Applications of genotyping by sequencing in aquaculture breeding and genetics

Applica ions o geno yping by sequencing in aquacul u e b eeding and gene ics Diego Robledo 1 , Ch is os Palaiokos as 1 , Luca Ba gelloni 2 , Paulino Ma  ınez 3 and Ross Hous on 1 1 The Roslin Ins i u e and Royal (Dick) School o Ve e ina y S udies, Uni e si y o Edinbu gh, Midlo hian, UK 2 Depa men o Compa a i e Biomedicine and Food Science, Uni e si y o Pado a, Legna o, Pado a, I aly 3 Depa men o Zoology, Gene ics and Physical An h opology, Facul y o Ve e ina y, Uni e si y o San iago de Compos ela, Lugo, Spain Co espondence Ross Hous on, The Roslin Ins i u e and Royal (Dick) School o Ve e ina y S udies, The Uni e si y o Edinbu gh, Midlo hian, EH25 9RG UK. Email: [email p o ec ed] Recei ed 12 Oc obe 2016; accep ed 27 Decembe 2016. Abs ac Selec i e b eeding is inc easingly ecognized as a key componen o sus ainable p oduc ion o aquacul u e species. The up ake o genomic echnology in aquacul- u e b eeding has adi ionally lagged behind e es ial a med animals. Howe e , he apid de elopmen and applica ion o sequencing echnologies has allowed aquacul u e o na ow he gap, leading o subs an ial genomic esou ces o all majo aquacul u e species. While high-densi y single-nucleo ide polymo phism (SNP) a ays o some species ha e been de eloped ecen ly, di ec geno yping by sequencing (GBS) echniques ha e unde pinned many o he ad ances in aqua- cul u e gene ics and b eeding o da e. In pa icula , es ic ion-si e associa ed DNA sequencing (RAD-Seq) and subsequen a ia ions ha e been ex ensi ely applied o gene a e popula ion-le el SNP geno ype da a. These GBS echniques a e no dependen on p io genomic in o ma ion such as a e e ence genome assembly o he species o in e es . As such, hey ha e been widely u ilized by esea che s and companies ocussing on nonmodel aquacul u e species wi h ela- i ely small esea ch communi ies. Applica ions o RAD-Seq echniques ha e included gene a ion o gene ic linkage maps, pe o ming genome-wide associa- ion s udies, imp o emen s o e e ence genome assemblies and, mo e ecen ly, genomic selec ion o ai s o in e es o aquacul u e like g ow h, sex de e mina- ion o disease esis ance. In his e iew, we b ie ly discuss he his o y o GBS, he nuances o he a ious GBS echniques, bioin o ma ics app oaches and applica- ion o hese echniques o a ious aquacul u e species. Key wo ds: aquacul u e, geno yping, nex -gene a ion sequencing, es ic ion-si e associa ed DNA, selec i e b eeding, single nucleo ide polymo phism. Backg ound Despi e he c i ical ole o aquacul u e in global ood secu- i y, he as majo i y o wo ld ish and shell ish p oduc ion is based on s ocks wi hou ad anced selec i e b eeding p o- g ammes (Gjed em e al. 2012; Janssen e al. 2016). Aqua- cul u e b eeding schemes end o lag behind hei e es ial li es ock coun e pa s in e ms o he up ake o genomic echnologies, and o many aquacul u e species, molecula gene ic ools a e only applied o pedig ee econs uc ion (Cha anne e al. 2016). In compa ison, mos mode n b eeding p og ammes in li es ock a e now unde pinned by genomic selec ion (GS, Meuwissen e al. 2001), he bene i s o which a e well-illus a ed in dai y ca le (Hayes e al. 2009). GS ypically equi es genome-wide gene ic ma ke da a o a la ge numbe o indi idual animals. Up un il a ew yea s ago, ob aining gene ic ma ke s was cos ly and labo ious; hence, la ge numbe s o ma ke s we e only a ail- able o a hand ul o well-s udied species. Howe e , he ecen ad ances in nex -gene a ion sequencing (NGS) ha e g ea ly educed he cos o nucleic acid sequencing, and he e o e also gene ic ma ke disco e y. This has opened he doo o apid gene a ion o genome-wide gene ic ma - ke da ase s, ei he ia gene a ion and applica ion o SNP a ays, o di ec ly ia geno yping by sequencing (GBS) echniques (Da ey e al. 2011). GBS echniques ha e e o- lu ionized he ield o e olu iona y genomics ( e iewed in And ews e al. 2016) and ha e also led o se e al ad ances in gene ics and b eeding o aquacul u e species, he subjec o his e iew. ©2017 The Au ho s. Re iews in Aquacul u e Published by Wiley Publishing Asia P y L d670 This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s use, dis ibu ion and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed. Re iews in Aquacul u e (2018) 10, 670–682 doi: 10.1111/ aq.12193 Due o he high ecundi y o aquacul u e species, he majo i y o b eeding p og ammes a e based on collec ion o ai da a on close ela i es (e.g. ull siblings) o he selec- ion candida es, pa icula ly whe e he ai o in e es can- no be measu ed on he candida es hemsel es (e.g. ille quali y, disease esis ance). Wi hou gene ic ma ke s, his se -up enables amily selec ion, whe eby amily-le el es i- ma ed b eeding alues (EBVs) o selec ion candida es a e calcula ed using he da a collec ed on he ela i es. How- e e , o u ilize he wi hin- amily gene ic a ia ion in hese ai s, gene ic ma ke s a e necessa y o dis inguish be ween selec ion candida es. Implemen a ion o ma ke s in b eed- ing can b oadly be spli in o wo ca ego ies; ma ke -assis ed selec ion (MAS) and GS. MAS is based on he use o a - ge ed ma ke s linked o majo quan i a i e ai s loci (QTL) a ec ing he ai , and one o he i s examples in aquacul u e was hos esis ance o in ec ious panc ea ic nec osis i us (IPNV) in A lan ic salmon (Salmo sala , Hous on e al. 2008; Moen e al. 2009). Fo ai s wi h a polygenic a chi ec u e, GS is a mo e app op ia e app oach, whe eby he ela i es o he selec ion candida es become he ‘ aining’ popula ion wi h geno ypes and pheno ypes, and hose da a a e used o calcula e genomic b eeding al- ues (GEBVs) o selec ion candida es wi h geno ype da a only. This applica ion o