PiRS and su i al analysis ac oss cance coho :
This is he supplemen a y igu e o he main Figu e 5. The PiRS sco e calcula ed o all pa ien s
was sc eened o a i e a a cu o h eshold o ca ego ise pa ien s in o wo (low and high PiRS)
g oups.
Supplemen a y igu e 1: The PiRS sco es a e plo ed on he x-axis as hei pe cen ile. Su i al
analysis was ca ied ou o each pe cen ile and he esul ing p- alue om he su i al model is
plo ed in nega i e log10 scale in he y-axis. The ho izon al line e e s o p alue < 0.01 indica ing
he signi icance. The sco es abo e his line we e conside ed o ca ego ising he pa ien s.
The cu o s we e chosen based on he minimum p- alue ha also happened o be a ound he 50 h
PiRS quan ile. Fo COAD, GBM and SKCM, we empi ically chose he sco es o ge a nea ly equal
p opo ion o pa ien s be ween g oups wi h p < 0.01 signi icance. The cu -o sco es selec ed a e
gi en in he ollowing able along wi h he numbe o pa ien s in each g oup.
Figu e 5 in he main documen shows he Kaplan-Meie plo s o unde s and he di e ence in
su i al be ween he g oups. Pa ien s in he low PiRS had be e su i al han hose in he high
PiRS g oup.
Supplemen a y me hod:
1. Node weigh s we e calcula ed independen ly o each sample. In addi ion, a coho -le el analysis
was also ca ied ou by conside ing di e en pa ien samples as biological eplica es (pool o
umou samples e sus pool o non- umou con ol) in he coho . The esul ing Mu Pa hs showed
simila en ichmen pa e ns as he indi idual pa ien -speci ic Mu Pa hs.
2. Co ela ion o he isk sco e (PiRS) wi h he clinical pa ame e s:
We a e able o see di e ences in he esul ing iPanels ob ained ac oss he s ages o all 6 cance
ypes which includes se e al well-classi ied sub ypes in each. The su i al analysis and subsequen
PiRS calcula ion o indi idual pa ien s based on hei iPanels show an inc easing sco e end wi h
inc easing s ages. Supplemen a y igu e 2 depic s he di e en ia ing end o he PiRS sco e
be ween Li ing and Deceased pa ien s ac oss cance s ages.
Supplemen a y igu e 2: Di e ences in he PiRS sco es among he Li ing (g een) and Deceased ( ed) pa ien s ac oss cance s ages in each coho
Mo eo e , he Ne Sco e componen o he PiRS sco ing scheme explains he p ognos ic e ec o
he iden i ied iPanel genes wi h espec o he pa ien ’s clinical pa ame e s such as he cance s age,
disease du a ion, and s a us a he end o he s udy.
A case s udy o 6 pa ien s om TCGA-BRCA p esen ed in supplemen a y igu e 3 p o es ha PiRS
is capable o explaining he e ec o he d i e gene panel on he clinical pheno ype. We obse e
he associa ion o he sco e wi h he o e all su i al mon hs, cance s ages and s a us o he pa ien .
In scena io 1, he pa ien P2 wi h Deceased s a us ha e signi ican ly highe PiRS despi e ha ing
simila su i al mon hs as o P1. In scena io 2, hough he su i al mon hs we e he same o P3
and P4, PiRS is highe o he s age 4 pa ien (P4). In scena io 3, P6 ha e highe PiRS
co esponding o he highe su i al mon h and he cance s age.
Supplemen a y Figu e 3: Compa ison o he PiRS (P OPs indi idual Risk Sco e) o 6 pa ien s
om he TCGA-BRCA coho wi h he same d i e gene panel, TP53. Due o he di e ences in he
ne wo k opology o he Mu Pa hs and hei downs eam connec i i y o he pe u bed genes, TP53
has di e en Ne Sco es in all 6 pa ien s. This con ibu ed o he di e en PiRS in he pa ien s
despi e hem ha ing he same panel gene.
3. Benchma king:
We ca ied ou benchma king wi h 2 me hods - DawnRank and PRODIGY. DawnRank was
benchma ked agains D i e Ne , PARADIGM-Shi , CHASM and Oncod i e-FM and PRODIGY
agains DawnRank and SCS, while SCS was benchma ked agains DawnRank and OncoIMPACT.
Figu es shown he e, aken om hei espec i e pape s clea ly indica e DawnRank and PRODIGY
o be comp ehensi e bes -pe o ming me hods o iden i y d i e mu a ions. Hence, we chose o
benchma k agains hese 2 me hods.
