Appraising the relevance of DNA copy number loss and gain in prostate cancer using whole genome DNA sequence data

RESEARCH ARTICLE App aising he ele ance o DNA copy numbe loss and gain in p os a e cance using whole genome DNA sequence da a Niedzica Camacho 1,2,3 , Pe e Van Loo 4,5 , Sand a Edwa ds 1 , Jona han D. Kay 6,7 , Lucy Ma hews 1 , Ke s in Haase 4 , Je emy Cla k 8 , Nening Dennis 9 , Sa ah Thomas 9 , Ba ba a K emeye 10 , Jo ge Zamo a 10 , Adam P. Bu le 10 , Gunes Gundem 10,11 , Sue Me son 1 , Hayley Lux on 6,7 , S e e Hawkins 6 , Mohammed Gho i 10 , Luke Ma sden 12 , Adam Lambe 13 , Ka alin Ka aszi 13,14 , Gill Pel ende 14 , Cha lie E. Massie 6,15 , Zso ia Ko e- Ja ai 1 , Kei an Raine 10 , Da id Jones 10 , William J. Howa 16 , S e en Hazell 9 , Naomi Li ni 9 , Cy il Fishe 9 , Ch is ophe Ogden 9 , Pa deep Kuma 9 , Alan Thompson 9 , Da id Nicol 9 , E ik Maye 9 , Tim Dudde idge 9 , Yongwei Yu 17 , Hongwei Zhang 17 , Nimish C. Shah 18 , Vincen J. Gnanap agasam 19 , The CRUK-ICGC P os a e G oup ¶ , William Isaacs 20 , Tapio Visako pi 21 , F eddie Hamdy 14 , Dan Be ney 22 , Cla e Ve ill 23 , Anne Y. Wa en 18 , Da id C. Wedge 10,24 , And ew G. Lynch 25,26‡ , Ch is ophe S. Fos e 27‡ , Yong Jie Lu 22 , G. S e en Bo a 21‡ , Hayley C. Whi ake 6,7 , Ul an McDe mo 10‡ , Da id E. Neal 6,19‡ , Rosalind Eeles 1,9‡ , Colin S. Coope 1,8‡ , Daniel S. B ewe 8,28‡ * 1Di ision o Gene ics and Epidemiology, The Ins i u e O Cance Resea ch, London, Uni ed Kingdom, 2Human Oncology and Pa hogenesis P og am, Memo ial Sloan Ke e ing Cance Cen e , New Yo k, New Yo k, Uni ed S a es o Ame ica, 3Ma ie-Jose ´e and Hen y R. K a is Cen e o Molecula Oncology, Memo ial Sloan Ke e ing Cance Cen e , New Yo k, New Yo k, Uni ed S a es o Ame ica, 4Cance Genomics Labo a o y, The F ancis C ick Ins i u e, London, Uni ed Kingdom, 5Depa men o Human Gene ics, Uni e si y o Leu en, Leu en, Belgium, 6U o-Oncology Resea ch G oup, Cance Resea ch UK Camb idge Ins i u e, Camb idge, Camb idgeshi e, Uni ed Kingdom, 7Molecula Diagnos ics and The apeu ics G oup, Uni e si y College London, London, Uni ed Kingdom, 8No wich Medical School, Uni e si y o Eas Anglia, No wich, No olk, Uni ed Kingdom, 9Cance Gene ics Uni , Royal Ma sden NHS Founda ion T us , London, Uni ed Kingdom, 10 Cance , Ageing and Soma ic Mu a ion, Wellcome T us Sange Ins i u e, Hinx on, Camb idgeshi e, Uni ed Kingdom, 11 Epidemiology & Bios a is ics, Memo ial Sloan-Ke e ing Cance Cen e , New Yo k, New Yo k, Uni ed S a es o Ame ica, 12 Depa men o Physiology, Uni e si y o Ox o d, Ox o d, Ox o dshi e, Uni ed Kingdom, 13 Depa men o Oncology, CRUK/MRC Ox o d Ins i u e o Radia ion Oncology, Ox o d, Ox o dshi e, Uni ed Kingdom, 14 Nu ield Depa men o Su gical Sciences, Uni e si y o Ox o d, Ox o d, Ox o dshi e, Uni ed Kingdom, 15 CRUK Camb idge Cen e, Ea ly De ec ion P og amme, U ological Malignancies P og amme, Hu chison-MRC Resea ch Cen e, Camb idge, Camb idgeshi e, Uni ed Kingdom, 16 His opa hology and in si u hyb idiza ion Resea ch G oup, Cance Resea ch UK Camb idge Ins i u e, Camb idge, Camb idgeshi e, Uni ed Kingdom, 17 Depa men o Epidemiology, Second Mili a y Medical Uni e si y, Shanghai, China, 18 Depa men o His opa hology, Camb idge Uni e si y Hospi als NHS Founda ion T us , Camb idge, Camb idgeshi e, Uni ed Kingdom, 19 Academic U ology G oup, Depa men o Su ge y, Uni e si y o Camb idge, Camb idge, Camb idgeshi e, Uni ed Kingdom, 20 School o Medicine, Johns Hopkins Uni e si y, Bal imo e, Ma yland, Uni ed S a es o Ame ica, 21 Facul y o Medicine and Li e Sciences and BioMediTech Ins i u e, Uni e si y o Tampe e and Tampe e Uni e si y Hospi al, Tampe e, Finland, 22 Cen e o Molecula Oncology, Ba s Cance Ins i u e, The Ba s and London School o Medicine and Den is y, Queen Ma y Uni e si y o London, London, Uni ed Kingdom, 23 Depa men o Cellula Pa hology and Ox o d Biomedical Resea ch Cen e, Ox o d Uni e si y Hospi als NHS T us , Ox o d, Ox o dshi e, Uni ed Kingdom, 24 Ox o d Big Da a Ins i u e & Ox o d Cen e o Cance Gene Resea ch, Wellcome T us Cen e o Human Gene ics, Ox o d, Ox o dshi e, Uni ed Kingdom, 25 S a is ics and Compu a ional Biology Labo a o y, Cance Resea ch UK Camb idge Ins i u e, Camb idge, Camb idgeshi e, Uni ed Kingdom, 26 School o Ma hema ics and S a is ics/School o Medicine, Uni e si y o S And ews, S And ews, Fi e, Sco land, 27 HCA Pa hology Labo a o ies, HCA Heal hca e, London, Uni ed Kingdom, 28 O ganisms and Ecosys ems, The Ea lham Ins i u e, No wich, No olk, Uni ed Kingdom ‡ CSC and DSB a e join senio au ho s on his wo k. AGL, CSF, GSB, UM, DEN, RE, CSC, and DSB a e senio p incipal in es iga o s o his p ojec and he CRUK-ICGC P os a e G oup. ¶ Membe ship o he CRUK-ICGC P os a e G oup is p o ided in S3 Appendix. *[email p o ec ed] PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 1 / 28 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Ci a ion: Camacho N, Van Loo P, Edwa ds S, Kay JD, Ma hews L, Haase K, e al. (2017) App aising he ele ance o DNA copy numbe loss and gain in p os a e cance using whole genome DNA sequence da a. PLoS Gene 13(9): e1007001. h ps://doi.o g/10.1371/jou nal.pgen.1007001 Edi o : Rameen Be oukhim, B oad Ins i u e, UNITED STATES Recei ed: Feb ua y 15, 2017 Accep ed: Augus 28, 2017 Published: Sep embe 25, 2017 Copy igh : ©2017 Camacho e al. 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, 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 au ho and sou ce a e c edi ed. Da a A ailabili y S a emen : All whole genome sequencing iles and SNP6 iles a e a ailable om he Eu opean Genome-phenome A chi e (da abase accession numbe EGAS00001000262). Funding: We acknowledge suppo om Cance Resea ch UK (C5047/A22530, C309/A11566, C368/A6743, A368/A7990, C14303/A17197) and he Dallaglio Founda ion. We also acknowledge suppo om he Na ional Ins i u e o Heal h Resea ch (NIHR) (The Biomedical Resea ch Cen e a The Ins i u e o Cance Resea ch & The Royal Abs ac A a ie y o models ha e been p oposed o explain egions o ecu en soma ic copy num- be al e a ion (SCNA) in human cance . Ou s udy employs Whole Genome DNA Sequence (WGS) da a om umo samples (n = 103) o comp ehensi ely assess he ole o he Knud- son wo hi gene ic model in SCNA gene a ion in p os a e cance . 