genomic selec ion in aquacul u e b eeding is a a o ma i e s age, and mos examples o da e ha e ocussed on imp o ed b eeding o esis ance o in ec- ious diseases (e.g. Ødeg  a d e al. 2014; Tsai e al. 2015, 2016b; Vallejo e al. 2016; Dou e al. 2016; Palaiokos as e al. 2016). The majo i y o high- esolu ion gene ic s udies in aquacul u e species, and applica ions o genomic selec- ion, ha e been unde pinned by GBS echniques, ei he by di ec ly p o iding geno ype da a o by disco e ing ma ke s o he design o SNP a ays, which a e cu en ly only a ail- able o a hand ul o aquacul u e species (e.g. A lan ic sal- mon, Hous on e al. 2014; Y a~ nez e al. 2016; Paci ic oys e , C assos ea gigas, and Eu opean la oys e , Os ea edulis, Lap egue e al. 2014; channel ca ish, Ic alu us punc a us, Liu e al. 2014; common ca p, Cyp inus ca pio,Xue al. 2014a; ainbow ou , Onco hynchus mykiss, Pal i e al. 2015a). The mos common GBS echniques in ol e lib a y p epa a ion s eps ha esul in deep sequence da a a a epea able subse o si es dispe sed h oughou he genome, ypically using one o wo es ic ion enzymes (RE), al hough also new GBS echniques based on a ge ed sequencing ha e been ecen ly de eloped (i.e. GT-Seq, dis- cussed below). The eason behind his genome complexi y educ ion is ha high-co e age sequencing o a ypical aquacul u e species’ genome wi h enough dep h o con i- den ly call geno ypes is s ill p ohibi i ely expensi e o he numbe o animals equi ed o high- esolu ion gene ic s udies and b eeding p og amme applica ions. Genome complexi y educ ion ia RE is as and inexpensi e. Indeed, RE-based echniques ha e been commonplace in geno yping o many yea s, wi h RFLP and AFLP being widely applied o gene a e geno yping assays o limi ed numbe s o gene ic ma ke s. The ma iage o hese ideas wi h NGS has enabled a majo b eak h ough o gene ic s udies o complex ai s in nonmodel o ganisms, and hei applica ion o imp o e aquacul u e p oduc ion. RAD sequencing Res ic ion-si e associa ed DNA sequencing (RAD sequencing o RAD-Seq) co e s a ange o GBS echniques which com- bine he use o genome complexi y educ ion wi h REs and he high sequencing ou pu o NGS echnologies. RAD-Seq was i s desc ibed by Bai d e al. (2008), ollowing on om a simila idea based on mic oa ays (Mille e al. 2007). Some o he main easons o i s ins an success a e ha RAD-Seq does no equi e any p io genomic knowledge, i allows gen- e a ion o popula ion-speci ic geno ype da a (i.e. no asce - ainmen bias) and i o e s lexibili y in e ms o desi ed ma ke densi y ac oss he genome. The use o di e en REs o inno a i emodi ica ions o hebase echniqueallowsa high le el o con ol o e he numbe o ma ke s ob ained o a speci ic s udy. RAD-Seq and simila echniques a e also amenable ools o aquacul u e b eeding, whe e gene ic ma ke s ha e ypically been used in amily assignmen and pedig ee econs uc ion (Vandepu e & Ha ay 2014). Mass spawning species a e common in aquacul u e, whe e mixed ea ing and unknown pa en al con ibu ion necessi a e he use o geno yping o amily-based b eeding. RAD-Seq po en- ially acili a es a single expe imen whe eby pedig ees a e econs uc ed, gene ic di e si y is quan i ied, QTL can be mapped and genomic b eeding alues calcula ed (Palaiokos- as e al. 2016). Since he o iginal RAD-Seq pape by Bai d e al. (2008), se e al a ian s o his me hodology ha e been desc ibed. Th ee o hem ha e been ex ensi ely used in aqua- cul u e gene ics esea ch: he o iginal RAD-Seq (Bai d e al. 2008), 2b-RAD (Wang e al. 2012) and ddRAD (Pe e son e al. 2012). O he RAD-based echniques like ezRAD (Too- nen e al. 2013) o SLAF-seq (Sun e al. 2013) in oduced mino modi ica ions, which do no con e a majo ad an age o aquacul u e applica ions. All a ailable RAD-based ech- niques ha e been ecen ly e iewed in dep h elsewhe e (And ews e al. 2016); he e o e, he e we ha e ocused on hose mos ele an in aquacul u e b eeding. The main ea- u es o o iginal RAD-Seq, 2b-RAD and ddRAD a e shown in Table 1, and hey a e b ie ly desc ibed below. O iginal RAD-Seq In o iginal RAD-Seq (Bai d e al. 2008), genomic DNA samples om se e al animals a e indi idually diges ed wi h Re iews in Aquacul u e (2018) 10, 670–682 ©2017 The Au ho s. Re iews in Aquacul u e Published by Wiley Publishing Asia P y L d 671 RAD sequencing in aquacul u e a RE o choice. The diges ed DNA is hen andomly shea ed and pooled a e liga ion o adap o s wi h nucleo ide ba - codes o unique iden i ica ion o each sample. The esul ing es ic ion agmen s a e selec ed o sui able size ange (i.e. o Illumina sequencing, ypically 300–600 bp), and a e a subsequen polyme ase chain eac ion (PCR) s ep, he ag- men s a e sequenced. The esul is high-co e age sequence da a o lanking egions o he RE cu si es, which a e ypi- cally dispe sed qui e e enly h oughou he genome. As such, a genome-wide gene ic ma ke da ase can be p oduced ac oss a popula ion o indi iduals a a ac ion o he cos o whole genome esequencing. Illumina sequencing o sho agmen s ei he in ol es sequencing one (one ead, single end) o bo h ( wo eads, pai ed end) ends o each agmen and cu en ly gi es eads o up o 300 bp in leng h. Each lanking sequence o he RE cu si e is e e ed o as a RAD locus (o RAD- ag), and he high co e age o RAD ags acil- i a es simul aneous SNP de ec ion and geno yping. The numbe o RAD ags, and he e o e SNPs, gene a ed in he expe imen is uneable ia he choice o a e o mo e e- quen cu ing RE. The mos commonly used enzyme o da e is Sb I which has an eigh base ecogni ion si e and he e o e cu s ela i ely in equen ly h oughou he genome. Online ools a e a ailable o guide he choice o he mos app op i- a e RE acco ding o he equi emen s and budge o he s udy (Lepais & Wei 2014). In addi ion o