O e all Su i al
mon hs
PiRS
S a us
Scena io1
P1
19.9
9.06
Ali e
P2
24.7
154.4
Deceased
Scena io2
P3
17.21
18.6
T1
P4
18.76
81.06
T4
Scena io3
P5
74.87
5.07
T1
P6
139.06
26.3
T4
This igu e aken om he pape DawnRank: disco e ing pe sonalized d i e genes in cance ,
desc ibes he compa ison o DawnRank wi h D i e Ne and PARADIGM-Shi . All he measu es,
namely P ecision, Recall and F1-sco e show ha DawnRank pe o mance is supe io .
(Re e ence: Hou, J. P. & Ma, J. Dawn ank: disco e ing pe sonalized d i e genes in cance .
Genome Medicine,6 (2014).)
Simila ly, he abo e igu e 5 om P odigy: pe sonalized p io i iza ion o d i e genes depic s he
be e pe o mance o PRODIGY as compa ed o DawnRank, SCS and ne wo k cen ali y measu es
like closeness, be weeness and deg ee cen ali ies.
(Re e ence: Dins ag, G. & Shami , R. P odigy: pe sonalized p io i iza ion o d i e genes.
Bioin o ma ics 36, 1831–1839 (2019).)
O e all, ou algo i hm P OPs has been di ec ly compa ed wi h 2 and indi ec ly wi h a o al o 6
di e en algo i hms. This makes he benchma king qui e obus . We ha e explained he wo king
p inciple o each algo i hm in he able below.
Me hod
Wo king p inciple
Inpu
Ou pu
P OPs
Uses max co e g eedy
algo i hm o iden i y d i e
mu a ions ela ed o he
pe u bed genes in he
p o ein-p o ein in e ac ion
ne wo k
Pa ien -speci ic
mu a ion and
gene exp ession
lis
Pa ien -le el; Pe sonalised
panel o anked d i e
mu a ions and hei
co esponding p ognos ic isk
sco e and ac ionabili y s a us;
ypically concise panels
Dawn ank
Uses he andom walk app oach
o PageRank o iden i y
mu a ions ha ha e highe
connec i i y in he
p o ein-p o ein in e ac ion
ne wo k.
Pa ien -speci ic
mu a ion and
gene exp ession
lis
Pa ien -le el; Lis o anked
pa ien -speci ic d i e genes;
ypically concise panels
D i e Ne
Uses he ne wo k opology o he
ansc ip ional egula o y
ne wo k o iden i y d i e
mu a ions
Mu a ion and
gene exp ession
lis
Coho -le el lis o d i e
genes.
PARADIGM-Shi
Belie -p opaga ion algo i hm
ha in e s gene ac i i y based on
CNV and gene exp essions in
speci ic biological pa hways
Copy numbe
a ia ion,
Me hyla ion
s a us, Gene
exp ession and
speci ic
biological
pa hways
Coho -le el; P edic s he
unc ional s a us o he d i e
genes as ei he
loss-o - unc ion o
gain-o - unc ion mu a ions.
CHASM
Random o es classi ie ained
on 49 ea u es o dis inguish he
d i e mu a ions om
syn he ically gene a ed
passenge mu a ion lis .
Lis o
cance -speci ic
mu a ed genes
Coho -le el; Lis o
cance -speci ic d i e genes
Oncod i eFM
Compu es he unc ional impac
o each a ian in he coho and
he unc ional bias o hose
a ian s o occu in speci ic
genes and pa hways
Lis o
cance -speci ic
single-nucleo i
de a ian s
Coho -le el; Lis o
FM-biased d i e genes and
pa hways
PRODIGY
Uses he p ize-collec ing S eine
ee model o ank he d i e
mu a ions based on hei
agg ega ed impac on he
dys egula ed pa hways
Pa ien -speci ic
mu a ion and
gene exp ession
lis
Lis o genes
associa ed wi h
Pa ien -le el; Pa ien -speci ic
anked lis o d i e genes
(panel a e ypically la ge,
10-40+ genes)
REACTOME
pa hways.
SCS
Iden i ies d i e genes based on
hei impac on he gene
exp ession p o ile in he disease
s a e compa ed o a no mal s a e.
Pa ien -speci ic
mu a ion and
gene exp ession
lis
Pa ien -le el; Ranked lis o
pa ien -speci ic d i e genes.
OncoIMPACT
Iden i ies d i e genes ha a e
associa ed wi h he dys egula ed
ne wo k modules
Lis o
cance -speci ic
mu a ions and
gene exp ession
p o iles
Pa ien -le el; Pa ien -speci ic
d i e gene lis .
