唐振雙, 殷東, 尹立林, 馬云龍, 項(xiàng)韜, 朱猛進(jìn), 余梅, 劉小磊, 李新云, 邱小田, 趙書紅
豬基因組選擇“兩步走”策略的計(jì)算機(jī)模擬評(píng)估
唐振雙1, 殷東1, 尹立林1, 馬云龍1, 項(xiàng)韜1, 朱猛進(jìn)1, 余梅1, 劉小磊1, 李新云1, 邱小田2, 趙書紅1
1華中農(nóng)業(yè)大學(xué)動(dòng)物科技學(xué)院/農(nóng)業(yè)動(dòng)物遺傳育種與繁殖教育部重點(diǎn)實(shí)驗(yàn)室/農(nóng)業(yè)農(nóng)村部豬遺傳育種重點(diǎn)實(shí)驗(yàn)室/國(guó)家家畜工程技術(shù)研究中心,武漢 430070;2全國(guó)畜牧總站 北京 100107
基因組選擇育種自2001年被MEUWISSEN等提出以來,已廣泛應(yīng)用在奶牛、豬等重要家畜的育種中,并顯著加快了其重要經(jīng)濟(jì)性狀的遺傳改良速度。2017年,在全國(guó)畜牧總站的組織協(xié)調(diào)下,在全國(guó)生豬遺傳改良計(jì)劃框架內(nèi),豬全基因組選擇育種平臺(tái)項(xiàng)目正式啟動(dòng)?!尽勘M管基因組選擇在種豬選育中取得了良好的效果,基因分型技術(shù)的不斷升級(jí)也帶來了成本的持續(xù)下降,但對(duì)于我國(guó)多數(shù)核心育種場(chǎng)依然面臨著基因芯片分型個(gè)體數(shù)量不足、基因組選擇實(shí)施流程不完善等問題,限制了該技術(shù)的大規(guī)模推廣應(yīng)用。結(jié)合我國(guó)生豬育種的實(shí)際情況,研究提出了一種“終測(cè)選擇-早期選擇”的“兩步走”基因組選擇策略。“終測(cè)選擇”指在終測(cè)結(jié)束后利用一步法基因組BLUP對(duì)后備豬進(jìn)行遺傳評(píng)估,當(dāng)群體中芯片分型個(gè)體數(shù)量達(dá)到一定規(guī)模后進(jìn)行“早期選擇”。以杜洛克、長(zhǎng)白和大白三個(gè)種豬品種真實(shí)的50K基因芯片數(shù)據(jù)作為基礎(chǔ)群體對(duì)不同品種分別進(jìn)行大群模擬,共模擬4個(gè)世代,前3個(gè)世代作為基礎(chǔ)群體,第4個(gè)世代作為測(cè)試群體,每個(gè)個(gè)體模擬兩個(gè)性狀(中等遺傳力性狀和低遺傳力性狀),利用豬基因組選擇育種平臺(tái)基于HIBLUP軟件計(jì)算不同品種、不同性狀的育種值,比較一步法基因組BLUP和常規(guī)BLUP兩種方法的預(yù)測(cè)準(zhǔn)確性。根據(jù)測(cè)試群個(gè)體有無終測(cè)成績(jī)對(duì)其基因組育種值影響大小來評(píng)估早期選擇效果。分析表明在3個(gè)品種內(nèi)中等遺傳力性狀的終測(cè)選擇效果和早期選擇效果均好于低遺傳力性狀。一步法基因組BLUP的選擇準(zhǔn)確性均優(yōu)于常規(guī)BLUP的選擇準(zhǔn)確性,并且隨著測(cè)試群中芯片分型個(gè)體數(shù)量的增加、群體規(guī)模的擴(kuò)大,預(yù)測(cè)準(zhǔn)確性越來越高。一步法基因組BLUP的早期選擇效果好于常規(guī)BLUP,當(dāng)群體中芯片數(shù)量達(dá)到2 000張時(shí)就可以開展早期選擇,閹割排名后30%的個(gè)體,可以保證前1%的優(yōu)秀個(gè)體不會(huì)被錯(cuò)誤淘汰,并且隨著芯片數(shù)量的增加、早期選擇的效果會(huì)越來越好。基因組選擇“兩步走”的策略符合我國(guó)國(guó)情、容易在生豬育種中推廣實(shí)施。當(dāng)芯片數(shù)量較少時(shí),可以開展“終測(cè)選擇”,一定程度上提高選擇的準(zhǔn)確性,提高育種效率;當(dāng)芯片數(shù)量較多時(shí),可以開展“早期選擇”,對(duì)排名靠后的豬只個(gè)體進(jìn)行早期閹割,增加優(yōu)秀個(gè)體的測(cè)定量,增大選擇強(qiáng)度、加快遺傳進(jìn)展?!皟刹阶摺辈呗苑衔覈?guó)生豬產(chǎn)業(yè)基因組選擇育種的實(shí)際需求,該策略的實(shí)施將有利于推動(dòng)我國(guó)豬基因組選擇的應(yīng)用、加快種豬改良進(jìn)程。