sequencing and geno yping indi iduals, he app oach is also amenable o geno yping pooled popula ions o bulk-seg egan analysis (Bai d e al. 2008; Hohenlohe e al. 2010). One o he main d awbacks o he o iginal echnique is ha shea ing by soni- ca ion is andom and a iable, po en ially hinde ing he e i- ciency and he ep oducibili y o RAD-Seq (Da ey e al. 2013). Howe e , his andom shea ing s ep can also be a bene i , as he a iable size o he genomic agmen s ancho ed a he RE cu si e acili a es he assembly o a con- ig based on he pai ed-end eads. This augmen s anno a ion o he RAD loci when he e is no e e ence genome a ailable, and also he design o speci ic p ime s o e-geno yping o a ge ed SNPs. In addi ion, he pai ed-end da a om RAD-Seq allow iden i ica ion and emo al o pu a i e PCR duplica es ( eads o igina ed om he same o iginal DNA agmen , he e o e p esen ing iden ical sequences), which can hinde analysis and in e p e a ion o Illumina sequenc- ing da a (Schweyen e al. 2014). While he e a e se e al sou ces o po en ial bias and e o in RAD-Seq echniques (see e iew by And ews e al. 2016), se e al heo e ical and empi ical s udies ha e demons a ed ha RAD-Seq does en- de ep oducible geno yping da a ac oss di e en labo a o- ies, popula ions and e en species (e.g. DaCos a & So enson 2014; Gonen e al. 2015). 2b-RAD The i s majo modi ica ion o he o iginal RAD echnique was e med 2b-RAD (Wang e al. 2012). The main inno a- ion in 2b-RAD is he use o ype IIB REs, which sha e he ea u e o cu ing he genomic DNA a bo h sides o he ecogni ion si e a a ixed dis ance, esul ing in p o uding noncohesi e ends. The esul is sho genomic DNA ag- men s o iden ical size a each IIB RE si e in he genome. Lib a y cons uc ion in he 2b-RAD p o ocol is simple. Fol- lowing DNA diges ion, adap o s a e liga ed o he agmen s, and speci ic ba codes a e added o each sample h ough PCR ampli ica ion using degene a ed linke s. Samples a e hen pooled and sequenced ypically using Illumina echnol- ogy, bu allowing o uns o sho e ead leng h due o he smalle size o he agmen s in compa ison o o iginal RAD (2b-RAD agmen s a e 33–36 bp). The use o ype IIB REs heo e ically acili a es he sampling and sequencing o iden- ical si es ac oss indi iduals, ci cum en ing he po en ial bias o RAD-Seq caused by he andom shea ing s ep. I also a oids he ime-consuming and po en ially e o -p one size- selec ion s ep, which cha ac e izes he majo i y o o he RAD me hods. Addi ionally, 2b-RAD is cu en ly he only membe o he RAD amily ha allows emo al o loci exhibi ing s and bias (Pu i z e al. 2014a). The possibili y o p oduce indi idually ba coded lib a ies allows a ge ed adjus men be o e pooling o ob ain mo e equal ep esen a- ion o indi idual samples. The main ca ea o his me hod is ha i p oduces sho sequencing eads (33–36 bp), which a e less amenable o alignmen o e e ence genome assem- blies, and hinde s ollow-up applica ions such as he design o indi idual SNP assays (due o lack o SNP lanking sequence). Howe e , his is no an issue i a d a genome sequence is a ailable o he species, as is becoming he case in many aquacul u e ish species. Table 1 Summa y o he di e en geno yping by sequencing (GBS) echniques Technique Key ea u es Ad an ages Disad an ages RAD-Seq Diges ion wi h one RE •Pai ed-end con igs •PCR duplica e emo al •Complex lib a y p epa a ion 2bRAD Diges ion wi h ype IIB REs •No size-selec ion s ep •High ep oducibili y •Easy lib a y p epa a ion •S and bias de ec ion •Sho agmen s •Remo al o PCR duplica es no possible ddRAD Diges ion wi h wo di e en REs •Can mul iplex many samples •Easy lib a y p epa a ion •Flexibili y o e SNP densi y •Repea abili y dependen on size- selec ion s ep Re iews in Aquacul u e (2018) 10, 670–682 ©2017 The Au ho s. Re iews in Aquacul u e Published by Wiley Publishing Asia P y L d 672 D. Robledo e al. ddRAD Pe e son e al. (2012) de eloped a new RAD-Seq pla o m using a double diges ion o genomic DNA wi h wo REs (ddRAD), hus elimina ing he shea ing s ep o o iginal RAD. The ddRAD p o ocol is mo e lexible han RAD-Seq o 2b-RAD in e ms o a ge ed ma ke densi y; he numbe o agmen s and SNPs can be eadily ailo ed by combin- ing di e en RE pai s. Due o he ypical use o a a e and a common cu ing enzyme, ddRAD esul s in ewe sequenced si es han RAD-Seq, acili a ing highe sequence co e age and/o mo e indi iduals mul iplexed wi hin a sin- gle sequencing lane. Highe mul iplexing is possible due o combina ional mul iplex indexing, whe eby a i s ba code is in oduced in he liga ion s ep and a second du ing he PCR. The e o e, a la ge numbe o samples can po en ially be sequenced in a single lane han wi h he o he RAD ech- niques. Compa ed o he RAD-Seq p o ocol, he wo k low o p epa a ion o ddRAD lib a ies is simple , quicke and also subs an ially cheape . Howe e , he wo k low is s ill mo e complex han he 2b-RAD p o ocol and equi es a size-selec ion s ep. To ensu e epea abili y o sampled ddRAD loci ac oss samples and lib a ies, consis ency o size selec ion is pa amoun (And ews e al. 2016). A simpli ied a ia ion o he ini ial ddRAD p o ocol, whe e bo h P1 and P2 adap o s wi h indi idual ba codes a e liga ed p io o size selec ion (Palaiokos as e al. 2015a), u he educes hands-on ime o lib a y p epa a ion. RAD bioin o ma ic analyses The ad en o NGS posed impo an challenges in e ms o da a s o age, ans e and analysis, which necessi a ed he de elopmen o specialized ha dwa e and so wa e. Conse- quen ly, he imp o emen o NGS-based sequencing pla - o ms occu ed in andem wi h con inuous de elopmen and imp o emen o sui able bioin o ma ics ools o anal- yse he la ge da ase s. A weal h o so wa e is a ailable o analysing da a o igina ing om he RAD amily o