64 ecu en egions o loss and gain we e de ec ed, o which 28 we e no el, including egions o loss wi h mo e han 15% equency a Ch 4p15.2-p15.1 (15.53%), Ch 6q27 (16.50%) and Ch 18q12.3 (17.48%). Comp ehensi e mu a ion sc eens o genes, lincRNA encoding sequences, con- ol egions and conse ed domains wi hin SCNAs demons a ed ha a wo-hi gene ic model was suppo ed in only a mino p opo ion o ecu en SCNA losses examined (15/ 40). We ound ha ecu en b eakpoin s and egions o in e sion o en occu wi hin Knud- son model SCNAs, leading o he iden i ica ion o ZNF292 as a a ge gene o he dele ion a 6q14.3-q15 and NKX3.1as a wo-hi a ge a 8p21.3-p21.2. The impo ance o al e a ions o lincRNA sequences was illus a ed by he iden i ica ion o a no el mu a ional ho spo a he KCCAT42,FENDRR,CAT1886 and STCAT2 loci a he 16q23.1-q24.3 loss. Ou da a con i m ha he bu den o SCNAs is p edic i e o biochemical ecu ence, de ine nine indi- idual egions ha a e associa ed wi h elapse, and highligh he possible impo ance o ion channel and G-p o ein coupled- ecep o (GPCR) pa hways in cance de elopmen . We con- cluded ha a wo-hi gene ic model accoun s o abou one hi d o SCNA indica ing ha mechanisms, such haploinsu iciency and epigene ic inac i a ion, accoun o he emaining SCNA losses. Au ho summa y Cance is a gene ic disease whe e changes in DNA cause al e a ions in he con ol o cellula sys ems leading o unchecked g ow h. Copy numbe changes, including dupli- ca ions, ampli ica ions, and dele ions, a e a common ype o DNA change obse ed in cance cells bu i is no always clea which o he changes a e impo an in d i ing can- ce de elopmen . We ha e examined his class o gene ic al e a ion in p os a e cance by DNA sequencing he whole genome in 103 cance s. 64 ecu en copy numbe changes we e de ec ed, o which 28 we e new. Fo gene ic losses ou s udy comp ehen- si ely assessed he ole o a model called he “Knudson wo-hi gene ic model” whe e al e a ions in bo h alleles o a gene is equi ed o gene a e unc ional al e a ions. This model was only suppo ed a mino p opo ion o ecu en dele ions (15/40). This obse a ion indica es ha o he mechanisms, such haploinsu iciency and epigene ic inac i a ion, may accoun o he majo i y o dele ions. Ou s udies highligh se e al no el changes including hose in non-coding lincRNA sequences, he iden i ica ion ZNF292as a a ge gene o a ecu en dele ion on ch omosome 6, and he common Knudson dele ions a he NKX3.1loci on ch omosome 8. In oduc ion Soma ic copy-numbe al e a ions (SCNAs) occu e y equen ly in human cance and exac ly how hese al e a ions con ibu e o cance de elopmen is a subjec o conside able in e es . App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 2 / 28 Ma sden NHS Founda ion T us and he p ojec "P os a e Cance : Mechanisms o P og ession and T ea men (PROMPT)" [G0500966/75466]). We hank he Wellcome T us , Bob Champion Cance T us , The O chid Cance appeal, The RoseT ees T us , The No h Wes Cance Resea ch Fund, Big C, The King amily, and The Masonic Cha i able Founda ion o unding. This esea ch is suppo ed by he F ancis C ick Ins i u e which ecei es i s co e unding om Cance Resea ch UK (FC001202), he UK Medical Resea ch Council (FC001202), and he Wellcome T us (FC001202). The unde s had no ole in s udy design, da a collec ion and analysis, decision o publish, o p epa a ion o he manusc ip . Compe ing in e es s: I ha e ead he jou nal’s policy and he au ho s o his manusc ip ha e he ollowing compe ing in e es s: RE has ecei ed educa ional g an s om Illumina and GenP obe ( o me ly Tepnel), Vis a Diagnos ics and Janssen Pha maceu icals, as well as hono a ia om Succin Communica ions o alks on p os a e cance gene ics. Mapping o SCNAs has iden i ied ecu en si es o al e a ions in many cance ypes, bu only a small p opo ion o such si es ha e unambiguously been assigned o speci ic cance genes [1]. Se e al models based on clonal e olu ion and selec ion may be in oked o explain ecu en egions o ch omosomal loss in he au osomes and sex ch omosomes. In he classic model o cance de elopmen p oposed by Knudson [2], mu a ions a e equi ed in each o he wo cop- ies o a single gene: he loss o an allele is conside ed as one mu a ion and he emaining allele would be al e ed by loss (homozygous dele ion), mu a ion o ea angemen . This is due o s ong posi i e selec i e p essu e o al e a ion o bo h alleles. In p inciple, inac i a ion o he emaining allele migh also in ol e epigene ic inac i a ion h ough DNA me hyla ion. Hap- loinsu iciency is also an es ablished mechanism o cance de elopmen , whe e loss o only a single allele is equi ed o cance de elopmen ; exp ession o he no mal allele is e ained in he cance , albei a a lowe le el. This model is suppo ed in p os a e cance by ansgenic mouse s udies o NKX3.1and p27 Kip1 [3,4]. Solimini e al. [5] ha e p o ided a model, whe e he collec i e con ibu ion o many genes may p o ide selec i e ad an age o a cance cell, possibly o e lapping wi h he Knudson model. De e al. and Fudenbe g e al. ha e p oposed a model whe e 3D ch oma in o ganiza ion and spa ial co-localiza ion o DNA egions du ing eplica ion may explain he gene a ion o copy numbe al e a ions [6,7]. Ano he explana ion is ha al e a ions a e simply he hallma k o an uns able genome and ha e no pa icula unc ional signi icance, o example e lec ing agile genomic si es. Simila conside a ion can be gi en o egions o gene ic gain and ampli i- ca ion whe e o e exp ession o one o mo e genes is belie ed in many cases o d i e cance de elopmen [8]. P os a e cance is he second mos common cance in men wo ldwide and in 2012 an es i- ma ed 307,000 men died om p os a e cance wo ldwide [9]. Se e al s udies ha e in es iga ed SCNA in p os a e cance [10–12], and c i ically, i has been es ablished ha he bu den o SCNA is associa ed wi h subsequen biochemical ecu ence ( ising P os a e Speci ic An igen, PSA, le - els a e adical p os a ec omy) and me as asis independen o ini ial PSA le els and Gleason sco es. Simila obse a ions we e epo ed in a ecen s udy whe e a 100-loci (276 genes) copy numbe signa u e was p edic i e o biochemical ecu ence [13]. Howe e , only a small p opo - ion o he gains and losses ha e been unambiguously assigned o speci ic d i ing genes. The cu en s udy is he i s o implemen a a ge ed app oach in p os a e cance whe e Whole Genome DNA Sequencing (WGS) da a is used o comp ehensi ely examine mu a ion da a in ela ion o he p esence o SCNAs. We used WGS da a om each pa ien o sc een ecu en SCNA egions o po en ially unc ional al e a ions no only in p o ein coding genes, bu also in genomic egions encoding lincRNAs, in con ol egions, and in o he con- se ed DNA sequences. Unde s anding mechanisms o SCNA gene a ion and he iden i ica- ion o a ge genes, linked o clinical ou comes, may assis in iden i ying no el bioma ke s and he apeu ic a ge s. Resul s Genome wide copy numbe p o iles ASCAT 2.2 [14,15] was used o iden i y soma ic copy numbe al e a ions (SCNAs; S2 Table; ypes o al e a ion de ined in S1 Table) in WGS da a o malignan samples aken om 103 p os a e cance pa ien s (S2 Table). In cases o pa ien s wi h mul iple umo samples a single p o ile was used. The ela ionship be ween samples wi hin a pa ien and he umou e olu ion o he 13 pa ien s whe e he e we e mul iple umou samples has been p e iously in es iga ed [16,17]. ETS gene s a us was in e ed om WGS da a (S3 Table). App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 3 / 28 When compa ed o p os a ec omy cases, me as a ic cance s had highe p opo ions o e - aploid genomes as de ined by ASCAT (53% s 19%; Fishe ’s exac es , p= 0.0042; S4 Table), signi ican ly la ge numbe s o SCNAs (mean 83 s 21; Mann-Whi ney U p = 2.17x10 -08 ), highe copy numbe bu den (pe cen age o he genome al e ed) (mean 31.24% s 7.694%; Mann-Whi ney U p = 1.30x10 -08 ), and longe a e age SCNA size (Mann-Whi ney U p = 4.37x10 -04 ). Pa ien s wi h mo e han 44 SCNAS ( he mean numbe o SCNAs in pa ien s ha had p og essed wi hin six mon hs) had wo se p ognosis (Log- ank es p= 0.027, median ol- low up o 16.5 mon h; Fig 1A and 1B) han pa ien s wi h ewe SCNAs, consis en wi h o he publica ions [13,18,19] linking highe numbe s o SCNAs o poo e ou come. Pa ien s wi h highe copy numbe bu den had wo se p ognosis (Log- ank es p= 0.023; Fig 1C and 1D). Classi ica ion o p os a e cance based on copy numbe p o iles Hie a chical clus e ing analysis was applied on o e lapping SCNAs p esen in a leas i e pa ien s. A bina y ma ix was cons uc ed on he basis o a pa ien ha ing a egion o ampli ica- ion o dele ion (1) o no (0). Fi e majo clus e s we e obse ed: C1-C5 (Fig 2;S5 Table). C5 and C4 we e composed mainly o me as a ic cases (6/13 and 5/7) while he emaining me as a ic cases we e placed in C2 (3/39) and C3 (2/16). Pa ien s in C1 had ewe SCNAs han hose in C2-C5 (mean 6.79 s 40.32 SCNAs) and lowe copy numbe bu den (mean 1.64% s 14.98%). P os a ec omy/TURP pa ien s in C1 had no signi ican di e ence in Gleason Sco es (X 2 es p= 0.213; Fig 3A) o in hei le els o PSA a diagnosis (Mann-Whi ney U p = 0.929). P os a ec- omy pa ien s in C1 had a signi ican ly be e p ognosis (Log- ank es p= 0.028; Fig 3B). De ec ing egions o ecu en al e a ion We de ined a ecu en egion o al e a ion as he minimal egion o o e lap ha con ains suppo ing al e a ions om i e o mo e pa ien s (minimal MRA, Fig 4A). 