基因組選擇;豬;兩步走;終測(cè)選擇;早期選擇
【研究意義】種豬質(zhì)量是影響生豬產(chǎn)業(yè)效益的核心要素,育種則是提升種豬質(zhì)量的關(guān)鍵措施[1]。基因組選擇育種是當(dāng)前效率最高的育種技術(shù),已廣泛應(yīng)用在奶牛、豬、雞等多個(gè)物種中[2-4]。2017年,在全國(guó)畜牧總站的組織協(xié)調(diào)下,我國(guó)豬全基因組選擇育種平臺(tái)項(xiàng)目正式啟動(dòng)[5]?!厩叭搜芯窟M(jìn)展】20世紀(jì)四十年代,HAZEL通過建立選擇指數(shù)的方法來選擇個(gè)體,標(biāo)志著現(xiàn)代育種正式開始[6]。七十年代,HENDERSON開創(chuàng)基于表型和系譜信息快速求解BLUP(best linear unbiased perdition)的方法,該方法極大地提高了選擇的準(zhǔn)確性,隨后的幾十年間BLUP成為遺傳評(píng)估的主要方法[7-9]。九十年代初,標(biāo)記輔助選擇(marker- assisted selection, MAS)方法興起,但較低的標(biāo)記密度使估計(jì)的準(zhǔn)確性較差[10-11]。隨著測(cè)序技術(shù)和計(jì)算機(jī)的快速發(fā)展,2001年MEUWISSEN等提出基因組選擇(genomic selection, GS),該方法能對(duì)一些低遺傳力的性狀、無法直接度量的性狀進(jìn)行選擇[12],并且能顯著提高育種值估計(jì)的準(zhǔn)確性、縮短世代間隔、降低群體近交水平的增長(zhǎng)速度等[13-15]。2010年,AGUILAR和CHRISTENSEN等提出“一步法”選擇(single-step genomic BLUP, SSGBLUP)用于育種值估計(jì),該方法能同時(shí)利用基因組信息和系譜信息對(duì)有基因型個(gè)體和無基因型個(gè)體進(jìn)行育種值的估計(jì),具有較高的預(yù)測(cè)準(zhǔn)確性[16-18]?!颈狙芯壳腥朦c(diǎn)】雖然基因組選擇在多個(gè)物種上都取得了良好的效果,國(guó)內(nèi)部分生豬龍頭企業(yè)也已嘗試?yán)迷摲椒ㄟM(jìn)行種豬選育,但核心育種群體的芯片分型個(gè)體數(shù)量不足、基因組選擇實(shí)施流程不完善等因素仍然限制著該技術(shù)的大規(guī)模推廣應(yīng)用[2,19]?!緮M解決的關(guān)鍵問題】本研究提出了一種“終測(cè)選擇-早期選擇”的“兩步走”基因組選擇策略,并以杜洛克、長(zhǎng)白和大白3個(gè)品種的真實(shí)基因芯片數(shù)據(jù)作為基礎(chǔ)群體,通過大群模擬對(duì)終測(cè)選擇和早期選擇效果進(jìn)行評(píng)估。結(jié)果表明“兩步走”策略有利于公司盡早開展基因組選擇育種、提高種豬效率和降低育種成本,也有利于我國(guó)豬基因組選擇育種技術(shù)快速推廣。
基于目前基因組選擇在我國(guó)豬育種領(lǐng)域應(yīng)用的現(xiàn)狀,本研究提出豬基因組選擇“兩步走”的策略,第一步:終測(cè)選擇,第二步:早期選擇。
開展終測(cè)選擇,即終測(cè)結(jié)束后利用SSGBLUP對(duì)后備豬進(jìn)行遺傳評(píng)估。終測(cè)選擇需要在終測(cè)前一個(gè)月收集生長(zhǎng)、體型外貌正常豬只的組織樣品進(jìn)行基因分型,終測(cè)時(shí)獲得基因型數(shù)據(jù)和表型數(shù)據(jù),在選留前一天利用全部的表型、系譜和基因型數(shù)據(jù),使用SSGBLUP計(jì)算目標(biāo)性狀的基因組育種值(genomic estimated breeding values,GEBV),進(jìn)一步構(gòu)建綜合選擇指數(shù),根據(jù)指數(shù)的排名進(jìn)行選留和分流。終測(cè)選擇能夠在芯片數(shù)量較少時(shí)盡早地開展基因組選擇育種,提高選擇準(zhǔn)確性和育種效率。