ech- niques. In he cu en e iew, a gene al amewo k o da a analysis will be desc ibed, a he han a emp ing o p o- ide a comp ehensi e o e iew o all a ailable ools. Acco dingly, he mos popula , s aigh o wa d o use and egula ly upda ed o he a ailable ools a e highligh ed in e ms o a sugges ed o de o usage ha migh o m a com- ple e RAD analysis pipeline. Expe imen al design and simula ion Sequencing and lib a y cons uc ion ypically accoun o he bulk o he cos o any expe imen u ilizing NGS. This leads o a balancing exe cise, whe eby esea che s s i e o include as many samples as possible pe sequencing lane (mul iplexing), wi hou comp omising he ead co e age equi ed o accu a e SNP geno ype calling. The e o e, wo key a iables o a RAD expe imen a e he choice o he RE (a ec ing how many si es a e sequenced), and he desi ed ead co e age pe locus. In silico simula ion is a aluable ool o any well-designed RAD expe imen . The R-based package SimRAD (Lepais & Wei 2014) can be u i- lized o simula ion-based p edic ion o he expec ed num- be o loci o each RE (o hei combina ion) and he genome o s udy. Al hough simula ion es ima es a e likely o di e om he empi ical da a, aluable in o ma ion can be gained o op imize expe imen al design be o e commi - ing o he high cos associa ed wi h lib a y cons uc ion and sequencing. Demul iplexing lib a ies The iles ha a e gene a ed by he sequence ( ypically Fas Q iles) equi e demul iplexing in o indi idual samples based on nucleo ide ba codes. The mos popula packages o his ask include S acks (Ca chen e al. 2011) and pyRAD (Ea on 2014). S anda d quali y con ol p ocedu e is o disca d sequence eads below use -de ined accep able quali y sco es, e oneous ba codes and eads missing he cha ac e is ic sequence pa e n ob ained om he RE. Fol- lowing demul iplexing, sequence iles co esponding o each indi idual a e gene a ed o downs eam analyses, including SNP calling and geno yping. SNP iden i ica ion One o he key ad an ages o RAD-Seq app oaches o non- model o ganisms (including many aquacul u e species) is he abili y o iden i y and geno ype SNPs wi hou equi ing a e e ence genome o he o ganism unde s udy. This app oach, commonly de ined in he li e a u e as de no o assembly, can be pe o med using ei he S acks (Ca chen e al. 2011), pyRAD (Ea on 2014) o dDocen (Pu i z e al. 2014b); howe e , he la e is limi ed o ddRAD o ezRAD da a. The de no o app oach in ol es iden i ica ion and assembling o RAD loci in each indi idual, based on use - de ined pa ame e s ela ed o ead co e age equi ed pe locus, and sequence di e gence be ween loci (Ca chen e al. 2011). Iden i ica ion o SNPs and in e ence o alleles wi hin RAD loci is pe o med using a maximum-likelihood-based algo i hm (Hohenlohe e al. 2010), which unde akes s a- is ical es s a each nucleo ide posi ion o assess he likeli- hood o a pa icula diploid geno ype. In doing so, he model implici ly es ima es and accoun s o sequencing e o a e (Ca chen e al. 2011). The S acks so wa e does no cu en ly suppo SNP iden i ica ion and geno yping in he pai ed-end (P2) ead, unless ancho ed o a second RE (e.g. in ddRAD). The e o e, in o iginal RAD expe imen s Re iews in Aquacul u e (2018) 10, 670–682 ©2017 The Au ho s. Re iews in Aquacul u e Published by Wiley Publishing Asia P y L d 673 RAD sequencing in aquacul u e using S acks, he P2 ead is ypically used o quali y con- ol (e.g. emo al o PCR duplica es), and o cons uc ing pai ed-end ‘mini-con igs’ which acili a e BLAST alignmen and geno yping assay design (E e e al. 2011). The simul- aneous use o P1 and P2 eads in he case o dDocen , and he applica ion o an alignmen -clus e ing algo i hm in he case o pyRAD, allow he iden i ica ion o inse ion/dele ion polymo phisms (indels) and iden i ica ion o SNPs in he P2 eads. Due o he dec easing cos o NGS, e e ence genome sequences a e becoming a ailable o many impo an aquacul u e species. The numbe o species wi h e e ence genome assemblies is apidly inc easing (A lan ic cod, Gadus mo hua, S a e al. 2011; Paci ic oys e , Zhang e al. 2012; Eu opean sea bass, Dicen a chus lab ax, Tine e al. 2014; ainbow ou , Be helo e al. 2014; Japanese eel, Anguilla japonica, Kai e al. 2014; hal -smoo h ongue sole, Cynoglossus semilae is, Chen e al. 2014; common ca p, Xu e al. 2014b; No he n pike, Esox lucius, Rondeau e al. 2014; Nile ilapia, O eoch omis nilo icus, B awand e al. 2015; Asian sea bass, La es calca i e , Vij e al. 2016; Medi e anean mussel, My ilus gallop o incialis, Mu ga ella e al. 2016; u bo , Scoph halmus maximus, Figue as e al. 2016; A lan ic salmon, Lien e al. 2016; channel ca ish, Chen e al. 2016), and new sequencing da a will imp o e genome quali y and anno a ion. The e o e, e e ence- guided RAD-Seq app oaches a e likely o be inc easingly u ilized. Bo h S acks and dDocen can u ilize e e ence gen- ome in o ma ion, using s anda d alignmen ools ollowed by simila SNP calling algo i hms o he de no o app oach desc ibed abo e. Po en ial bias and sou ces o e o While he bioin o ma ic pipelines o he RAD-like app oaches a e becoming inc easingly s anda dized, he e emains po en ial in insic ba ie s ha mus be o e - come o ensu e he gene a ion o accu a e and epea able SNP da ase s. One example ha is pa icula ly ele an o he aquacul u e esea ch communi y is dis inguishing be ween genuine allelic SNPs and pa alogous a ian s esul ing om ances al whole genome duplica ion. This is pa icula ly a challenge o salmonid species, and s a egies o accoun o his include (i) assessing ead co e age o pa e ns sugges i e o pa alogous a ia ion, (ii) checking o excessi e he e ozygosi y a loci and (iii) sequencing (double) haploid indi iduals as he basis o il e ing ou pa alogous sequence a ian