40 egions o dele- ion and 24 egions o gain we e iden i ied (S6 Table). To iden i y po en ial d i e genes, we Fig 1. Soma ic copy numbe al e a ions in 103 pa ien s. (A) Soma ic copy numbe al e a ions in elapse ee (n = 59) and p og essed pa ien s (n = 10) a e adical p os a ec omy a six mon hs and (B) associa ion wi h ime o biochemical ecu ence ( he wo g oups we e de ined ela i e o he mean numbe o SCNAs in p og essed pa ien s (44 SCNAs) wi hin he coho ). P og essed pa ien s had signi ican ly mo e SCNAs han elapse ee pa ien s (a mean o 19 compa ed o 44; Mann-Whi ney U p = 0.0133). (C) Copy numbe bu den in elapse ee and p og essed pa ien s a e adical p os a ec omy a six mon hs and (D) associa ion wi h ime o biochemical ecu ence ( he wo g oups we e de ined ela i e o he mean copy numbe bu den in p og essed pa ien s wi hin he coho ). P og essed pa ien s had lowe copy numbe bu den (mean o 7.359 compa ed o a mean o 11.710; Mann-Whi ney U p = 0.0166). h ps://doi.o g/10.1371/jou nal.pgen.1007001.g001 App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 4 / 28 sc eened sequencing da a o de e mine whe he each MRA con ained poin subs i u ions, inse ions and/o dele ions wi hin he coding egion o genes (Table 1,S7 Table), in cance - ela ed and conse ed lincRNAs om MiT ansc ip ome [20] (S8 Table), in p omo e egions (S9 Table) and in DNA High-occupancy a ge (HOT) egions [21] (S10 Table). To educe he e ec o ou lie measu emen s in ou da a, when sc eening o mu a ions we used a la ge minimal common egion o al e a ion, e e ed o as he ex ended MRA, which is de ined by emo ing he wo al e a ions ha we e closes o he 5’ bounda y o he minimal MRA and simila ly he wo al e a ions closes o he 3’ bounda y (S1 Fig,S6 Table). We also applied he s a is ical me hod GISTIC [22] o de ine signi ican egions o gain and loss (15 gains and 19 losses, esidual q<0.05, S11 Table). 14 losses and one gain we e common o ou lis o mini- mal MRAs and 5 losses and 14 gains we e only de ec ed by GISTIC (S6 Table). 16 ou o 34 sig- ni ican GISTIC egions ( esidual q<0.05) we e also de ec ed in o he p os a e cance s udies ha applied GISTIC [23,24], only ou o which we e no de ec ed by ou app oach. Fig 2. Clus e dendog am and hea map o soma ic copy numbe al e a ions in 103 pa ien s. Clus e dendog am and hea map o ampli ica ions ( ed) and dele ions (blue) ep esen ing he soma ic copy numbe p o ile o he 103 pa ien s calcula ed using unsupe ised hie a chical clus e ing wi h Manha an dis ance and comple e in e -clus e linkage. Samples wi h SNP6.0 da a a e indica ed wi h an as e isk. Con idence in e als de e mined by mul iscale boo s ap esampling a e displayed in S2 Fig. h ps://doi.o g/10.1371/jou nal.pgen.1007001.g002 App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 5 / 28 Known egions o gene ic al e a ion The mos equen MRAs we e loca ed a 8p21.3-p21.2 (60% loss), 6q15 (49%), 13q14.13 (46%) and 10q23.31 (39%), in ag eemen wi h p e ious s udies [18,24–26] (Fig 5,S3 Fig). 8p21.3-p21.2 is an example o a loss whe e haploinsu iciency has been p oposed as he p i- ma y mechanism. The minimal MRA con ains 16 coding genes, including NKX3.1, a sugges ed a ge gene o his egion [24,25] (Fig 5B). The lincRNA KCCAT306 (MiT ansc ip iome ID) was he only sequence mu a ed mo e han once (S8 Table). NKX3.1was howe e , a ec ed by a homozygous dele ion in h ee pa ien s (S12 Table). 1.1MBp om he minimal MRA he e is a egion a ec ed by homozygous loss in i e pa ien s con aining PPP2R2a,EBF2,BNIP3L, PNMA2 and DPYSL2.BNIP3L, which coun e ac s he apop o ic induce BNIP3, is ano he p oposed a ge [26]. By compa ison, 10q23.31 p o ided an example whe e al e a ions in bo h alleles a e obse ed, consis en wi h he Knudson wo-hi model [2], wi h PTEN as he a ge . A high p o- po ion o he dele ions we e homozygous losses (11/39, S12 Table,Fig 5C). In se en cases when one allele was dele ed, a mu a ion was ound a he emaining PTEN allele (Table 1). PTEN al e a ions we e mo e common in me as a ic disease han in he p os a ec omy se ies (93% s 29%; Fishe ’s exac es , p= 1.0x10 -08 ). Consis en wi h p e ious s udies [27,28], some- imes only he 5’ end o he PTEN gene was los . 17p13.1 (35%) is ano he example o his class (S2T Fig,Table 1) whe e TP53 was a ec ed by mu a ion in nine pa ien s who all had an associ- a ed SCNA and homozygous loss was seen in wo pa ien s. MAP3K7, a p oposed a ge o he 6q14.3-q15 dele ion [25,26] is loca ed 2.4Mbp away om he ex ended MRA (Fig 5A). ZNF292was mu a ed on he emaining allele in one pa ien and has been ound o be ea anged in p os a e cance [29]. The lincRNAs KHCAT274 and KHCAT370 (MiT ansc ip iome IDs) in he ex ended MRA we e mu a ed ou imes on he allele ha emained a e dele ion (Fig 5A,S8 Table). The known dele ion a 17q21.31 [18,23,26] (Fig 5F) had a minimal MRA ha con ained UBTF, a known usion pa ne o ETV4[30]. Fig 3. Clus e associa ions wi h clinical ac o s. (A) Clus e s associa ion wi h Gleason Sco e (p os a ec omies and TURP samples only). (B) Clus e s associa ion o isk o biochemical ecu ence wi hin C1 and C2-C5. A he ime o w i ing he single p os a ec omy case in C5 had a ollow up ime o 23 mon hs and had no p og essed. h ps://doi.o g/10.1371/jou nal.pgen.1007001.g003 App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 6 / 28 Fig 4. F equency plo ep esen ing egions o ampli ica ion ( ed) and dele ion (blue). Ac oss samples om (A) all 103 pa ien s, (B) ETS- posi i e cance s, and (C) ETS-nega i e cance s. SCNAs p esen in mo e han one pa ien we e plo ed. The g een lines ep esen he cu -o (5 pa ien s) o de ining ecu en ly al e ed egions. Ch omosome numbe s a e indica ed in black. Genes po en ially in ol ed in p os a e cance de elopmen in he minimal egions o al e a ion a e anno a ed in black. h ps://doi.o g/10.1371/jou nal.pgen.1007001.g004 App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 7 / 28 RB1 and BRCA2 a e wo p oposed candida es o he MRA a Ch omosome 13q14.13 (Fig 5D) [18,25,26]. RB1 is loca ed 1.3MBp om he minimal MRA and wi hin he ex ended MRA. In h ee pa ien s, a egion o homozygous loss spans RB1 (S12 Table). BRCA2 was 9.5MBp ou - side he ex ended MRA. Dele ions con aining only one gene in he minimal MRA, we e de ec ed a 3p13 con aining RYBP (33.01%, S2D Fig), 12p13.1 con aining CDKN1B(27.18%, S2N Fig), and 