MOA a e able o ecapi ula e s uc u al analogies o he compounds as well
as clinically ele an ela ionships be ween diseases, d ugs and diseases, and chemicals and diseases. Finally, ou
me hods p o ide a sys ema ic way o in e obus implica ions o exposu es o human diseases, going beyond
www.na u e.com/scien i ic epo s/
7
SCIENTIFIC REPORTS | (2019) 9:179 | DOI:10.1038/s41598-018-37411-y
speci ic oxicology endpoin s, which can be di icul o link o human pa hogenesis. An impo an egula o y
and e hical issue is he possibili y o de i e o ganism-le el in o ma ion om in i o assays. Ma ching MOA
signa u es o d ugs es ed in i o wi h pa e ns o molecula al e a ions o pa ien s has al eady p o ed alid in
sugges ing d ug eposi ioning7. Mo eo e , we ha e ecen ly demons a ed ha a gene ne wo k-based analysis o
omics da a allows o highligh molecula pa hways consis en ly al e ed by ENMs exposu e in i o and in i o39.
Along he same lines, he e we in eg a ed MOA signa u es de i ed in i o (ENMs, d ugs, and some chemicals)
and in i o (diseases and some chemicals). Impo an ly, he associa ions be ween ENMs and neu odegene a i e
diso de s compu a ionally p edic ed by INSIdE NANO a e ecapi ula ed in a whole body in i o exposu e model
in ish, and also oden s. I should be no ed ha omics sc eening in i o can be used o iden i y he MOA asso-
cia ed wi h an exposu e, which is he ensemble o he p ima y molecula al e a ions caused by ha exposu e.
In his sense, in i o expe imen s can be o g ea alue in in e ing pa hways o oxici y. We acknowledge ha
he cu en lack o da a conce ning ENM MOA poses a challenge in espec o he po en ial o INSIdE NANO
and u u e i e a ions o he ool will ake in o accoun new da a as hese become a ailable. Howe e , despi e his
po en ial limi a ion, we we e al eady able o de i e meaning ul and s a is ically obus simila i ies be ween ENMs,
d ugs, chemicals, and human diseases. In conclusion, we ha e de eloped INSIdE NANO, a no el compu a ional
pla o m o he sys ema ic con ex ualiza ion o ENMs MOA in ela ion o human diseases, d ug ea men s, and
chemical exposu es. Ou analysis o he la ge in eg a ed da a se unde lying INSIdE NANO has poin ed owa ds
no el associa ions o speci ic me al and me al oxide nanopa icles wi h neu odegene a i e diso de s, and unde -
sco es he u ili y o ansc ip omics analysis in i o o he p edic ion o possible in i o e ec s o ENMs. These
esul s sugges ha epidemiological s udies o he possible ela ionships be ween exposu e o me al based nan-
opa icles and neu odegene a ion a e wa an ed o es ablish whe he ENMs a e a isk ac o o such diso de s.
Me hods
Da a in eg a ion. Fo each pheno ypic en i y, a lis o associa ed genes is gi en. In pa icula , a se o genes
is associa ed o each disease and chemical, while an o de ed lis o genes esul ing om di e en ial exp ession
analysis is buil o each d ug and ENM in he da a se . In o de o cons uc a simila i y ne wo k be ween he
pheno ypic en i ies all he pai -wise simila i ies be ween hem we e e alua ed (Figs1C and 2a).
Gene se e sus gene se simila i y. The Jacca d Index was used o compu e he pai -wise simila i y be ween gene
se s ( wo diseases, wo chemicals o a disease and a chemical). Gi en wo se s A and B he Jacca d index is de ined
as: =
∩
∪
||
||
JA
B(,)
AB
AB
. This measu e is 0 i he in e sec ion be ween A and B is emp y, while i is 1 i i con ains
exac ly he same elemen s. Fo each chemical, wo se s o genes we e conside ed: hose whose exp ession is
up- egula ed and hose whose exp ession is down- egula ed by he chemical exposu e. Fo he down- egula ed
genes, he Jacca d Index was mul iplied by −1 in o de o ake in o accoun he e ec s on he genes.