當(dāng)群體中芯片分型個(gè)體數(shù)量達(dá)到一定規(guī)模后進(jìn)行早期選擇。早期選擇是指在仔豬出生初期,利用基礎(chǔ)群數(shù)據(jù)及仔豬個(gè)體基因芯片數(shù)據(jù),對(duì)仔豬進(jìn)行育種值評(píng)估和綜合選擇指數(shù)計(jì)算,根據(jù)指數(shù)排名進(jìn)行選留和分流。早期選擇重點(diǎn)針對(duì)小公豬開展,具體做法是在仔豬出生當(dāng)天,收集待評(píng)估的初生仔豬的組織樣本,盡快進(jìn)行基因型檢測(cè),仔豬閹割前一天,將最新的基因型數(shù)據(jù)、系譜數(shù)據(jù)和表型數(shù)據(jù)上傳至豬基因組選擇育種平臺(tái),利用SSGBLUP模型計(jì)算目標(biāo)性狀基因組育種值和綜合選擇指數(shù),根據(jù)指數(shù)排名高低對(duì)小公豬進(jìn)行選留、對(duì)淘汰個(gè)體進(jìn)行閹割。對(duì)母豬也可以進(jìn)行早期選擇,基因型檢測(cè)做法同公豬,選留可以在斷奶前一天進(jìn)行,綜合指數(shù)高的選留、低的個(gè)體淘汰分流到擴(kuò)繁群或商品群。所有早期選擇的種豬,只要正常生長(zhǎng)就進(jìn)入測(cè)定站,終測(cè)后再進(jìn)行一次終測(cè)選擇,最終選定后備豬進(jìn)入核心群。
基因組選擇“兩步走”的策略符合我國(guó)國(guó)情,其優(yōu)勢(shì)主要有兩點(diǎn):一是當(dāng)芯片數(shù)量較少時(shí)可以開展“終測(cè)選擇”,一定程度上提高選擇的準(zhǔn)確性,提高育種效率;二是當(dāng)芯片數(shù)量比較多時(shí),可以開展“早期選擇”,對(duì)排名靠后的豬只個(gè)體進(jìn)行早期閹割,增加優(yōu)秀個(gè)體的測(cè)定量,增大選擇強(qiáng)度、加快遺傳進(jìn)展。
本研究采用Simer軟件(https://github.com/xiaolei- lab/SIMER),基于1 030頭杜洛克、530頭長(zhǎng)白和2 030頭大白的50K(50 697 SNPs)基因芯片數(shù)據(jù)進(jìn)行計(jì)算機(jī)模擬[20],利用基因芯片數(shù)據(jù)的染色體、物理位置和基因分型等信息,模擬群體的繁殖過程,在群體生成過程中同時(shí)模擬系譜數(shù)據(jù)、基因型數(shù)據(jù)和表型數(shù)據(jù)。
每個(gè)品種模擬4個(gè)世代,其中N1—N3世代的核心群作為基礎(chǔ)群體,N4世代的核心群作為測(cè)試群體。各群體的參數(shù)設(shè)置如下:每個(gè)世代核心群規(guī)模固定,杜洛克基礎(chǔ)母豬為1 000頭,長(zhǎng)白基礎(chǔ)母豬為500頭,大白基礎(chǔ)母豬為2 000頭,各品種核心群公豬為30頭。杜洛克產(chǎn)仔數(shù)為8頭/窩,長(zhǎng)白為13頭/窩,大白為14頭/窩,每窩產(chǎn)公豬、母豬的比例為1﹕1。杜洛克母豬的留種率為15%,長(zhǎng)白和大白母豬留種率為10%,各品種公豬的留種率為1%。每個(gè)世代各品種公豬更新率為100%,母豬更新率為60%。根據(jù)上述設(shè)置,杜洛克核心群每個(gè)世代能產(chǎn)8 000頭仔豬,從中挑選600頭母豬和30頭公豬作為核心群。長(zhǎng)白核心群每個(gè)世代能產(chǎn)6 500頭仔豬,從中挑選300頭母豬和30頭公豬作為核心群。大白核心群每個(gè)世代能產(chǎn)28 000頭仔豬,從中挑選1 200頭母豬和30頭公豬作為核心群。在測(cè)試群有表型的個(gè)體中隨機(jī)抽取10%或30%的個(gè)體進(jìn)行基因型數(shù)據(jù)模擬,群體內(nèi)所有個(gè)體都有完整的系譜記錄。
在基礎(chǔ)群體內(nèi)隨機(jī)抽取50%的個(gè)體進(jìn)行表型模擬,測(cè)試群內(nèi)所有個(gè)體都模擬表型,每個(gè)個(gè)體模擬2個(gè)性狀,性狀1的遺傳力設(shè)為0.3、性狀2的遺傳力設(shè)為0.1,每個(gè)性狀受100個(gè)QTNs(Quantitative Trait Nucleotides)控制,且QTN效應(yīng)分布服從正態(tài)分布。模擬使用的參數(shù)見表1。