14q24.1 con- aining RAD51B(10.68%, Fig 5E) bu none had mu a ions in he emaining allele (Table 1,S6 Table). Gains a Ch 8q exhibi ed a complex s uc u e wi h h ee b oad peaks (Fig 6A,S4 Fig). The p oposed a ge o 8q24.21, MYC, was loca ed 48kb om he minimal MRA and was p esen in he ex ended MRA. Thi een mu a ions we e ound in o al a PCAT1 and CCAT1 wi h se en accompanying ch omosome gain (S8 Table). The ampli ica ion a 7p11.2, p esen in 11.65% o pa ien s, con ained EGFR in he minimal MRA (Fig 6B). The ex ended MRA egion con ained a o al o se en mu a ions wi h he highes numbe in lincRNA CAT941. Ampli ica- ion o 14q13.3-q21.1 (Fig 6C) con ained only FOXA1 and MIPOL1. No el egions o ecu en gene ic al e a ion 24 addi ional no el ecu en egions o loss and gain we e de ec ed (S6 Table,S1 Appendix), including egions o loss wi h mo e han 15% equency a 4p15.2-p15.1 (15.53%), 6q27 (16.50%) and 18q12.3 (17.48%). Regions o gain wi h he highes equencies we e a 1q25.3 (8.74%), 5p15.31 (9.71%), and 10q21.1-q21.3 (8.74%). Fo wo o hese egions, 22q12.1-q12.2 Table 1. Genes ha a e ecu en ly al e ed by coding-changing mu a ions in egions o ecu en soma ic copy numbe al e a ion de ined by he ex ended MRAs. Gene Gene name Numbe o mu a ions Pa ien s wi h a SCNA and mu a ion Pa ien s wi h SCNA (%) Type o SCNA FOXA1 (14q21.1) Fo khead box p o ein A1 11 1 6 (5.83) Amp TP53 (17p13.1) Tumo p o ein p53 9 9 37 (35.92) Del PTEN (10q23.31) Phospha ase and ensin homolog 9 7 39 (37.86) Del HMCN1 (1q25.3) Hemicen in-1 6 1 8 (7.77) Amp KIF26B (1q44) Kinesin amily membe 26B 4 2 7 (6.80) Amp KIAA1614 (1q25.3) Uncha ac e ized P o ein KIAA1614 4 1 8 (7.77) Amp TCF12 (15q21.3) T ansc ip ion ac o 12 3 1 15 (14.56) Del CACNA1E (1q25.3) Calcium channel, ol age-dependen , R ype, alpha 1E subuni 3 1 8 (7.77) Amp PHYKPL/AGXT2L2 (5q35.3) 5-Phosphohyd oxy-L-lysine phospho-lyase 3 0 8 (7.77) Amp SCN4A (17q23.3) Sodium channel p o ein ype 4 subuni alpha 3 0 4 (3.88) Amp C17o 58 (17q24.2) Ch omosome 17 open eading ame 58 3 0 4 (3.88) Amp RYR2 (1q43) Ryanodine ecep o 2 3 0 12 (11.65) Amp ZFHX4 (8q21.11) Zinc inge homeobox 4 2 2 31 (30.10) Amp TRPA1 (8q13.3) T ansien ecep o po en ial ca ion channel, sub amily A, membe 1 2 2 31 (30.10) Amp NBN/NBS1 (8q21.3) Nib in 2 2 32 (31.07) Amp COL27A1 (9q32) Collagen, ype XXVII, alpha 1 2 1 5 (4.85) Amp AMBP (9q32) Alpha-1-mic oglobulin/Bikunin p ecu so 2 1 5 (4.85) Amp UIMC1 (5q35.2) Ubiqui in in e ac ion mo i con aining 1 2 1 6 (5.83) Amp SLC26A2 (5q32) Solu e ca ie amily 26 (anion exchange ), membe 2 2 1 8 (7.77) Amp h ps://doi.o g/10.1371/jou nal.pgen.1007001. 001 App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 8 / 28 Fig 5. Examples o minimal ecu en dele ions. Dele ions a (A) 6q14.3-q15, (B) 8p21.3-p21.2, (C) 10q23.31, (D) 13q14.13, (E) 14q24.1 and (F) 17q21.31. The genomic loca ion o he MRA and linked genes a e displayed. Fo he loss a 8p a ecu en egion o homozygous loss (8:25417422–26386565) close o he MRA is indica ed. h ps://doi.o g/10.1371/jou nal.pgen.1007001.g005 App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 9 / 28 Me hods Sample coho 141 malignan samples om 105 pa ien s wi h p os a e cance wi hin he Cance Resea ch UK ICGC P os a e Cance P ojec we e included in his s udy: 88 cance s collec ed a e adical p os a ec omy, including 10 umo and h ee blood samples om h ee men wi h complex o mul i ocal disease; i e cance s om men diagnosed wi h p os a e cance in Shanghai; wo cance s collec ed om ansu e h al esec ion o he p os a e (TURP); 8 lymph node me a- s a ic samples om non-and ogen dep i ed pa ien s; and 36 malignan samples om 7 men wi h me as a ic disease belonging o he PELICAN apid au opsy p og am [17]. Samples we e collec ed subjec o ICGC s anda ds o e hical consen . E hical app o al o his wo k was ob ained om he espec i e local e hics commi ees (NHS Sou h Wes London REC [10/ H0806/113]; NHS Eas Midlands–De by REC [01/4/061]; NHS Eas o England—Camb idge REC [03/018]; John Hopkins IRB [NA_00003925]; Changhai Hospi al E hics commi ee; Join Chinese Uni e si y o Hong Kong-New Te i o ies Eas Clus e Clinical Resea ch E hics Com- mi ee [CRE-2011.373]) and om The T en Mul icen e Resea ch E hics Commi ee [MREC/ 01/4/061]. Explici in o med w i en consen was ob ained om all pa icipan s o he use and s o age o hei gene ic ma e ial and issue samples in esea ch, including whole genome sequencing. Explici w i en consen was also ob ained o any da a de i ed om hese sam- ples, along wi h coded clinical/heal h da a, o be placed on an in e na ional da abase (ICGC) and made a ailable o ICGC membe s and o he in e na ional esea che s unde open o con- olled access. P os a ec omy samples we e collec ed as desc ibed p e iously [65]. Blood sam- ples we e used as no mal con ols excep o PELICAN samples whe e no mal issue was used. DNA p epa a ion and DNA sequencing DNA om whole blood samples and ozen issue was ex ac ed and quan i ied using a ds- DNA assay (UK-Quan -iT PicoG een dsDNA Assay Ki o DNA) ollowing manu ac u e ’s ins uc ions wi h a Fluo escence Mic opla e Reade (Bio ek Syne gyHT, Bio ek). Accep able DNA had a concen a ion o a leas 50ng/μl in TE (10mM T is/1mM EDTA), wi h an OD 260/280 be ween 1.8–2.0. Fo aCGH a leas 10μl equi alen o 500ng we e used o hyb idisa- ion o he A yme ix SNP6.0 a ay. WGS was pe o med a Illumina, Inc. (Illumina Sequenc- ing Facili y, San Diego, CA USA) o he BGI (Beijing Genome Ins i u e, Hong Kong) as desc ibed p e iously o a a ge dep h o 50X o he umo samples and 30X o ma ched con- ols [66]. The Bu ows-Wheele Aligne (BWA) was used o align he sequencing da a o he GRCh37 human genome [66]. Gene a ion o segmen ed copy numbe p o iles Pseudo-SNP6.0 p o iles we e c ea ed o each umo and ma ched con ol om whole genome sequencing da a. In a small numbe o cases da a was ob ained di ec ly om A yme ix SNP6.0 a ays (S18 Table) and escaled and ans o med using PennCNV [67]. ASCAT 2.2 was hen used o gene a e segmen ed copy numbe p o iles and es ima e he pu i y and ploidy o he sample using a penal y o 50 and gamma alue o 1 [14,15]. No e ha gi en he a ge sequencing co e age o 50x/30x, he e may be s ochas ic egions o low co e age whe e ue segmen s a e no de ec ed. Two p os a ec omy samples wi h low es ima ed umo con en we e emo ed om u he analysis. Da a om he ma ched con ol is used o educe alse posi i es. A conse a i e il e ing s a egy was used o ensu e he bes quali y o segmen s a e ob ained and gi e us he g ea es con idence in he minimal egions o al e a ion ha we call. App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 16 / 28 8876 segmen s we e emo ed, o which 6692 we e emo ed based on he ollowing il e ing c i e ia: • Segmen was wi hin cen ome ic o elome ic egions (ex ended by 1,000,000 bp). • Segmen was wi hin low mapabili y egions as de ined by ENCODE (ex ended by 1,000,000 bp). • Segmen copy numbe o majo allele (nMajo ) >25 and copy numbe o majo allele (nMino ) = = 0, o nMajo >= 10 and leng h was less han 500,000bp. • Segmen whe e he mino allele copy numbe is la ge han he majo allele copy numbe . • Segmen had less han 50 suppo ing SNPs and leng h was less han 100,000bp The emaining 2184 segmen s we e il e ed ou upon isual inspec ion in a blind ashion. Visual inspec ion included checking ha he change in copy