Gene ank e sus gene ank simila i y. A e impo ing he p e-p ocessed NanoMine and CMAP da ase s in R,
a con as ma ix o each da ase was cons uc ed by using he limma package; only he subse o sha ed genes in
bo h da ase s was conside ed. Fo he NanoMine da ase , he con as s we e de ined o compa e each sample
exposed o an ENM agains he con ols. Likewise, o he CMAP da a, con as is de ined conside ing each d ug
e sus he un ea ed con ols. Subsequen ly, he genes we e anked by using he ollowing sco e
±⋅−logFClog P al()
, esul ing in o de ed gene lis s ha ing he mos up egula ed genes on he op and he mos
down egula ed genes in he bo om. The Kendall Tau Dis ance40 was hen used o e alua e he simila i y be ween
ENMs, d ugs and ENMs-d ugs based on he anked lis s o genes. The Kendall Tau dis ance be ween wo lis s T1
and T2 is de ined as ollow:
=| <<
∧> ∨>
∧< |
KT Tiji jTiTi
Ti Tj Ti Tj
Ti Tj
(, ){(, ): ,((()())
(()())) (( () ())
(()()))}
(1)
12 11
22 11
22
whe e T1 and T2 a e wo anked lis s o genes. Thei alues ange be ween 0 and n(n − 1), whe e n is he lis leng h.
A alue o 0 means ha elemen s in he lis a e in he same o de ; A alue o n(n − 1) means ha elemen s in
he lis a e in he opposi e o de . Values we e inally no malized o he ange [−1; 1] whe e −1 co esponds o
n(n − 1) and 1 co esponds o 0.
Gene ank e sus gene se simila i y. The Gene Se En ichmen Analysis (GSEA)41, based on he
Kolmogo o -Smi no es , was used o compu e he pai wise simila i y be ween an ENM and a disease, and an
ENM and a chemical, a d ug and a disease, and a d ug and a chemical. The Kolmogo o -Smi no es 42 can be
used o compa e a sample wi h a e e ence p obabili y dis ibu ion. The empi ical dis ibu ion unc ion Fn o n
iid obse a ions Xi is de ined as =∑−∞
=
Fx Ix
x() [,](
)
n
n
i
n
i
1
1 whe e
−∞IxX[,]( )
i
is he indica o unc ion
de ined on a se X ha indica es he membe ship o an elemen o a subse A o X, ha ing he alue 1 o all ele-
men s o A and he alue 0 o all elemen s o X no in A. The Kolmogo o -Smi no s a is ic o a gi en cumula-
i e dis ibu ion unc ion F(x) is Dn = supx|Fn(x) − F(x)|. As in43, he Kolmogo o -Smi no s a is ic was used
wi hou he absolu e alue in o de o p ese e he sign. This helps unde s anding i he genes in he se s a e up o
down- egula ed.
Pheno ypic Ne wo k In e ence. The pai wise simila i y ma ix was used as an adjacency ma ix o con-
s uc a weigh ed undi ec ed ne wo k whe e he nodes a e he en i ies and he simila i ies be ween hem
www.na u e.com/scien i ic epo s/
8
SCIENTIFIC REPORTS | (2019) 9:179 | DOI:10.1038/s41598-018-37411-y
ep esen he edge weigh s. Each simila i y measu e has a di e en ange o alues. To make hem compa able,
hese alues we e scaled in he uni o m ange 0–1 by means o he cumula i e unc ion. Unlike he simila i y
alue, he signs ha e no been al e ed, and hen edges in he ne wo k ha e a sign ha indica e i he co ela ion
be ween a couple o nodes is posi i e o nega i e. The esul ing ne wo k is comple ely connec ed. To educe he
numbe o nodes and analyze only s ong connec ions, we used a anking sys em o cu edges. Fo each e ex we
anked i s neighbo s basing on he simila i y sco e; hen we can que y he ne wo k by se ing a pe cen age o he
op edges o selec (e.g. i s 10%, 20%, 30% o he ank). Since ankings a e no symme ic, when we cu he
anked lis we compu e he mu ual neighbo hood o a node i de ined as
=≤∧≤ij ank j h anki h() {:(())(())}
ij
, whe e anki(j) is he posi ion o node j in he anked lis o nodes
connec ed o i and h is he use de ined h eshold.