群體內(nèi)芯片數(shù)量較少時(shí)實(shí)施終測(cè)選擇。在測(cè)試群(N4世代)內(nèi),分別采用基于系譜信息的BLUP和SSGBLUP兩種方法估計(jì)育種值,預(yù)測(cè)的準(zhǔn)確性用EBV(estimated breeding values)與TBV(true breeding values)之間的皮爾森相關(guān)系數(shù)表示,相關(guān)系數(shù)越大,表明育種值估計(jì)的準(zhǔn)確性越高。
表1 模擬參數(shù)
當(dāng)群體中芯片分型個(gè)體數(shù)量積累到一定規(guī)模后實(shí)施早期選擇,根據(jù)測(cè)試群個(gè)體有無終測(cè)成績(jī)對(duì)其基因組育種值影響大小來評(píng)估早期選擇效果。在每個(gè)品種的基礎(chǔ)群體(N1—N3世代)內(nèi)隨機(jī)抽取1 000或3 000個(gè)個(gè)體進(jìn)行基因型數(shù)據(jù)的模擬作為積累的芯片數(shù)據(jù),利用基礎(chǔ)群體內(nèi)積累的芯片數(shù)據(jù)、表型數(shù)據(jù)、系譜數(shù)據(jù)及測(cè)試群(N4世代)內(nèi)新檢測(cè)的芯片數(shù)據(jù)和系譜數(shù)據(jù),分別采用BLUP和SSGBLUP兩種方法計(jì)算測(cè)試群內(nèi)不同性狀的育種值。在測(cè)試群內(nèi),計(jì)算有終測(cè)表型時(shí)GEBV排名前1%或5%的個(gè)體被無表型時(shí)GEBV排名前30%、50%或70%覆蓋的比例,個(gè)體相同條件下覆蓋比例越高,表明早期選擇效果越好。
育種值估計(jì)和綜合選擇指數(shù)計(jì)算均在豬基因組選擇育種平臺(tái)(http://ubreed.pro:90/ GSPlatform/webroot/ user/login.html)完成,該平臺(tái)使用HIBLUP軟件(https://www.hiblup.com/)進(jìn)行育種值估計(jì)。
3個(gè)品種的數(shù)據(jù)概況見表2,基因型數(shù)據(jù)模擬基于真實(shí)的50K芯片數(shù)據(jù),標(biāo)記密度圖見電子附圖1。在每個(gè)品種內(nèi)隨機(jī)挑選1 000個(gè)個(gè)體,利用rMVP軟件進(jìn)行主成分分析(https://cran.r-project.org/web/packages/ rMVP/index.html),結(jié)果如圖1所示,3個(gè)品種的模擬數(shù)據(jù)質(zhì)量良好、滿足分析要求[21]。
芯片數(shù)量較少時(shí)實(shí)施終測(cè)選擇,即使用GEBV代替EBV進(jìn)行選種。在3個(gè)品種內(nèi)對(duì)比分析BLUP和SSGBLUP兩種方法育種值估計(jì)的準(zhǔn)確性,分析時(shí)數(shù)據(jù)使用情況見表2。在這里由于想對(duì)基因芯片較少的情況進(jìn)行分析,所以只用了測(cè)試群體的基因型數(shù)據(jù),不包括基礎(chǔ)群體的基因型數(shù)據(jù),分析結(jié)果見圖2和電子附表1。結(jié)果表明:(1)隨著測(cè)試群中芯片分型個(gè)體數(shù)量的增加、群體規(guī)模的擴(kuò)大,預(yù)測(cè)準(zhǔn)確性越來越高;(2)性狀1的預(yù)測(cè)準(zhǔn)確性高于性狀2,表明中等遺傳力的性狀育種值預(yù)測(cè)的準(zhǔn)確性比低遺傳力的性狀要高;而且不論是性狀1還是性狀2,SSGBLUP的準(zhǔn)確性始終高于BLUP,表明盡管芯片數(shù)量不多,但是終測(cè)選擇使用SSGBLUP效果更好。因此,建議在芯片數(shù)量較少時(shí)開展終測(cè)選擇,這有利于提高選擇的準(zhǔn)確性和育種效率。
表2 3個(gè)品種數(shù)據(jù)的概況
圖1 3個(gè)品種的主成分分析圖