s (e.g. E e e & Seeb 2014; Hous on e al. 2014; Pal i e al. 2015a,b). Ano he po en ial sou ce o e o o all RAD-Seq s udies is he p oblem o RAD allele d opou (Gau ie e al. 2013), whe e mu a ions wi hin he ecogni ion sequence o he RE seg ega ing in he popula ion a e a common sou ce o null alleles. The ex en o he issue is ela ed o he leng h o he RE ecogni ion sequence, and i is he e o e po en ially mo e o a p oblem o ddRAD (which equi es wo REs) e sus o he me hods (Gonen e al. 2015; And ews e al. 2016). Bo h ead co e age le els and assessmen o seg ega ion dis o ion in pedi- g eed c osses can assis in iden i ying and emo ing, o accoun ing o , hese null alleles. Finally, he concep o PCR duplica es is aised abo e, and his is due o p e e - en ial ampli ica ion o ce ain clonal DNA agmen s de i ed om he o iginal genomic DNA agmen s. PCR duplica es can gi e ise o he si ua ion whe e one allele is o e ep esen ed in he esul ing sequence da a and causes p oblems wi h di e en ia ing homozygous and he e ozygous indi iduals a ha locus (Schweyen e al. 2014). Applica ions o RAD sequencing in aquacul u e Since i s i s desc ip ion by Bai d e al. (2008), RAD-Seq has quickly sp ead h ough di e en ields o gene ic esea ch, and i has been used in di e en aquacul u e spe- cies o cons uc gene ic maps (e.g. Recknagel e al. 2013; Gonen e al.2014), o compa a i e genomics (e.g. Kakioka e al. 2013; Manousaki e al. 2015), o mapping genes associa ed wi h p oduc ion ai s (e.g. Hous on e al. 2012; Shao e al. 2015; Fu e al. 2016), mapping sex de e mining loci (e.g. Palaiokos as e al. 2013a,b), s udying popula ion dynamics (e.g. B adic e al. 2013), o ishe ies manage- men (e.g. Ogden e al. 2013), assembling e e ence gen- omes (e.g. Tine e al. 2014) o gene a ing SNP esou ces o u u e SNP a ay de elopmen (e.g. Hous on e al. 2014; Pal i e al. 2014). A summa y o he s udies pe - o med di ec ly ele an o aquacul u e is de ailed below and in Table 2. Gene ic ma ke disco e y o SNP a ay de elopmen Ea ly s udies using RAD-Seq ypically ocussed on sim- ply gene a ing a gene ic ma ke esou ce o nonmodel o ganisms. When he genome size o he a ge species is la ge, hen whole genome ( e)sequencing is a guably no cos -e ec i e o SNP disco e y ac oss many indi- iduals, and genome complexi y educ ion is ad an a- geous. As such, RAD-Seq and simila echniques enabled a s ep change in he numbe o gene ic ma ke s (SNPs) a ailable o se e al species (e.g. s u geon, Aci- pense genus, Ogden e al. 2013; o ainbow ou , Pal i e al. 2014), and hese ha e subsequen ly been used o se e al high- esolu ion gene ic s udies. SNPs gene a ed by RAD echniques ha e also been applied o p oduce SNP a ays o se e al aquacul u e species, including A lan ic salmon (Hous on e al. 2014), ainbow ou Re iews in Aquacul u e (2018) 10, 670–682 ©2017 The Au ho s. Re iews in Aquacul u e Published by Wiley Publishing Asia P y L d 674 D. Robledo e al. (Pal i e al. 2015a) and Paci ic oys e (Lapegue e al. 2014). Wi h he educ ion in sequencing cos s o e ecen yea s, whole genome ( e)sequencing (i.e. pool- sequencing, Schl€ o e e e al. 2014) has become inc eas- ingly iable. Howe e , RAD-like echniques s ill hold a signi ican ad an age o SNP disco e y when (i) he e is no e e ence genome a ailable, and (ii) only a med- ium densi y SNP esou ce is equi ed. Linkage maps and e e ence genome assembly Res ic ion-si e associa ed DNA sequencing echniques ha e been widely used in aquacul u e species o cons uc - ing gene ic maps based on ecombina ion e en s in de ined c osses. Such medium densi y SNP linkage maps a e use ul ools o downs eam applica ions such as QTL mapping, compa a i e genomic and gene mining, o popula ion Table 2 Summa y o aquacul u e-o ien ed s udies using es ic ion-si e associa ed DNA sequencing (RAD-Seq) S udy Species Aim Technique Samples SNPs Families Salmonids Hous on e al. (2012) Salmo sala Disease esis ance QTL (IPNV) RAD 32 6712 Two amilies Gonen e al. (2014) Salmo sala Linkage map RAD 96 8257 Two amilies Campbell e al. (2014) Onco hynchus mykiss Disease esis ance QTL (BCWD and IHNV) RAD 456 4661 40 amilies Pal i e al. (2014) Onco hynchus mykiss SNP esou ce RAD (92) 19 145 168 19 gene ic lines Pal i e al. (2015b) Onco hynchus mykiss Disease esis ance QTL (BCWD) RAD 252 5612/4946 Two amilies Liu e al. (2015b) Onco hynchus mykiss Co isol esponse o c owding QTL RAD 234 4874 One amily Liu e al. (2015b) Onco hynchus mykiss Disease esis ance QTL (BCWD) and spleen size QTL RAD 301 7849 Two hal -sib amilies Vallejo e al. (2016) Onco hynchus mykiss Genomic selec ion (BCWD) RAD 711 24 465 81 amilies E e e and Seeb (2014) Onco hynchus shawy scha The mo ole ance and g ow h QTL RAD 422 3534 Six amilies La son e al. (2016) Onco hynchus ne ka The mo ole ance and g ow h QTL RAD 491 11 457 Fi e amilies Nonsalmonid ish Palaiokos as e al. (2013b) O eoch omis nilo icus Sex de e mina ion QTL RAD 88 3904/4477 Two amilies Palaiokos as e al. (2015a) O eoch omis nilo icus Sex de e mina ion QTL ddRAD 372 1279 Fi e amilies Palaiokos as e al. (2013a) Hippoglossus hippoglossus Sex de e mina ion QTL RAD 93 7572/5954 2 hal -sib amilies Palaiokos as e al. (2015b) Dicen a chus lab ax Sex de e mina ion QTL RAD 187 6706 4 +4 hal -sib amilies Wang e al. (2015a,b) Scoph halmus maximus Sex de e mina ion and g ow h QTL RAD 151 6647 One amily B own e al. (2016) Polyp ion oxygeneios Sex de e mina ion and g ow h QTL ddRAD 59 1609 One amily Manousaki e al. (2015) Pagellus e y h inus Linkage map ddRAD 99 920 One amily Shao e al. (2015) Pa alich hys oli aceus Disease esis ance QTL (Vib io anguilla um) RAD 218 13 362 One amily Palaiokos as e al. (2016) Spa us au a a Disease esis ance genomic selec ion 