Cliques Sea ch. A g aph o ne wo k is a ma hema ical abs ac ion ha ep esen s a se o objec s (nodes) and
hei ela ionships (edges). Fo mally, a g aph G is de ined as he pai G = (V, E), whe e
=…V n1, ,
is he se
o he nodes o he g aph, and
=…Ee em1, ,
is he se o he edges. Each edge in E is a connec ion be ween a pai
o nodes (x, y) in V. I a ele an so ing o de in he pai (x, y) is p esen , hen he g aph G will be said o be o i-
en ed (o di ec ed), whe e x will be he sou ce o he edge and y he des ina ion. On he o he side, i he e is no
ele an o de , he g aph G will be said o be uno ien ed (o undi ec ed). In an undi ec ed g aph G, a clique is
de ined as a subg aph G′ = (V′, E′) o G wi h V′ in V and E′ in E, whe e all he pai s o nodes in G′ a e connec ed
by an edge. INSIdE nano is an indi ec g aph, whe e he e ices a e labeled by he class o he pheno ypic en i ies
(ENM, d ug, chemical and disease). The he e ogeneous cliques wi h ou (o h ee) di e en e ex classes we e
sys ema ically e ie ed wi hin he ne wo k by an exhaus i e sea ch algo i hm implemen ed in phy on
(Supplemen a y Fig.S2).
Valida ion o he Simila i y Measu es
The pai wise pheno ypic simila i ies based on he MoA we e sys ema ically compa ed wi h o he independen ly
compu ed simila i ies based on di e en cha ac e is ics, such as he molecula s uc u e o he d ugs and chem-
icals, he symp oms o he diseases, he use in clinical p ac ice o d ugs, and he pa hogenic oles o chemi-
cal exposu es. (See Fig.2b). The 2D d ug s uc u es, in he o m o smiles ec o s, we e downloaded om he
D ugBank Da abase (h ps://www.d ugbank.ca)44. Simila ly, he smiles o chemical compounds we e e ie ed
om he Chemspide Da abase (h p://www.chemspide .com/). The pai wise d ug-d ug, chemical-chemical, and
d ug-chemical simila i ies we e compu ed wi h he Op imal s ing alignmen algo i hm implemen ed in he R
package “s ingdis ”45. The associa ions be ween d ugs and diseases, based on clinical indica ions o d ugs, we e
downloaded om he MEDI P esc ip ion Da abase (h ps://medschool. ande bil .edu)11,12. In his case, he sim-
ila i y was de ined as a bina y sco e, whe e 1 deno es ha a gi en d ug is used o ea a ce ain disease, while he
0 means no p esc ip ion indica ion. The associa ions be ween chemicals and diseases we e downloaded om he
Compa a i e Toxicogenomics Da abase (h p://c dbase.o g/). The simila i ies be ween diseases we e e ie ed
om he Supplemen a y ma e ials o a p e ious s udy by Zhou e al.46, whe e a symp om-based human disease
ne wo k was buil om a ious da abases. The compa isons be ween he simila i y ma ices de i ed om MOA
and he o he s we e pe o med by he Man el Tes , which is used o e alua e he co ela ion be ween pai s o
simila i y ma ices, by adop ing a pe mu a ion es p ocedu e47.
S a is ical e alua ion o pheno ypic cliques. In o de o s a is ically alida e he se s o cliques ela ed
o each disease, a pe mu a ion es was pe o med. The o iginal adjacency ma ix was andomly shu led 1,000
imes. Fo each clique, a p alue was compu ed by coun ing how many imes he s eng h o connec ion in he
o iginal clique ( he sum o he weigh s o i s edges) is highe han he s eng h o connec ion o he same clique
connec ed by pe mu a ed edges. The ob ained p alues a e hen co ec ed wi h he Fd me hod. Only he cliques
wi h p alue (<0.05) we e conside ed.