評(píng)估早期選擇效果時(shí),假設(shè)群體已經(jīng)積累1 000張或3 000張芯片,在此基礎(chǔ)上分別根據(jù)GEBV和EBV在測(cè)試群內(nèi)進(jìn)行早期選擇,通過對(duì)比最優(yōu)秀個(gè)體保留比例來分析早期選擇效果,不同閹割比例下早期選擇的效果見圖3和電子附表2—4。結(jié)果發(fā)現(xiàn)SSGBLUP早期選擇效果好于BLUP,并且隨著芯片數(shù)量的增加、早期選擇的效果越來越好;性狀1的早期選擇效果略好于性狀2;不同品種、不同性狀的選擇效果趨勢(shì)類似,隨著群體規(guī)模的擴(kuò)大,早期選擇的效果也愈明顯。以杜洛克群體為例,在性狀1中,當(dāng)閹割排名后70% 時(shí),測(cè)試群中最優(yōu)秀的1%(有終測(cè)表型時(shí)的GEBV排名),根據(jù)常規(guī)BLUP預(yù)測(cè),有85.71%的個(gè)體不會(huì)被錯(cuò)誤閹割,而根據(jù)SSGBLUP預(yù)測(cè)(1 700張芯片),有95.71%的個(gè)體不會(huì)被錯(cuò)誤閹割;在性狀2中,比例分別是75.71%和97.14%。當(dāng)閹割比例為30%時(shí),測(cè)試群中最優(yōu)秀的1%(有終測(cè)表型時(shí)的GEBV排名)都不會(huì)被錯(cuò)誤閹割。這表明早期選擇是可行的,而且SSGBLUP比常規(guī)BLUP更好。
圖2 育種值估計(jì)的準(zhǔn)確性
該圖展示杜洛克品種早期選擇的部分結(jié)果。SSGBLUP方法估計(jì)育種值時(shí),基礎(chǔ)群體有1000張芯片、測(cè)試群有700張芯片;基于最優(yōu)秀的前1%(有終測(cè)表型時(shí)的GEBV排名)個(gè)體保留比例進(jìn)行早期選擇效果評(píng)估
通過數(shù)據(jù)模擬分析,分別對(duì)杜洛克、長(zhǎng)白、大白3個(gè)主流瘦肉豬品種進(jìn)行了分析,結(jié)果表明基因組選擇“兩步走”策略是可行的,也提示了“兩步走”的階段。育種公司可以根據(jù)基因芯片累積的數(shù)量,以2 000張為界限將基因組選擇分成終測(cè)選擇和早期選擇兩個(gè)階段實(shí)施,這種策略有利于育種公司提高選擇準(zhǔn)確性和育種效率、降低育種成本,同時(shí)也有利于在我國(guó)推廣豬基因組選擇育種技術(shù)。
比較了BLUP和SSGBLUP兩種方法育種值估計(jì)的準(zhǔn)確性,結(jié)果表明即使在芯片分型個(gè)體數(shù)量較少的情況下SSGBLUP也能提高估計(jì)的準(zhǔn)確性,本研究與前人的研究結(jié)果一致[22]。SSGBLUP可以同時(shí)利用系譜信息和基因組信息構(gòu)建親緣關(guān)系矩陣(H矩陣),由于使用了基因組信息,使個(gè)體間親緣關(guān)系估計(jì)得更準(zhǔn),所以育種值估計(jì)的準(zhǔn)確性比常規(guī)BLUP更高[23-24]。另外,有研究表明在低遺傳力性狀的評(píng)估中,間接法與直接法效果類似,但以Bayes為理論基礎(chǔ)的間接法在估計(jì)遺傳參數(shù)時(shí)需要多次迭代才能收斂,因此需要更多的計(jì)算時(shí)間和內(nèi)存[22, 25-26]。與其相比,SSGBLUP在保持一定計(jì)算準(zhǔn)確性的前提下,能夠顯著縮短計(jì)算時(shí)間、降低計(jì)算負(fù)擔(dān),而且,SSGBLUP無需構(gòu)建參考群就能實(shí)施基因組選擇[8, 26]?;诖?,建議在芯片較少的情況下,終測(cè)結(jié)束后利用SSGBLUP開展遺傳評(píng)估、進(jìn)行選種,這樣可盡早開展基因組選擇、提高育種效率。