2b-RAD 777 12 085 75 amilies Wang e al. (2015a,b) La es calca i e G ow h QTL ddRAD 144 3349 One amily Fu e al. (2016) Hypoph halmich hys nobilis G ow h QTL 2b-RAD 119 3323 One amily In e eb a es Jiao e al. (2014) Chlamys a e i Sex de e mina ion and g ow h QTL 2b-RAD 98 7458 One amily Li and He (2014) Pinc ada uca a G ow h QTL RAD 100 1381 One amily Shi e al. (2014) Pinc ada uca a G ow h QTL 2b-RAD 98 10 577 One amily Tian e al. (2015) Apos ichopus japonicas G ow h QTL 2b-RAD 102 11 306 One amily Lu e al. (2016) Ma supenaeus japonicus The mo ole ance and g ow h QTL RAD 152 9829 One amily Dou e al. (2016) Pa inopec en yessoensis Genomic selec ion (g ow h) 2b-RAD 349 2364 Fi e amilies Ren e al. (2016) Halio is di e sicolo G ow h QTL RAD 142 3317 One amily Re iews in Aquacul u e (2018) 10, 670–682 ©2017 The Au ho s. Re iews in Aquacul u e Published by Wiley Publishing Asia P y L d 675 RAD sequencing in aquacul u e genomic s udies. Fo example, RAD-based linkage maps ha e been c ea ed o A lan ic salmon (Gonen e al. 2014), channel ca ish (Li e al. 2014), Japanese lounde (Shao e al. 2015), u bo (Wang e al. 2015b) and Asian seabass (Wang e al. 2015a). Gene ic maps based on RAD-Seq ha e also con ibu ed o mapping and o ien a ion o sca olds o e e ence genome assemblies o key aquacul u e species such as Eu opean sea bass (Tine e al. 2014), ainbow ou (Be helo e al. 2014), Japanese eel (Kai e al. 2014), hal - smoo h ongue sole (Chen e al. 2014) and u bo (Figue as e al. 2016). While NGS echnology has enabled apid and cheap e e ence genome assemblies, hey a e ypically ag- men ed and incomple e. Fu he , assembly e o s a e qui e common, and linkage maps can also assis wi h esol ing mis-assemblies (Fie s 2015; Tsai e al. 2016a). Aquacul u e species ypically ha e an amenable amily s uc u e o high- esolu ion linkage maps, due o he high ecundi y esul ing in la ge ull and hal sibling amilies. Linkage maps can also be used in conjunc ion wi h physical e e - ence genome sequences o de ec a ia ion in ecombina- ion a es ac oss he genome, wi h implica ions o downs eam applica ions (e.g. LD be ween ma ke s and QTL in associa ion mapping s udies). Mapping QTL associa ed wi h ai s o economic impo ance The a e o applica ion o genomic echnology o aquacul- u e species ends o e lec he deg ee o scien i ic and comme cial in e es o hose species. This is ypically mo i- a ed by he in e es o unde s anding he gene ic basis o economically-impo an p oduc ion ai s, o example g ow h, disease esis ance o sex de e mina ion. Resea ch- e s wo king in he high- alue salmonid species we e amongs he i s o exploi RAD-Seq echniques, e alua - ing esis ance o di e en pa hogens causing high economic losses, including in ec ious panc ea ic nec osis in A lan ic salmon (Hous on e al. 2012), and in ec ious hema opoi- e ic nec osis (Campbell e al. 2014) and bac e ial cold wa e disease (Campbell e al. 2014; Liu e al. 2015a; Pal i e al. 2015b) in ainbow ou . Based on ea ly successes, and gi en he impo ance o disease esis ance o mode n aquacul u e b eeding p og ammes (Y a~ nez e al. 2014), la ge-scale p ojec s ha e been es ablished o apply RAD-like echniques o de ec ma ke s, and e en ually he genes and causal mu a ions in ol ed, o imp o ing esis ance. Fo example, he Eu opean Union unded FISHBOOST p ojec (www. ishboos .eu) is using RAD sequencing echniques o geno ype se e al housand animals om la ge-scale disease challenge expe imen s in ainbow ou , common ca p, Eu opean sea bass, gil head sea b eam (Spa us au a a) and u bo . These geno ype and pheno ype da a will be used o es ima e gene ic pa ame e s, map disease esis ance QTL and e alua e genomic p edic ion app oaches o disease esis ance b eeding. In addi ion o disease esis ance, RAD-Seq associa ion s udies ha e been widely applied o mapping QTL a ec - ing a ange o o he p oduc ion- ele an ai s, pa icula ly in salmonid species. These include spleen size (Liu e al. 2015a) and co isol esponse (Liu e al. 2015b) in ainbow ou , and he mal ole ance and g ow h in Onco hynchus ne ka, he sockeye salmon (La son e al. 2016). Ou wi h he salmonid gene a, RAD-Seq has been pe o med o map loci a ec ing disease esis ance in oli e lounde (Pa aly- ch hys oli aceous, Shao e al. 2015), and g ow h in bighead ca p (Hypoph halmich hys nobilis,Fue al. 2016) and u - bo (Wang e al. 2015b). In addi ion, RAD-like echniques ha e been e y popula o ma ke disco e y and QTL mapping in bi al e shell ish including Chinese scallop (A gopec en i adians; Jiao e al. 2014), Akoya pea l oys e (Pinc a a uca a; Li & He 2014; Shi e al. 2014), a iously colou ed abalone (Halio is di e sicolo ; Ren e al. 2016; Yesso scallop (Pa inopec en yessoensis; Dou e al. 2016) and ha e also been applied in he sh imp ku uma p awn (Ma - supenaeus japonicas;Lue al. 2016) and one echinode m, he sea cucumbe (Apos ichopus japonicus; Tian e al. 2015). In e es ingly, 2b-RAD has been he mos common echnique in bi al es, while in in ish, adi ional RAD has been mo e widely u ilized. Using RAD o s udy sex de e mina ion Sex de e mina ion (SD) is one o he mos c i ical ai s o many aquacul u e species, as pheno ypic sex is o en no e i- den in ju eniles and sexual dimo phism in g ow h a e is commonly obse ed. SD is complex in many ish species, o en wi h polygenic con ol and an en i onmen al compo- nen ( e iewed in Ma  ınez e al. 2014), and he applica ion o la ge geno yping p ojec s has been s ongly ecommended o sc een o SD loci in ish (e.g. Pan e al. 2016). RAD-like echniques ha e clea ly boos ed ou knowledge o SD in aquacul u e, wi h s udies in Nile ilapia (Palaiokos as e al. 2013a, 2015a), A lan ic halibu (Hippoglossus hippoglossus, Palaiokos as e al. 