INSIdE nano ool. INSIdEnano is a web-based ool (publicly a ailable a h p://inano.bioby e.de) ha high-
ligh s connec ions be ween pheno ypic en i ies based on hei e ec s on he genes. The da a collec ion, p epos-
sessing and in eg a ion s a egies we e implemen ed in R, as desc ibed abo e. The g aphical ool and he ou ine
o scan he ne wo k we e implemen ed in Py hon and Ja asc ip using he d3 lib a y o he G aphical Use
In e ace (GUI). INSIdE nano was de eloped in a clien -se e s uc u e: he clien is esponsible o managing
he use in e ace, collec ing he use inpu and displaying he ou pu s. The se e , ins ead, p ocesses he da a
om he da abase acco ding o he use inpu s, and ou pu s he esul s o he clien . The ool p o ides wo di e -
en ypes o que ies. The simple que y allows he use o in es iga e connec ions o a speci ic elemen in he ne -
wo k. Gi en a node and a h eshold, he ool shows all i s neighbo s di ided in o ou ca ego ies: ENMs, diseases,
d ugs and chemicals. The condi ional que y allows he use o que y he ne wo k by applying di e en il e s o
sea ch o he cliques. Since he pu pose o he analysis is o compa e he beha io o a gi en elemen wi h espec
o he o he s, he use mus speci y a leas wo di e en ypes o i ems. Mo eo e , he le el o simila i y necessa y
o epo a connec ion be ween selec ed i ems, he numbe o i ems ha mus be in he same esul ing cliques,
and he numbe o que y i ems being connec ed o he o he nodes in he sub-ne wo k a e eques ed as inpu .
Fi s , he ool e ie es he sub-ne wo k o all he elemen s, connec ed o he que y i ems ha sa is y he use
inpu . Then i scans he ne wo k in sea ch o cliques. The cliques can con ain h ee he e ogeneous elemen s, ha
will be any one o he possible combina ions o h ee elemen s be ween ENMs, d ugs, chemicals and diseases in
he sub-ne wo k (e.g., an ENM, a d ug, a chemical; a nano, a d ug, a disease; e c.,), o hey will con ain exac ly 4
elemen s (an ENM, a chemical, a d ug and a disease). Those cliques a e hen g ouped wi h espec o he na u e
o he connec ions be ween each couple o i ems ha hey con ain. As a esul o he analysis, he ool gi es he
oppo uni y o isualize he sub-ne wo k o all he nodes connec ed o he que ied en i ies ha sa is y he use
www.na u e.com/scien i ic epo s/
9
SCIENTIFIC REPORTS | (2019) 9:179 | DOI:10.1038/s41598-018-37411-y
equi emen s. I displays he lis o all he cliques wi h he oppo uni y o analyze each one o hem and inspec he
genes unde lying he connec ions. Mo eo e , di ec links o ele an ex e nal sou ces o in o ma ion a e a ailable
o each pheno ype. A comple e u o ial is a ailable a h p://inano.bioby e.de/help.cgi and in he Supplemen a y
ma e ials ile.
Re e ences
1. K ug, H. F. Nanosa e y esea ch—a e we on he igh ack? Angewand e Chemie In . Ed. 53, 12304–12319, h ps://doi.o g/10.1002/
anie.201403367 (2014).
2. Valsami-Jones, E. & Lynch, I. How sa e a e nanoma e ials? Science 350, 388–389 (2015).
3. Cos a, P. M. & Fadeel, B. Eme ging sys ems biology app oaches in nano oxicology: owa ds a mechanism-based unde s anding o
nanoma e ial haza d and isk. Toxicol. Appl. Pha macol. 299, 101–111 (2016).
4. Lamb, J. e al. The connec i i y map: using gene-exp ession signa u es o connec small molecules, genes, and disease. Science 313,
1929–1935 (2006).