numbe was in ag eemen wi h he o e all ploidy o he sample, ha he e was a change in BAF and o al copy numbe , and ha he SNPs wi hin he segmen didn’ ha e high a iance. In he majo i y o cases segmen s we e emo ed because ASCAT did no ge he app op ia e i and he e was no ob ious copy numbe change. In his and o he ICGC p ojec s we ha e gene ally ound ha isual inspec- ion o any class o al e a ion is an excellen me hod o il e ing ou inco ec calls made by he algo i hms. Simila esul s we e ob ained om SNP6.0 and pseudo-SNP6.0 p o iles (S6 Fig). We also applied he s a is ical me hod GISTIC 2.0.16 [22] o de ine signi ican egions o gain and loss, using he de aul pa ame e s. No e ha all p e- il e ed segmen s we e used when he GISTIC algo i hm was applied. Con e sion o sequencing da a o a pseudo-SNP6.0 p o ile The numbe o eads o each nucleo ide base a each p obe posi ion on he A yme ix SNP6.0 na32 hg19 anno a ion was ex ac ed om sequence using deepSNV [68]. The loga- i hm o he a io (LogR) be ween allele A and B and he B allele equency (BAF) we e calcu- la ed as ollows: LogRi¼log2ðAiþBiÞ BAF ¼Bi ðAiþBiÞ whe e A i is he ead coun o allele A o p obe iand B i is he ead coun o allele B. Fo CN p obes: LogRi¼log2ðNiÞ whe e N i is he o al numbe o eads a he posi ion o p obe i. The LogR alues we e no mal- ised: LogRi¼LogRimedianðLogRÞ SNP p obes wi h no eads a A and B alleles o wi h a BAF o 1 o 0 we e emo ed. Using he SNP6 a ay p obe posi ions o WGS da a wi h ASCAT is a common app oach, used success- ully in he ICGC pancance analysis o whole genomes p ojec (h ps://dcc.icgc.o g/pcawg) [69]. A w appe p og am o ease o use is a ailable [70]. App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 17 / 28 Calcula ion o he minimal egions o al e a ion (MRA) Da a om pa ien s wi h mul iple umo samples is collapsed in o one copy numbe p o ile pe pa ien using all de ec ed egions. O e lapping egions o SCNAs we e iden i ied ac oss all pa ien s by applying he GenomicRanges co e age unc ion [71]; ampli ica ions and dele ions we e analysed independen ly. The minimal MRA is de ined by iden i ying he egion mos e- quen ly al e ed in each peak egion o o e lap be ween i e o mo e pa ien s. This is an a bi- a y h eshold, bu simpli ying segmen calling o a andom p ocess and assuming ha a copy numbe will be called in a egion 1% o he ime by chance (which we would conside high), hen gi en i e pa ien s, i is signi ican ly likely ha he ue numbe o segmen s is g ea e han 1% (p= 0.0039; Exac binomial es ). The ex ended MRA expands he minimal MRA by aking he s a posi ion o he o e lapping segmen wi h he hi d la ges s a posi ion and he end posi ion o he segmen wi h he hi d smalles end posi ion (S1 Fig). Mu a ional sc een Subs i u ions, inse ions and dele ions we e de ec ed using he Cance Genome P ojec Well- come T us Sange Ins i u e pipeline. An upda ed e sion o his pipeline is a ailable as a Docke image (Alignmen : h ps://docks o e.o g/con aine s/quay.io/w sicgp/docks o e-cgpmap; Va i- an -calling: h ps://docks o e.o g/con aine s/quay.io/w sicgp/docks o e-cgpwgs). The Bu ows- Wheele Aligne ’s Smi h-Wa e man Alignmen (BWA-SW) 0.5.9- 16+ ugo was used o align he sequencing da a om each lane o he GRCh37 e e ence human genome using pa ame e s -l 32 - 6 [72]. Lanes ha pass quali y con ol a e me ged in o a single well-anno a ed sample BAM ile wi h PCR duplica e eads emo ed. Subs i u ions we e de ec ed using CaVEMan 1.3, an in-house bespoke algo i hm de eloped a he Sange Ins i u e (h p://cance i .gi hub.io/ CaVEMan/), wi h a cu -o ‘soma ic’ p obabili y o 95%. CaVEMan u ilises a Bayesian expec a- ion maximiza ion (EM) algo i hm: Gi en he e e ence base, copy numbe s a us and ac ion o abe an umo cells p esen in each cance sample, CaVEMan gene a es a p obabili y sco e o po en ial geno ypes a each genomic posi ion. Fu he pos -p ocessing il e s we e applied o elimina e alse posi i e calls a ising om genomic ea u es ha gene a e mapping e o s and sys- ema ic sequencing a i ac s. In compa isons wi h o he mu a ion calle s i has been ound o be amongs he op pe o me s in e ms o sensi i i y and speci ici y [73]. Only subs i u ions ha we e a missense, nonsense, s a -los , o occu ed in a non-coding ansc ip we e conside ed. Inse ions and dele ions we e called using a ligh ly modi ied e sion o pindel 4.2 [74] (h p:// cance i .gi hub.io/cgpPindel/). Only indels ha we e in ame, ameshi , o occu ed in a non- coding ansc ip we e conside ed. S uc u al a ian s we e de ec ed using B ass (B eakpoin s ia assembly) 1.0.3, an in-house bespoke algo i hm de eloped a he Sange Ins i u e (h ps:// gi hub.com/cance i /BRASS). In b ie , he i s s ep is o combine disco dan ead pai s in o po en ial egions whe e a b eakpoin migh occu . Nex , eads a ound each po en ial egion, including hal -unmapped eads, a e ga he ed and a local de no o assembly using Vel e is pe - o med [75]. By analysing he De B uijn g aph pa e n he b eakpoin can be iden i ied down o base pai esolu ion. Any b eakpoin s whe e an exac loca ion could no be de e mined we e emo ed. A posi i e ETS s a us was assigned i a b eakpoin be ween ERG,ETV1 o ETV4 and p e iously epo ed pa ne DNA sequences [76] we e de ec ed. Genes in each MRA we e iden- i ied using good quali y ansc ip s anno a ed in he EnsEMBL Genome Re e ence Conso ium Human Build 37.p13 (GRCh37.p13). The mu a ional sc een o homozygous loss was pe o med in any egion ha had a homozy- gous loss in a leas one sample. Fo mu a ions ha occu wi hin p omo e s, EPDNew human e sion 003 [77] was used o de ine p omo e egions. Conse ed egions o DNA we e de e - mined om UCSC phas Cons sco es o mul iple alignmen s o 45 e eb a e genomes o he App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 18 / 28 hg19 human genome using a h eshold p- alue o 0.95. We used 7,942 long noncoding RNAs (lincRNAs) om he MiT ansc ip ome p ojec ha we e de ec ed in 7,256 RNA sequencing lib a ies om umo s, no mal issues and cell lines and de ined as being ei he cance -associ- a ed o con aining conse ed egions [20]. S a is ical es s and su i al analyses All s a is ical es s we e pe o med in R [78]. Fo compa isons be ween g oups he non- pa ame ic Mann-Whi ney U es was used o con inuous a iables and he X 2 es wi h Ya es’ con inui y co ec ion o Fishe ’s exac es was used o ca ego ical a iables. Fo all s a is ical es s used he e is he assump ion o independence be ween da a and we ha e ensu ed his is he case by only using one copy numbe p o ile pe pa ien in all compa i- sons. The sample used o he copy numbe p o ile was chosen a andom (see S2 Table o selec ion). In all cases whe e he X 2 es was applied no cells in he con ingency able had an expec ed alue less han 5. Clinical associa ions we e de e mined using s anda d s a is ical es s wi h Benjamin-Hoch- be g mul iple es ing co ec ion applied. The log- ank es was used o pe o m su i al analy- ses wi h biochemical ecu ence a e p os a ec omy as he end poin o su i al analyses. Reasonable assump ions we e made i.e. censo ing is un ela ed o p og ess, su i al p obabili- ies a e he same o subjec s ec ui ed ea ly and la e in he s udy, and e en s happen a he imes speci ied. 84 ou o 86 p os a ec omy pa ien s we e used; wo pa ien s had incomple e clinical da a (0040 and 0052). Fo he pa ien s whe e da a was a