本研究的結(jié)果表明當(dāng)芯片數(shù)量達(dá)到2 000張左右時(shí)可以開展早期選擇,在這種情況下,閹割后30%—50%的個(gè)體,最優(yōu)秀的前1%個(gè)體是比較安全的;而閹割后70%的個(gè)體時(shí),前1%的個(gè)體會(huì)有一定風(fēng)險(xiǎn)。另外,本研究是基于模擬數(shù)據(jù)進(jìn)行早期選擇效果的評(píng)估,在真實(shí)情況下,由于系譜記錄錯(cuò)誤或缺失、表型測(cè)定數(shù)量和質(zhì)量等諸多因素的影響,預(yù)測(cè)的準(zhǔn)確性會(huì)打折扣。閹割比例為50%和30%的結(jié)果類似,SSGBLUP在兩類性狀上都取得了較好的早期選擇效果,雖然在低遺傳力性狀的評(píng)估中會(huì)出現(xiàn)個(gè)別BLUP的選擇效果好于SSGBLUP選擇效果的情況,主要原因是早期選擇效果評(píng)估過程復(fù)雜,單次模擬就需要進(jìn)行144次的育種值估計(jì)和比較分析,但單次模擬的結(jié)果基本符合預(yù)期。通過觀察閹割比例為70%的結(jié)果,可知SSGBLUP的早期選擇效果要明顯好于BLUP的效果。
基于此,建議在2 000張芯片時(shí)閹割后30%的個(gè)體,隨著芯片數(shù)量的增多,閹割比例可以進(jìn)一步上升。早期選擇可以對(duì)無種用價(jià)值的公豬實(shí)施早期閹割,有利于減少肉公豬比例,降低育種成本[27]。同時(shí),早期選擇還能夠通過擴(kuò)大初選范圍,增加選擇強(qiáng)度,加快遺傳進(jìn)展。因此,隨著育種公司基因組選擇育種的逐漸開展,建議擇機(jī)實(shí)施早期選擇。
本研究提出基因組選擇“兩步走”的策略,第一步,當(dāng)芯片分型個(gè)體數(shù)量較少的時(shí)候開展終測(cè)選擇;第二步,當(dāng)芯片分型個(gè)體數(shù)量達(dá)到一定規(guī)模后,開展早期選擇。研究結(jié)果表明即便在芯片數(shù)量較少時(shí),SSGBLUP的育種值估計(jì)的準(zhǔn)確性也會(huì)高于常規(guī)BLUP,終測(cè)選擇有效;當(dāng)芯片數(shù)量達(dá)到2 000張左右時(shí)開展早期選擇,閹割后30%的個(gè)體是比較安全的,隨著芯片數(shù)量增加,閹割比例可以上升,早期選擇不但能夠降低育種成本,還能提升選擇準(zhǔn)確性和選擇強(qiáng)度,實(shí)施早期選擇對(duì)提高育種公司效益是有幫助的。本研究提出的基因組選擇“兩步走”的策略容易實(shí)施推廣,符合我國(guó)當(dāng)前的國(guó)情,也有利于推動(dòng)豬基因組選擇育種技術(shù)在養(yǎng)豬業(yè)中的廣泛應(yīng)用。
[1] 彭中鎮(zhèn). 《全國(guó)生豬遺傳改良計(jì)劃》解讀及實(shí)施中有關(guān)問題的討論(待續(xù)). 豬業(yè)科學(xué), 2011, 28(8): 100-103. doi:10.3969/j.issn.1673- 5358.2011.08.037.
PENG Z Z. Discussion on the relevant issues in the interpretation and implementation of the “National Swine Improvement Program”. Swine Industry Science, 2011, 28(8): 100-103. doi:10.3969/j.issn. 1673-5358.2011.08.037. (in Chinese)
[2] KNOL E. ASAS-EAAP Exchange Speaker Talk: Is genomic selection beneficial for the pig industry? Journal of Animal Science, 2019, 97l(1): 1-7.
[3] CESARANI A, GASPA G, CORREDDU F, CELLESI M, DIMAURO C, MACCIOTTA N P P. Genomic selection of milk fatty acid composition in Sarda dairy sheep: Effect of different phenotypes and relationship matrices on heritability and breeding value accuracyJournal of Dairy Science, 2019, 102l(4): 3189-3203.