5. Io io, F. e al. Disco e y o d ug mode o ac ion and d ug eposi ioning om ansc ip ional esponses. P oc. Na l. Acad. Sci. 107,
14621–14626 (2010).
6. Napoli ano, F. e al. D ug eposi ioning: a machine-lea ning app oach h ough da a in eg a ion. J. Chemin o ma ics 5, 30 (2013).
7. Dudley, J. T. e al. Compu a ional eposi ioning o he an icon ulsan opi ama e o in lamma o y bowel Disease. Sci. T ansl.
Medicine 3 (2011).
8. Da is, A. P. e al. The compa a i e oxicogenomics da abase’s 10 h yea anni e sa y: upda e 2015. Nucleic Acids Res. 43, D914–D920
(2015).
9. Lamb, J. The connec i i y map: a new ool o biomedical esea ch. Na . Re . Cance 7, 54–60, h ps://doi.o g/10.1038/n c2044
(2007).
10. Kong, L. e al. Nanomine - in eg a i e human ansc ip omics da a esou ce o nanopa icle esea ch. PloS One 8, e68414 (2013).
11. Wei, W.-Q., Mosley, J. D., Bas a ache, L. & Denny, J. C. Valida ion and enhancemen o a compu able medica ion indica ion esou ce
(medi) using a la ge p ac ice-based da ase . In AMIA Annual Symposium P oceedings, ol. 2013, 1448 (Ame ican Medical
In o ma ics Associa ion, 2013).
12. Wei, W.-Q. e al. De elopmen and e alua ion o an ensemble esou ce linking medica ions o hei indica ions. J. Am. Med.
In o ma ics Assoc. 20, 954–961 (2013).
13. Obe dö s e , G., Elde , A. & Rinde knech , A. Nanopa icles and he b ain: cause o conce n? J. Nanosci. Nano echnol. 9, 4996–5007,
h ps://doi.o g/10.1166/jnn.2009.GR02, NIHMS150003 (2009).
14. Miglio e, L., Uboldi, C., Di Bucchianico, S. & Coppedè, F. Nanoma e ials and neu odegene a ion. En i on. Mol. Mu agen. 56,
149–170, h ps://doi.o g/10.1002/em.21931 (2015).
15. Pea son, B. L. e al. Iden i ica ion o chemicals ha mimic ansc ip ional changes associa ed wi h au ism, b ain aging and
neu odegene a ion. Na . Commun. 7, 11173 (2016).
16. An onini, J. M., San ama ia, A. B., Jenkins, N. T., Albini, E. & Lucchini, R. Fa e o manganese associa ed wi h he inhala ion o
welding umes: po en ial neu ological e ec s. Neu o oxicology 27, 304–310 (2006).
17. Chin-Chan, M., Na a o-Yepes, J. & Quin anilla-Vega, B. En i onmen al pollu an s as isk ac o s o neu odegene a i e diso de s:
Alzheime and Pa kinson diseases. F on . Cell. Neu osci. 9, 124, h ps://doi.o g/10.3389/ ncel.2015.00124 (2015).
18. Sh edo a, A. A., Kagan, V. E. & Fadeel, B. Close encoun e s o he small kind: ad e se e ec s o man-made ma e ials in e acing wi h
he nano-cosmos o biological sys ems. Annu. Re . Pha macol. Toxicol. 50, 63–88, h ps://doi.o g/10.1146/annu e .
pha m ox.010909.105819 (2010).
19. Bas ian, S. e al. Toxici y o ungs en ca bide and cobal -doped ungs en ca bide nanopa icles in mammalian cells in i o. En i on.
Heal. Pe spec i es 117, 530–536, h ps://doi.o g/10.1289/ehp.0800121 (2009).
20. Song, B., Liu, J., Feng, X., Wei, L. & Shao, L. A e iew on po en ial neu o oxici y o i anium dioxide nanopa icles. Nanoscale Res.
Le . 10, 342 (2015).
21. Tuomela, S. e al. Gene exp ession p o iling o immune-compe en human cells exposed o enginee ed zinc oxide o i anium
dioxide nanopa icles. PloS One 8, e68415 (2013).