ailable om mul iple samples, he pa ien was classi ied as ha ing he minimal egion o al e a ion i i occu ed in any o he samples. Pa hway analysis The en ichmen analysis was pe o med by explo ing he Reac ome canonical pa hways using he Reac ome plugin in Cy oscape [79] using he genes con ained wi hin he ex ended MRAs. In eg a ion o Reac ome and Kyo o Encyclopaedia o Genes and Genomes (KEGG) canonical pa hways pa hway analysis [80] was used o cons uc a map o he majo pa hways al e ed in p os a e cance wi hin ou coho . Iden i ying genes a ec ed by ea angemen s Rea angemen b eakpoin s wi hin he ex ended MRA and wi hin he signi ican GISTIC egions (q- alue <= 0.05) we e selec ed and used o de ine associa ed gene allele-speci ic e en s. Dele ions ha had common egions de ec ed by ASCAT2.2 and B ass o by ASC AT2.2 only we e included in he analysis. A gene was de ined as being a ec ed by a ea - angemen , and hence a bi-allelic e en , i an inse ion, in e sion o in a-ch omosomal b eakpoin occu ed wi hin he gene egion o he gene occu ed wi hin an in e sion. I a gene was in e ed wice and he b eakpoin s o he ea angemen we e ou side he gene coo dina es i was assumed he gene was s ill unc ional; hese e en in e sions we e e- mo ed om he analysis. Da a access Sequencing da a has been deposi ed a he Eu opean Genome-phenome A chi e (EGAS00001000262, S1 Table). App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 19 / 28 Suppo ing in o ma ion S1 Fig. Diag amma ic explana ion on minimal and ex ended MRA. (PDF) S2 Fig. Hie a chical clus e ing o 103 p os a e cance samples wi h mul iscale boo s ap esampling. The da a a e bina y alues co esponding o he p esence/absence (1/0) o egions o copy numbe gain and loss in each o he umou samples. p- alues we e calcula ed ia hie - a chical clus e analysis wi h mul iscale boo s ap esampling o 1000 using Wa d’s me hod and he Manha an dis ance. The analysis was pe o med using he p clus package in R. Val- ues a b anches a e AU (App oxima ely Unbiased) p- alues (le , ed), and BP (Boo s ap P obabili y) alues ( igh , g een). Clus e s signi ican ly suppo ed by he da a (AU 95) a e indica ed by he ed ec angles. (PDF) S3 Fig. Copy numbe al e a ion segmen s de ec ed by ASCAT ha o e lap wi h dele ion MRAs. Each dele ion is ep esen ed as a dis inc colou as shown in he key. Dele ions a e as ollows: neu al LOH (loss o one allele wi h duplica ion o he emaining allele); hemizygous dele ion LOH (loss o one allele); homozygous loss (loss o he wo alleles); and o he loss (loss o one allele copy ollowing whole genome duplica ion). Cases o which mo e han one sam- ple was a ailable a e all indica ed, howe e con ibu ion o he equency o he SCNA was de ined on a pe pa ien basis. The egions o SCNA a e o de ed by leng h: op-smalles , bo - om-la ges . Each block has been labelled wi h sample ID. (a) ch 1 p31.1, (b) ch 1 q42.2-q42.3, (c) ch 2 q21.3-q22.1, (d) ch 3 p13, (e) ch 4 p15.2-p15.1, ( ) ch 4 q22.3, (g) ch 4 q27-q28.1, (h) ch 4 q34.3, (i) ch 5 q13.1-q13.2, (j) ch 7 q31.32-q31.33, (k) ch 9 p22.3, (l) ch 11 p13, (m) ch 11 q23.2, (n) ch 12 p13.1, (o) ch 12 q24.33, (p) ch 14 q24.1, (q) ch 14 q32.13, ( ) ch 15 q21.3, (s) ch 16 q23.1-q24.3, ( ) ch 17 p13.1, (u) ch 18 p11.32-p11.31, ( ) ch 18 q12.3, (w) ch 18 q23-q22.3, (x) ch 19 p12, (y) ch 19 q13.31, (z) ch 20 p13, (a2) ch 20 p12.1, (b2) ch 20 q13.33, (c2) ch 21 q22.2, (d2) ch 21 q22.3, (e2) ch 22 q12.1-q12.2, ( 2) ch 22 q13.31, (g2) Ch 6q14.3-q15, (h2) Ch 8p21.3-p21.2, (i2) Ch 10q23.31, (j2) Ch 13q14.13, (k 2) Ch 17q21.3, (l2) Ch 5q21.1, (m2) Ch 6p25.1-p24.3, (n2) Ch 6q27. (PDF) S4 Fig. Copy numbe al e a ion segmen s de ec ed by ASCAT ha o e lap wi h ampli ica- ion MRAs. The ch omosome gain e en s a e ep esen ed wi h dis inc colou blocks depending on he ype o SCNA: gain (any gain in he numbe o no mal allele copies) and ampli ica ion LOH (loss o one allele wi h any gain o he emaining allele). The egions o SCNA a e o de ed by leng h: op-smalles , bo om-la ges . Each block has been labelled wi h sample ID. Cases o which mo e han one sample was a ailable a e all indica ed, howe e con ibu ion o he e- quency o he SCNA was de ined on a pe pa ien basis. (a) ch 1 q21.3-q22, (b) ch 1 q25.3, (c) ch 1 q43-q44, (d) ch 2 q24.3, (e) ch 5 p15.31, ( ) ch 5 q33.3-q35.2, (g) ch 9 q33.1, (h) ch 10 q21.1-q21.3, (i) ch 11 q13.4-q13.5, (j) ch 11 q14.3, (k) ch 12 q23.1, (l) ch 13 q11-q12.11, (m) ch 13 q33.3, (n) ch 13 q33.3-q34, (o) ch 14 q13.3-q21.1, (p) ch 16 p13.3, (q) ch 16 p13.3, ( ) ch 16 p13.12-p13.11, (s) ch 17 q22-q23.1, ( ) Ch 8q11.1-q11.21, (u) Ch 7p11.2, ( ) Ch 8q21.1-q12.1, (w) Ch 8q24.21, (x) Ch 3q22.1-q21.3. (PDF) S5 Fig. Commonly al e ed pa hways in ETS posi i e and nega i e cance s. Blue and ed blocks indica e genes con ained in egions o dele ion and ampli ica ion espec i ely. G ey blocks indica e genes wi h no al e a ion ha we e equi ed o ep esen a ion o he pa hway. Pu ple and blue squa es indica e he pe cen age o samples wi h a copy numbe al e a ion in App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 20 / 28 ha gene in ETS nega i e and posi i e samples. (PDF) S6 Fig. A pla o m compa ison o ASCAT p o iles on a single sample. (a) one p o ile om SNP6.0 and (b) one om NGS da a. (PDF) S1 Table. Classi ica ion o soma ic copy numbe al e a ions. (DOCX) S2 Table. Copy numbe p o iling o 103 pa ien s. Including ploidy, deg ee o con amina ion and numbe o SCNAs. Clinico-pa hological cha ac e is ics o he s udy coho consis ing o 105 pa ien s. In o ma ion on pa ien s unde going adical p os a ec omy (88) and TURP (2) is displayed. (XLSX) S3 Table. ETS gene s a us. (XLSX) S4 Table. Summa y cha ac e is ics o he genomes and soma ic copy numbe al e a ions (SCNAs). Le side: Samples classi ied by disease s a us i.e. samples om p os a ec omies om pa ien s ee o me as a ic disease (PT) o samples om pa ien s wi h me as a ic disease (M). Righ side: P os a ec omy samples whe e he e was a leas six mon hs ollow up (n = 69) classi- ied by whe he he e was apid biochemical ecu ence wi hin six mon hs o p os a ec omy (PG) o no (RF) ( igh side). The wo TURP samples a e no summa ized in his able. Resul s om s a is ical es s a e shown ha es whe he he e a e signi ican di e ences be ween ei he me as a ic and p os a ec omy pa ien s o p os a ec omy pa ien s ha ha e biochemical ecu ence wi hin six mon hs o no . (DOCX) S5 Table. Summa y o clinico-pa hological cha ac e is ics o he pa ien s in he de ined se s o clus e s. (DOCX) S6 Table. Minimal egions o dele ions and ampli ica ions wi h linked genes. This able includes: Compa ison o MRAs, ex ended MRAs and GISTIC-de ec ed egions; ampli ica ions and dele ions ound in p e ious p os a e cance s udies.; pe cen age o he minimal egions o al e a ions ha a e conse ed egions; and clinical co ela ions and ETS associa ions o mini- mal egions o soma ic copy numbe al e a ion. (XLSX) S7 Table. Genes wi h mu a ions in ex ended MRA, GISTIC, and homozygous loss egions. Possible haploinsu iciency a ge s a e iden i ied as genes whe e he e a e a leas h ee mu a- ions and a no mal allele e ained. (XLSX) S8 Table. Mu a ions in lncRNAs and conse ed RNAs loca ed wi hin he minimal egions o soma ic copy numbe al e a ion. lncRNAs we e examined ha we e de ined by he MiT ansc ip ome p ojec as being ei he cance -associa ed o con aining conse ed egions. (XLSX) S9 Table. Mu a ions in p omo e egions o genes wi hin he minimal egions o al e a ion. (DOCX) App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 21 / 28 S10 Table. Mu a ions in DNA High-occupancy a ge (HOT) egions. (XLSX) S11 Table. GISTIC egions o dele ion and ampli ica ion. q- alues: The q- alue o he peak egion. Residual q- alues: The q- alue o he peak egion a e emo ing (“peeling o ”) ampli- ica ions o dele ions ha o e lap o he , mo e signi ican peak egions in he same ch omo- some. Wide Peak Limi s: The “wide peak” bounda ies mos likely o con ain he a ge ed genes. These a e lis ed in genomic coo dina es and ma ke (o p obe) indices. (XLSX) S12 Table. Lis o egions o homozygous loss ha occu in g ea e han wo pa ien s. (DOCX) S13 Table. Mu a ions in conse ed DNA sequences wi hin he minimal egions o soma ic copy numbe al e a ion. (XLSX) S14 Table. S uc u al ea angemen s in egions o dele ion in ol ing genes ound o be in o close o minimal egions o al e a ion. (XLSX) S15 Table. S uc u al ea angemen s in egions o dele ion in ol ing genes ound o be in o close o GISTIC de ec ed egions. (XLSX) S16 Table. Pa hway en ichmen analysis o ETS posi i e and nega i e cance s using mu a ed genes in signi ican egions o ampli ica ion and dele ion. The analysis was pe - o med using he Reac ome plugin o analysis o canonical pa hways in Cy oscape. Only sig- ni ican (FDR calcula ed using he Benjamini-Hochbe g p ocedu e p<0.05) en iched pa hways wi h po en ial in ol emen in cance a e lis ed. (XLSX) S17 Table. Summa y able o he GISTIC de ec ed dele ions ha ollow he Knudson hi model. (DOCX) S18 Table. Summa y o cases and copy numbe pla o ms. (DOCX) S1 Appendix. Example copy numbe plo s (BAF and logR) o ou umou samples (and associa ed con ols) o each o he 24 no el MCRs ha we ha e iden i ied. The black lines indica e he segmen de ec ed by ASCAT and he blue lines indica e he MCR egion. (PDF) S2 Appendix. An example s a is ical conside a ion o he Knudson 2-hi model. (PDF) S3 Appendix. Membe ship o he CRUK-ICGC p os a e g oup. (PDF) Acknowledgmen s We hank he Human Resea ch Tissue Bank a Camb idge Adenb ooke’s Hospi al, ha is sup- po ed by he NIHR Camb idge Biomedical Resea ch Cen e. We hank Da e Holland om App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 22 / 28 he In as uc u e Managemen Team, Pe e Clapham om he In o ma ics Sys ems G oup a he Wellcome T us Sange Ins i u e, and Bob Ge agh y om he Camb idge Ins i u e. The au ho s would like o hank hose men wi h p os a e cance and he subjec s who ha e dona ed hei ime and hei samples o his esea ch. We also would like o acknowledge he suppo o Jo Bu ge, Ma ie Co co an, Anne Geo ge, and Sa a S ea n o sample cu a ion. PVL is a Win- on G oup Leade in ecogni ion o he Win on Cha i able Founda ion’s suppo owa ds he es ablishmen o The F ancis C ick Ins i u e. Au ho Con ibu ions Concep ualiza ion: Niedzica Camacho, Da id E. Neal, Rosalind Eeles, Colin S. Coope , Dan- iel S. B ewe . Da a cu a ion: Niedzica Camacho, Lucy Ma hews, Ba ba a K emeye , Jo ge Zamo a, Gunes Gundem, Mohammed Gho i, Tapio Visako pi, Yong Jie Lu, Daniel S. B ewe . Fo mal analysis: Niedzica Camacho, Ke s in Haase, Ba ba a K emeye , Daniel S. B ewe . Funding acquisi ion: Ch is ophe S. Fos e , Ul an McDe mo , Da id E. Neal, Rosalind Eeles, Colin S. Coope . In es iga ion: Niedzica Camacho, Sand a Edwa ds, Jona han D. Kay, Lucy Ma hews, Je emy Cla k, Nening Dennis, Sa ah Thomas, Jo ge Zamo a, Sue Me son, Hayley Lux on, S e e Hawkins, Luke Ma sden, Adam Lambe , Ka alin Ka aszi, Gill Pel ende , Cha lie E. Massie, William J. Howa , S e en Hazell, Naomi Li ni, Cy il Fishe , Ch is ophe Ogden, Pa deep Kuma , Alan Thompson, Da id Nicol, E ik Maye , Tim Dudde idge, Yongwei Yu, Hon- gwei Zhang, Nimish C. Shah, Vincen J. Gnanap agasam, F eddie Hamdy, Dan Be ney, Cla e Ve ill, Anne Y. Wa en, And ew G. Lynch, Ch is ophe S. Fos e , Yong Jie Lu, G. S e en Bo a, Hayley C. Whi ake , Ul an McDe mo , Rosalind Eeles, Colin S. Coope , Dan- iel S. B ewe . Me hodology: Niedzica Camacho, Pe e Van Loo, Jona han D. Kay, Lucy Ma hews, Je emy Cla k, Daniel S. B ewe . P ojec adminis a ion: Sand a Edwa ds, Sa ah Thomas, Adam P. Bu le , Gunes Gundem, Mohammed Gho i, Cha lie E. Massie, Da id C. Wedge, And ew G. Lynch, Ch is ophe S. Fos e , Da id E. Neal, Rosalind Eeles, Colin S. Coope , Daniel S. B ewe . Resou ces: Nening Dennis, Sa ah Thomas, Sue Me son, Hayley Lux on, S e e Hawkins, Luke Ma sden, Adam Lambe , Ka alin Ka aszi, Gill Pel ende , Zso ia Ko e-Ja ai, Kei an Raine, Da id Jones, William J. Howa , S e en Hazell, Naomi Li ni, Cy il Fishe , Ch is ophe Ogden, Pa deep Kuma , Alan Thompson, Da id Nicol, E ik Maye , Tim Dudde idge, Yongwei Yu, Hongwei Zhang, Nimish C. Shah, Vincen J. Gnanap agasam, William Isaacs, Tapio Visako pi, F eddie Hamdy, Dan Be ney, Cla e Ve ill, Anne Y. Wa en, Da id C. Wedge, And ew G. Lynch, Ch is ophe S. Fos e , Yong Jie Lu, G. S e en Bo a, Hayley C. Whi ake , Ul an McDe mo , Da id E. Neal, Rosalind Eeles, Colin S. Coope , Daniel S. B ewe . So wa e: Niedzica Camacho, Pe e Van Loo, Jo ge Zamo a, Adam P. Bu le , Mohammed Gho i, Kei an Raine, Da id Jones, Daniel S. B ewe . Supe ision: Pe e Van Loo, Je emy Cla k, Adam P. Bu le , Zso ia Ko e-Ja ai, Vincen J. Gna- nap agasam, Da id C. Wedge, G. S e en Bo a, Hayley C. Whi ake , Ul an McDe mo , Da id E. Neal, Rosalind Eeles, Colin S. Coope , Daniel S. B ewe . App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 23 / 28 Valida ion: Niedzica Camacho, Ke s in Haase, Daniel S. B ewe . Visualiza ion: Niedzica Camacho, Daniel S. B ewe . W i ing – o iginal d a : Niedzica Camacho, Pe e Van Loo, Colin S. Coope , Daniel S. B ewe . W i ing – e iew & edi ing: Niedzica Camacho, Sand a Edwa ds, Ke s in Haase, Je emy Cla k, Cha lie E. Massie, Da id C. Wedge, And ew G. Lynch, Ch is ophe S. Fos e , Yong Jie Lu, G. S e en Bo a, Da id E. Neal, Rosalind Eeles, Colin S. Coope , Daniel S. B ewe . Re e ences 1. Be oukhim R, Me mel CH, Po e D, Wei G, Raychaudhu i S, Dono an J, e al. The landscape o soma ic copy-numbe al e a ion ac oss human cance s. Na u e. 2010; 463: 899–905. h ps://doi.o g/10. 1038/na u e08822 PMID: 20164920 2. Knudson AG. Mu a ion and cance : s a is ical s udy o e inoblas oma. P oc Na l Acad Sci U S A. 1971; 68: 820–3. PMID: 5279523 3. Magee JA, Abdulkadi SA, Milb and J. Haploinsu iciency a he Nkx3.1 locus. Cance Cell. 2003; 3: 273–283. h ps://doi.o g/10.1016/S1535-6108(03)00047-3 PMID: 12676585 4. Gao H, Ouyang X, Banach-Pe osky W, Bo owsky AD, Lin Y, Kim M, e al. A c i ical ole o p27kip1 gene dosage in a mouse model o p os a e ca cinogenesis. P oc Na l Acad Sci U S A. Na ional Acad- emy o Sciences; 2004; 101: 17204–9. h ps://doi.o g/10.1073/pnas.0407693101 PMID: 15569926 5. Solimini NL, Xu Q, Me mel CH, Liang AC, Schlabach MR, Luo J, e al. Recu en hemizygous dele ions in cance s may op imize p oli e a i e po en ial. Science. 2012; 337: 104–9. h ps://doi.o g/10.1126/ science.1219580 PMID: 22628553 6. De S, Micho F. DNA eplica ion iming and long- ange DNA in e ac ions p edic mu a ional landscapes o cance genomes. Na Bio echnol. 2011; 29: 1103–1108. h ps://doi.o g/10.1038/nb .2030 PMID: 22101487 7. Fudenbe g G, Ge z G, Meye son M, Mi ny LA. High o de ch oma in a chi ec u e shapes he landscape o ch omosomal al e a ions in cance . Na Bio echnol. 