[4] MRODE R, OJANGO J M K, OKEYO A M, MWACHARO J M. Genomic selection and use of molecular tools in breeding programs for indigenous and crossbred cattle in developing countries: Current status and future prospects. Frontiers in Genetics, 2018, 9: 694. doi:10.3389/fgene.2018.00694.
[5] 丁向東, 邱小田, 王志剛, 楊紅杰, 陳瑤生, 張勤. 全國(guó)生豬遺傳改良計(jì)劃(2009—2020)給我國(guó)生豬種業(yè)帶來的變化和未來工作建議. 中國(guó)畜牧雜志, 2020(6): 169-174.
DING X D, QIU X T, WANG Z G, YANG H J, CHEN Y S, ZHANG Q. The changes and suggestions brought by the National Pig Genetic Improvement Program (2009-2020) for pig breeding industry in China. Chinese Journal of Animal Science, 2020(6): 169-174. (in Chinese)
[6] HAZEL L N, LUSH J L. The efficiency of three methods of selectionJournal of Heredity, 1942, 33l(11): 393-399.
[7] HENDERSON. Inverse of a matrix of relationships due to sires and maternal grandsiresElsevier, 1975, 58l(12): 1917-1921.
[8] 尹立林, 馬云龍, 項(xiàng)韜, 朱猛進(jìn), 余梅, 李新云, 劉小磊,趙書紅.全基因組選擇模型研究進(jìn)展及展望畜牧獸醫(yī)學(xué)報(bào), 2019, 50(2): 233-242.
YIN L L, MA Y L, XIANG T, ZHU M J, YU M, LI X Y, LIU X L, ZHAO S H. The progress and prospect of genomic selection models. Acta Veterinaria et Zootechnica Sinica, 2019, 50(2): 233-242. (in Chinese)
[9] VANRADEN P M. Efficient methods to compute genomic predictions. Journal of Dairy Science, 2008, 91(11): 4414-4423. doi:10.3168/ jds.2007-0980.
[10] FERNANDO R L, GROSSMAN M. Marker assisted selection using best linear unbiased predictionGenetics Selection Evolution, 1989, 21l(4): 467.
[11] GODDARD M E, HAYES B J. Mapping genes for complex traits in domestic animals and their use in breeding programmes. Nature Reviews Genetics, 2009, 10(6): 381-391. doi:10.1038/nrg2575.
[12] MEUWISSEN T H E, HAYES B J,GODDARD M E. Prediction of total genetic value using genome-wide dense marker mapsGenetics, 2001, 1571(4): 1819-1829.
[13] WRAY N R, YANG J, HAYES B J, PRICE A L, GODDARD M E, VISSCHER P M. Pitfalls of predicting complex traits from SNPs. Nature Reviews Genetics, 2013, 14(7): 507-515. doi:10.1038/nrg3457.
[14] HICKEY J M, CHIURUGWI T, MACKAY I, POWELL W, HENDRE P. Genomic prediction unifies animal and plant breeding programs to form platforms for biological discoveryNature Genetics, 2017, 491(9): 1297-1303.
[15] HILL W G. Applications of population genetics to animal breeding, from wright, fisher and lush to genomic prediction. Genetics, 2014, 196(1): 1-16. doi:10.1534/genetics.112.147850.
[16] CHRISTENSEN O F, LUND M S. Genomic prediction when some animals are not genotypedGenetics Selection Evolution, 2010, 42l(1): 2.
[17] AGUILAR I, MISZTAL I, JOHNSON D L, LEGARRA A, TSURUTA S, LAWLOR T J. Hot topic: a unified approach to utilize phenotypic, full pedigree, and genomic information for genetic evaluation of Holstein final scoreJournal of Dairy Science, 2010, 931(2): 743-752.
[18] GUO X, CHRISTENSEN O F, OSTERSEN T, WANG Y, LUND M S, SU G. Improving genetic evaluation of litter size and piglet mortality for both genotyped and nongenotyped individuals using a single-step method. Journal of Animal Science, 2015, 93(2): 503-512. doi:10. 2527/jas.2014-8331.
[19] 周磊, 楊華威, 趙祖凱, 楊紅杰, 劉劍鋒. 基因組選擇在我國(guó)種豬育種中應(yīng)用的探討. 中國(guó)畜牧雜志, 2018, 54(3): 4-8. doi:10. 19556/j.0258-7033.2018-03-004.