2013b), Eu opean sea bass (Palaiokos as e al. 2015b) and u bo (Wang e al. 2015b) inding pu a i e sex de e mining loci. Con olling sex a io is no only in e - es ing o ob ain highe g ow h a es, bu also o a oid size dispe sion o o delay sexual ma u i y. Fu he , he e a e some clea examples, like he s u geon, whe e he comme cial ad an age o ea ing ish o one sex o e he o he is ob ious. Genomic selec ion app oaches While QTL mapping and MAS app oaches can be success- ul when he gene ic a chi ec u e o a ai sugges s a gene o majo e ec (e.g. IPNV esis ance, Hous on e al. 2008; Re iews in Aquacul u e (2018) 10, 670–682 ©2017 The Au ho s. Re iews in Aquacul u e Published by Wiley Publishing Asia P y L d 676 D. Robledo e al. Moen e al. 2009), imp o emen o polygenic ai s using genomic da a is mo e e ec i ely achie ed using genomic p edic ion o b eeding alues (Meuwissen e al. 2001). S udies o genomic selec ion in aquacul u e we e i s ca - ied ou in salmonid ish, wi h simula ed (Sonesson & Meuwissen 2009; Lillehamme e al. 2013) and empi ical (Ødeg  a d e al. 2014; Tsai e al. 2015, 2016b; Vallejo e al. 2016) da a, demons a ing he clea ad an ages o e pedi- g ee-based me hods. S udies using a ying ma ke densi ies o p edic ion in salmonids ha e highligh ed ha as ew as a housand SNPs may be adequa e o achie ing he gain in selec ion accu acy e sus pedig ee app oaches (Ødeg  a d e al. 2014; Tsai e al. 2015, 2016b). The e o e, i is eason- able o assume ha RAD-like echniques may be use ul o genomic selec ion in aquacul u e b eeding, as ypical RAD SNP da ase s comp ise a ew housand SNPs. Indeed, he po en ial o his app oach has al eady been highligh ed o esis ance o bac e ial cold wa e disease in ainbow ou (Vallejo e al. 2016), o g ow h in Yesso scallop (Dou e al. 2016), and o esis ance o pas eu ellosis in gil head sea b eam (Palaiokos as e al. 2016). Gene ic aceabili y and aquacul u e sus ainabili y One o he main conce ns o aquacul u e p oduce s and consume s is o minimize he en i onmen al impac o ish a ming. In his sense, aceabili y ools a e essen ial o assess he impac o aquacul u e escapees in na u al popula ions o dis inguish be ween a med and wild specimens. RAD-Seq has been u ilized o ob ain SNPs o s u geon aceabili y and conse a ion (Ogden e al. 2013), which will con ibu e o en o ce cu en legisla ion on aquacul u e and ishing p ac ices bu also aid on he handling o wild s ocks, c i ical o sus ainable aquacul u e. RAD-Seq is also he main ool o he Eu opean p ojec AquaT ace (aqua ace.eu), he esul s o which ha e been ecen ly p esen ed in he Eu o- pean Aquacul u e Socie y mee ing in Edinbu gh (Aquacul- u e Eu ope 2016). One o he AquaT ace objec i es was o assess he impac o escapees on na u al popula ions o Eu opean sea bass, gil head sea b eam and u bo , while also de eloping o ensically alida ed ools o aceabili y pu - poses. The esul s highligh ed he u ili y o RAD-Seq app oaches o cap u e popula ion o amily speci ic a ia ion making i a sui able ool o gene ic aceabili y and conse - a ion o na u al popula ions. This is o he ou mos impo - ance o sus ainable aquacul u e g ow h, leading o las ing economic bene i s, ood sa e y and social accep ance. RAD-Seq and SNP a ays, owa ds a peace ul co- exis ence The de elopmen o NGS has g ea ly inc eased he amoun o genomic esou ces a ailable in he mos impo an aquacul u e species, including genome assemblies o many o hem. Alongside RNA-Seq and whole genome sequenc- ing, RAD-Seq has con ibu ed signi ican ly o he a ailabil- i y o abundan gene ic ma ke s compa ed o a ew yea s ago. While RAD-Seq and simila echniques a e likely o emain he geno yping me hod o choice o species wi h ew genomic esou ces, se e al medium and high-densi y SNP a ays a e al eady a ailable o aquacul u e species (A lan ic salmon, Hous on e al. 2014; Y a~ nez e al. 2016; channel ca ish, Liu e al. 2014; common ca p, Xu e al. 2014a; ainbow ou , Pal i e al. 2015a; Paci ic oys e and Eu opean la oys e , Lap egue e al. 2014), and many mo e a e unpublished o cu en ly being p oduced and alida ed. Single nucleo ide polymo phism a ays a e a ype o DNA mic oa ay, whe e hyb idiza ion o allele-speci ic p obes esul s in a luo escen signal which can be mea- su ed o call a geno ype in a gi en loci. They ha e bo h ad an ages and disad an ages o e RAD-Seq app oaches (Table 3). Fo ins ance, he expe imen al p ocedu es and bioin o ma ic analyses a e much simple o he use o SNP a ays, equi ing less echnical knowledge and usually esul ing in a as e u na ound. The geno ype sco ing me hod is mo e obus and amenable o au oma ion, and he e o e less p one o e o s (Hong e al. 2012; Wall e al. 2014). The epea abili y and ep oducibili y a e highe o SNP a ays han RAD-Seq, and geno yped loci a e known in ad ance. Howe e , ha ing a ixed se o loci on he chip is also a disad an age, especially in species wi h s ong pop- ula ion s uc u e, because o asce ainmen bias whe eby he SNP se is biased o polymo phic ma ke s in he disco - e y popula ion(s). This p esen s a majo issue whe e aqua- cul u e s ains o a speci ic species a e highly a iable, and he u ili y o a SNP a ay will a y hugely depending on he ela ionship o he disco e y popula ion. RAD-like app oaches o e come his issue and also o e much g ea e lexibili y o he esea che in e ms o he a ge ed numbe o loci. Fu he , RAD-Seq cap u es a ia ion ha is speci ic o popula ions, amilies and indi iduals ha is likely o be missed om SNP a ay, which a e ypically biased owa ds common a ian s. Ano he pu a i e ad an age o RAD-like Table 3 Gene al compa ison o es ic ion-si e associa ed DNA sequencing (RAD-Seq) and single nucleo ide polymo phism (SNP) chips RAD-Seq SNP a ays Sample p ocessing Labo