2011; 29: 1109–13. h ps://doi.o g/10.1038/nb . 2049 PMID: 22101486 8. San a ius T, Shipley J, B ewe D, S a on MR, Coope CS. A census o ampli ied and o e exp essed human cance genes. Na Re Cance . 2010; 10: 59–64. h ps://doi.o g/10.1038/n c2771 PMID: 20029424 9. Fe lay J, Soe joma a am I, Dikshi R, Ese S, Ma he s C, Rebelo M, e al. Cance incidence and mo al- i y wo ldwide: sou ces, me hods and majo pa e ns in GLOBOCAN 2012. In J cance . 2015; 136: E359–86. h ps://doi.o g/10.1002/ijc.29210 PMID: 25220842 10. Holcomb IN, G o e DI, Kinnunen M, F iedman CL, Gallahe IS, Mo gan TM, e al. Genomic al e a ions indica e umo o igin and a ied me as a ic po en ial o dissemina ed cells om p os a e cance pa ien s. Cance Res. 2008; 68: 5599–608. h ps://doi.o g/10.1158/0008-5472.CAN-08-0812 PMID: 18632612 11. Liu W, Lai inen S, Khan S, Vihinen M, Kowalski J, Yu G, e al. Copy numbe analysis indica es monoclo- nal o igin o le hal me as a ic p os a e cance . Na Med. 2009; 15: 559–65. h ps://doi.o g/10.1038/nm. 1944 PMID: 19363497 12. Robbins CM, Tembe WA, Bake A, Sina i S, Moses TY, Becks om-S e nbe g S, e al. Copy numbe and a ge ed mu a ional analysis e eals no el soma ic e en s in me as a ic p os a e umo s. Genome Res. 2011; 21: 47–55. h ps://doi.o g/10.1101/g .107961.110 PMID: 21147910 13. Lalonde E, Ishkanian AS, Sykes J, F ase M, Ross-Adams H, E ho N, e al. Tumou genomic and mic oen- i onmen al he e ogenei y o in eg a ed p edic ion o 5-yea biochemical ecu ence o p os a e cance : a e ospec i e coho s udy. Lance Oncol. Lalonde e al. Open Access a icle dis ibu ed unde he e ms o CC BY; 2014; 15: 1521–1532. h ps://doi.o g/10.1016/S1470-2045(14)71021-6 PMID: 25456371 14. Van Loo P, No dga d SH, Lingjæ de OC, Russnes HG, Rye IH, Sun W, e al. Allele-speci ic copy num- be analysis o umo s. P oc Na l Acad Sci U S A. 2010; 107: 16910–5. h ps://doi.o g/10.1073/pnas. 1009843107 PMID: 20837533 15. Van Loo P, Nilsen G, No dga d SH, Vollan HKM, Bø esen-Dale A-L, K is ensen VN, e al. Analyzing cance samples wi h SNP a ays. Me hods Mol Biol. 2012; 802: 57–72. h ps://doi.o g/10.1007/978-1- 61779-400-1_4 PMID: 22130873 16. Coope CS, Eeles R, Wedge DC, Van Loo P, Gundem G, Alexand o LB, e al. Analysis o he gene ic phylogeny o mul i ocal p os a e cance iden i ies mul iple independen clonal expansions in neoplas ic App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 24 / 28 and mo phologically no mal p os a e issue. Na Gene . Na u e Publishing G oup, a di ision o Macmil- lan Publishe s Limi ed. All Righ s Rese ed.; 2015; 47: 367–372. h ps://doi.o g/10.1038/ng.3221 PMID: 25730763 17. Gundem G, Van Loo P, K emeye B, Alexand o LB, Tubio JMC, Papaemmanuil E, e al. The e olu ion- a y his o y o le hal me as a ic p os a e cance . Na u e. 2015; 520: 353–357. h ps://doi.o g/10.1038/ na u e14347 PMID: 25830880 18. Taylo BS, Schul z N, Hie onymus H, Gopalan A, Xiao Y, Ca e BS, e al. In eg a i e genomic p o iling o human p os a e cance . Cance Cell. 2010; 18: 11–22. h ps://doi.o g/10.1016/j.cc .2010.05.026 PMID: 20579941 19. Hie onymus H, Schul z N, Gopalan A, Ca e BS, Chang MT, Xiao Y, e al. Copy numbe al e a ion bu - den p edic s p os a e cance elapse. P oc Na l Acad Sci U S A. 2014; 111: 11139–44. h ps://doi.o g/ 10.1073/pnas.1411446111 PMID: 25024180 20. Iye MK, Nikna s YS, Malik R, Singhal U, Sahu A, Hosono Y, e al. The landscape o long noncoding RNAs in he human ansc ip ome. Na Gene . 2015; 47: 199–208. h ps://doi.o g/10.1038/ng.3192 PMID: 25599403 21. Boyle AP, A aya CL, B dlik C, Cay ing P, Cheng C, Cheng Y, e al. Compa a i e analysis o egula o y in o ma ion and ci cui s ac oss dis an species. Na u e. 2014; 512: 453–6. h ps://doi.o g/10.1038/ na u e13668 PMID: 25164757 22. Me mel CH, Schumache SE, Hill B, Meye son ML, Be oukhim R, Ge z G. GISTIC2.0 acili a es sensi- i e and con iden localiza ion o he a ge s o ocal soma ic copy-numbe al e a ion in human cance s. Genome Biol. 2011; 12: R41. h ps://doi.o g/10.1186/gb-2011-12-4- 41 PMID: 21527027 23. Cance Genome A las Resea ch Ne wo k. The Molecula Taxonomy o P ima y P os a e Cance . Cell. 2015; 163: 1011–25. h ps://doi.o g/10.1016/j.cell.2015.10.025 PMID: 26544944 24. F ase M, Sabelnyko a VY, Yamaguchi TN, Heisle LE, Li ings one J, Huang V, e al. Genomic hall- ma ks o localized, non-indolen p os a e cance . Na u e. Na u e Publishing G oup; 2017; h ps://doi. o g/10.1038/na u e20788 PMID: 28068672 25. Lapoin e J, Li C, Giacomini CP, Sala i K, Huang S, Wang P, e al. Genomic p o iling e eals al e na i e gene ic pa hways o p os a e umo igenesis. Cance Res. 2007; 67: 8504–10. h ps://doi.o g/10.1158/ 0008-5472.CAN-07-0673 PMID: 17875689 26. Liu W, Xie CC, Thomas CY, Kim S-T, Lindbe g J, Ege ad L, e al. Gene ic ma ke s associa ed wi h ea ly cance -speci ic mo ali y ollowing p os a ec omy. Cance . 2013; 119: 2405–12. h ps://doi.o g/10. 1002/cnc .27954 PMID: 23609948 27. Reid AHM, A a d G, B ewe D, Mi anda S, Riisnaes R, Cla k J, e al. No el, g oss ch omosomal al e - a ions in ol ing PTEN coope a e wi h allelic loss in p os a e cance . Mod Pa hol. 2012; 1: 1–9. h ps:// doi.o g/10.1038/modpa hol.2011.207 28. Ve hagen PCMS, an Duijn PW, He mans KGL, Looijenga LHJ, an Gu p RJHLM, S oop H, e al. The PTEN gene in locally p og essi e p os a e cance is p e e en ially inac i a ed by bi-allelic gene dele ion. J Pa hol. 2006; 208: 699–707. h ps://doi.o g/10.1002/pa h.1929 PMID: 16402365 29. Baca SC, P andi D, Law ence MS, Mosque a JM, Romanel A, D ie Y, e al. Punc ua ed e olu ion o p os a e cance genomes. Cell. Else ie Inc.; 2013; 153: 666–77. h ps://doi.o g/10.1016/j.cell.2013.03. 021 PMID: 23622249 30. Ba os-Sil a JD, Paulo P, Bakken AC, Ce ei a N, Lø M, Hen ique R, e al. No el 50Fusion Pa ne s o ETV1 and ETV4 in P os a e Cance . Neoplasia. 2013; 15: 720–IN6. h ps://doi.o g/10.1593/neo. 13232 PMID: 23814484 31. Be lin A, Lalonde E, Sykes J, Za a ana G, Chu KC, Ramna ine VR, e al. NBN gain is p edic i e o ad e se ou come ollowing image-guided adio he apy o localized p os a e cance . Onco a ge . 2014; 5: 11081–90. h ps://doi.o g/10.18632/onco a ge .2404 PMID: 25415046 32. Moka am P, Kuma K, B im H, Naghibalhossaini F, Sabe i- i oozi M, Nou aie M, e al. Dis inc high-p o- ile me hyla ed genes in colo ec al cance . PLoS One. 2009; 4: e7012. h ps://doi.o g/10.1371/jou nal. pone.0007012 PMID: 19750230 33. Tessema M, Belinsky SA. Mining he epigenome o me hyla ed genes in lung cance . P oc Am Tho ac Soc. 2008; 5: 806–10. h ps://doi.o g/10.1513/pa s.200805-045TH PMID: 19017733 34. Kusakabe M, Ku omi T, Wa anabe K, Emo o N, Aki N, Kage H, e al. Iden i ica ion o G0S2 as a gene equen ly me hyla ed in squamous lung cance by combina ion o in silico and expe imen al app oaches. In J cance . Wiley Online Lib a y; 2010; 126: 1895–902. h ps://doi.o g/10.1002/ijc.24947 PMID: 19816938 35. Hoque MO, Kim MS, Os ow KL, Liu J, Wisman GBA, Pa k HL, e al. Genome-wide p omo e analysis unco e s po ions o he cance me hylome. Cance Res. 2008; 68: 2661–70. h ps://doi.o g/10.1158/ 0008-5472.CAN-07-5913 PMID: 18413733 App aising he ele ance o copy numbe loss and gain in p os a e PLOS Gene ics | h ps://doi.o g/10.1371/jou nal.pgen.1007001 Sep embe 25, 2017 25 / 28