22. Hue a-Ga cía, E. e al. Ti anium dioxide nanopa icles induce s ong oxida i e s ess and mi ochond ial damage in glial cells. F ee.
Radic. Biol. Medicine 73, 84–94, h ps://doi.o g/10.1016/j. ee adbiomed.2014.04.026 (2014).
23. Til on, S. C. e al. Th ee human cell ypes espond o mul i-walled ca bon nano ubes and i anium dioxide nanobel s wi h cell-
speci ic ansc ip omic and p o eomic exp ession pa e ns. Nano oxicology 8, 533–548, h ps://doi.o g/10.3109/17435390.2013.803
624 (2014).
24. Bue ki-Thu nhe , T. e al. In i o mechanis ic s udy owa ds a be e unde s anding o zno nanopa icle oxici y. Nano oxicology 7,
402–416 (2013).
25. Tian, L. e al. Neu o oxici y induced by zinc oxide nanopa icles: age- ela ed di e ences and in e ac ion. Sci. Repo s 5, 16117,
h ps://doi.o g/10.1038/s ep16117. (2015).
26. Xie, Y., Wang, Y., Zhang, T., Ren, G. & Yang, Z. E ec s o nanopa icle zinc oxide on spa ial cogni ion and synap ic plas ici y in mice
wi h dep essi e-like beha io s. J. Biomed. Sci. 19, 14 (2012).
27. Siddiqi, N. J., Abdelhalim, M. A. K., El-Ansa y, A. K., Alhomida, A. S. & Ong, W. Iden i ica ion o po en ial bioma ke s o gold
nanopa icle oxici y in a b ains. J. Neu oin lamma ion 9, 123 (2012).
28. Schlump , U., Meye , M., Ul ich, J. & F iede, R. L. Neu ologic complica ions induced by gold ea men . A h i is & Rheum. 26,
825–831, h ps://doi.o g/10.1002/a .1780260702 (1983).
29. Gamba i, P., Os uni, P., Lazza in, P., Ta ola o, B. & Todesco, S. Neu o oxici y ollowing a e y high dose o o al gold (au ano in).
A h i is & Rheum. 27, 1316–1317, h ps://doi.o g/10.1002/a .1780271121 (1984).
30. Pio o, E. P. e al. Dex ome ho phan plus ul a low-dose quinidine educes pseudobulba a ec . Annals neu ology 68, 693–702
(2010).
31. Sode lund, D. M. & Bloomquis , J. R. Neu o oxic ac ions o py e h oid insec icides. Annu. Re . En omol. 34, 77–96, h ps://doi.
o g/10.1146/annu e .en.34.010189.000453 (1989).
32. Be gg en, E. e al. Chemical sa e y assessmen using ead-ac oss: assessing he use o no el es ing me hods o s eng hen he
e idence base o decision making. En i on. Heal. Pe spec i es 123, 1232 (2015).
33. Pa lewicz, G., Helman, G., P adeep, P. & Shah, I. Na iga ing h ough he mine ield o ead-ac oss ools: a e iew o in silico ools o
g ouping. Compu . Toxicol. 3, 1–18, h ps://doi.o g/10.1016/j.com ox.2017.05.003 (2017).
34. Gajewicz, A. Wha i he numbe o nano oxici y da a is oo small o de eloping p edic i e nano-qsa models? An al e na i e ead-
ac oss based app oach o illing da a gaps. Nanoscale 9, 8435–8448 (2017).
35. Oomen, A. G., Bleeke , E. A. J., Bos, P. M. J. & B oekhuizen, F. V. G ouping and ead-ac oss app oaches o isk assessmen o
nanoma e ials. In . J. En i on. Res. Public Heal. 12, 13415–13434, h ps://doi.o g/10.3390/ije ph121013415 (2015).
36. G odwin, H. e al. Nanoma e ial ca ego iza ion o assessing isk po en ial o acili a e egula o y decision-making. ACS Nano 9,
3409–3417 (2015).