ious S aigh o wa d Bioin o ma ic analysis Complex Negligible Tu na ound ime Long Medium Accu acy Medium-high High Repea abili y Medium High Design Adjus able Fixed Cos Low Medium Re iews in Aquacul u e (2018) 10, 670–682 ©2017 The Au ho s. Re iews in Aquacul u e Published by Wiley Publishing Asia P y L d 677 RAD sequencing in aquacul u e echniques is ha he di ec cos o he expe imen is cheape , al hough he addi ional ime equi ed o lib a y p epa a ion and bioin o ma ics analyses should be consid- e ed in o any compa ison. In he nea u u e, genomic selec ion (GS) is likely o be a key echnique o b eeding p og ammes o many aquacul u e species, due o he demons able inc ease in selec ion accu acy e sus cu en pedig ee-based me hods. SNP a ays a e now ou inely used in li es ock b eeding p og ammes o GS and a e inc easingly u ilized in ech- nologically ad anced aquacul u e b eeding. Se e al s udies ha e shown ha only mode a e SNP ma ke densi y is equi ed o e ec i e GS in salmon (Ødeg  a d e al. 2014; Tsai e al. 2015, 2016b). Vallejo e al. (2016) compa ed bo h RAD-Seq and SNP a ays o GS o BCWD esis- ance in ainbow ou , inding simila selec ion accu a- cies o bo h echniques despi e highe ma ke densi y om he SNP chip (~40k SNP a ay e sus ~10k RAD- Seq). This may e lec high le els o linkage disequilib- ium in ypical aquacul u e amily selec ion p og ammes, whe eby ai eco ding is o en pe o med on close ela- i es o he selec ion candida es. The e o e, he highe ma ke densi y associa ed wi h SNP chips may be ad an- ageous when p edic ing b eeding alues in animals mo e dis an ly ela ed o he aining popula ion (Tsai e al. 2016b), o in species wi h g ea e e ec i e popula ion sizes and/o lowe le els o linkage disequilib ium. Howe e , gi en he ela i ely sho genomes o many nonsalmonid aquacul u e species (i.e. Eu opean sea bass –~763 Mb, o u bo –~658 Mb; A lan ic salmon –~2970 Mb), he ypical ma ke densi y gene a ed by RAD-like echniques may be pe ec ly adequa e o e ec- i e GS. Howe e , his needs o be es ed, as he ecombi- na ion equency and pa e ns o linkage disequilib ium ac oss he genome a e pe inen o he ques ion o ade- qua e ma ke densi y. Fu he educ ions in ma ke den- si y equi emen s a e likely o be obse ed when geno ype impu a ion app oaches a e used, o example geno yping pa en s a high densi y, and o sp ing o a small subse o he ma ke s. As al eady men ioned, RAD me hods allow o subs an ial lexibili y in e ms o numbe o geno yped ma ke s. In addi ion, lowe ing a e age sequence co e age in he o sp ing wi h pa en s sequenced a high co e age could be used o gene a e geno ype da a a a much lowe cos . Ta ge ed GBS echniques Bo h RAD-Seq and SNP a ays will also ha e o com- pe e wi h ecen ly de eloped geno yping me hods based on a ge ed geno yping by sequencing. Fo example, geno yping-in- housands by sequencing (GT-Seq, Camp- bell e al. 2015) is a me hod o a ge ed sequencing which ollows a mul iplex PCR app oach, whe e hun- d eds o housands o loci (amplicons) a e selec ed o geno yping. In his me hod, a mul iplex PCR using loci-speci ic p ime s ha also con ain Illumina sequenc- ing p ime s is used o ampli y he a ge ed egions. Unique ba codes o each sample a e added wi h a sec- ond PCR eac ion, ollowed by pooling and sequencing o samples. Unlike RAD echniques, his me hod equi es p e ious knowledge o design he assays, and he numbe o SNPs geno yped in a single un is limi ed o a ew housand. Simila echnologies a e now p o ided by majo geno yping echnology p o ide s, and i appea s likely o become one o he mos cos -e ec i e sys ems o geno yping a ge ed SNPs. O he GBS a ge ed-sequen- cing echniques ha e also been ecen ly de eloped, o example RAD cap u e (Rap u e), whe e p eselec ed RAD ags a e isola ed using cap u e p obes and hen sequenced (Ali e al. 2016). These a ge ed GBS echniques ha e he po en ial o become majo playe s in aquacul u e b eed- ing and gene ics due o hei simplici y and lexibili y. Howe e , in pa , hey su e om he same limi a ion as SNP a ays ha hey equi e p io knowledge and selec- ion o he SNPs ha a e use ul in he popula ion o in e es . Fu u e ou look Res ic ion-si e associa ed DNA sequencing echniques ha e d i en a majo inc ease in he applica ion o genomics o aquacul u e species. While he ca alogue o SNP a ays o aquacul u e species will inc ease in he coming yea s, i is likely ha RAD echniques will con inue o be widely applied. We an icipa e ha bo h echniques will co-exis o se e al yea s, and he choice o RAD-Seq o SNP chip will depend on he species and p ojec -speci ic ac o s. Fo example, i may be ha high- alue aquacul u e species wi h la ge genomes (e.g. salmonids) a e mo e sui able o SNP a ays, while lowe - alue species wi h smalle genomes (and/o highe le els o LD) a e mo e sui able o RAD echniques, al hough i will also depend on he esou ces a ailable o each pa icula p ojec . Ta ge ed GBS ech- niques like GT-Seq a e likely o ind a niche in geno yping hund eds o se e al housands o p e iously iden i ied SNPs ac oss many samples. Fu he , RAD echniques a e likely o emain he gold s anda d o new aquacul u e spe- cies and/o hose p oduced on a smalle scale, whe e SNP a ays a e no a ailable, and genomic esou ces a e sca ce. E en ually he cos o gene a ing and analysing sequence da a may d op o a le el whe e genome complexi y educ- ion is no longe equi ed, bu i seems unlikely in he sho e m. The e o e, RAD sequencing will con inue o lou ish in aquacul u e esea ch in he ollowing yea s and is likely o be ou inely applied o deli e he bene i s o genomic Re iews in Aquacul u e (2018) 10, 670–682 ©2017 The Au ho s. Re iews in Aquacul u e Published by Wiley Publishing Asia P y L d 678 D. Robledo e al.