ZHOU L, YANG H W, ZHAO Z K, YANG H J, LIU J F. The application of genomic selection in Chinese pig breeding industry. Chinese Journal of Animal Science, 2018, 54(3): 4-8. doi:10.19556/ j.0258-7033.2018-03-004. (in Chinese)
[20] Antonio R, Richard C, Nabeel A, Andreia A, Alan A, Jonathan B, Christian B, Carol C, Richard C, Patrick D, et al. Design of a high density SNP genotyping assay in the pig using SNPs identified and characterized by next generation sequencing technologyPLoS One, 2009, 4l(8): e6524.
[21] YIN L L, ZHANG H H, TANG Z S, XU J Y, YIN D, ZHANG Z W, YUAN X H, ZHU M J, ZHAO S H, LI X Y, LIU X L. rMVP: A Memory-efficient, Visualization-enhanced, and Parallel-accelerated tool for Genome-Wide Association StudybioRxiv, 2020, 2020.08. 20.258491.
[22] MISZTAL I, LEGARRA A. Invited review: Efficient computation strategies in genomic selection. Animal, 2017, 11(5): 731-736. doi:10.1017/s1751731116002366.
[23] MISZTAL I, AGGREY S E, MUIR W M. Experiences with a single- step genome evaluation. Poultry Science, 2013, 92(9): 2530-2534. doi:10.3382/ps.2012-02739.
[24] Nielsen Bjarne, Ostersen Tage, Guosheng Su, Christensen O, Henryon M. Use Of Genomic SNP Information In Pig Breeding, 2010.
[25] 彭瀟, 尹立林, 梅全順, 王海燕, 劉小磊, 朱猛進(jìn), 李新云, 付亮亮, 趙書紅. 豬主要經(jīng)濟(jì)性狀的基因組選擇研究. 畜牧獸醫(yī)學(xué)報(bào), 2019, 50(2): 439-445. doi:10.11843/j.issn.0366-6964.2019.02.023.
PENG X, YIN L L, MEI Q S, WANG H Y, LIU X L, ZHU M J, LI X Y, FU L L, ZHAO S H. A study of genome selection based on the porcine major economic traits. Acta Veterinaria et Zootechnica Sinica, 2019, 50(2): 439-445. doi:10.11843/j.issn.0366-6964.2019.02.023. (in Chinese)
[26] ZHANG H, YIN L, WANG M, YUAN X, LIU X. Factors affecting the accuracy of genomic selection for agricultural economic traits in maize, cattle, and pig populations. Frontiers in Genetics, 2019, 10: 189. doi:10.3389/fgene.2019.00189.
[27] 張金鑫, 唐韶青, 宋海亮, 高虹, 蔣堯, 江一凡, 彌世榮, 孟慶利, 于凡, 肖煒, 云鵬, 張勤, 丁向東. 北京地區(qū)大白豬基因組聯(lián)合育種研究. 中國(guó)農(nóng)業(yè)科學(xué), 2019, 52(12): 2161-2170. doi:10.3864/j.issn. 0578-1752.2019.12.013.
ZHANG J X, TANG S Q, SONG H L, GAO H, JIANG Y, JIANG Y F, MI S R, MENG Q L, YU F, XIAO W, YUN P, ZHANG Q, DING X D. Joint genomic selection of Yorkshire in Beijing. Scientia Agricultura Sinica, 2019, 52(12): 2161-2170. doi:10.3864/j.issn.0578-1752.2019. 12.013. (in Chinese)
附表1 育種值估計(jì)的準(zhǔn)確性
附表2 杜洛克早期選擇的效果
附表3 長(zhǎng)白早期選擇的效果
附表4 大白早期選擇的效果
To Evaluate the “Two-Step” Genomic Selection Strategy in Pig by Simulation
1College of Animal Science and Technology, Huazhong Agricultural University /Key Laboratory of Agricultural Animal Genetics, Breeding and Reproduction of Ministry of Education/Key Laboratory of Swine Genetics and Breeding of Ministry of Agriculture and Rural Affairs/National Engineering and Technology Research Center for Livestock, Wuhan 430070;2National Animal Husbandry Service, Beijing 100107
【】Since genomic selection (GS) was proposed by Meuwissen et al. in 2001, it has been widely used in the breeding of dairy cows, pigs, and other livestock, and has significantly improved the speed of genetic gain of various economic traits. In 2017, with the organization and coordination of the National Grazing Headquarter Station and within the framework of the National Swine Improvement Program, the genomic selection platform for pig breeding was officially launched. Although genomic selection has made positive achievements in pig breeding, and the developing of advanced genotyping technology reduced the costs dramatically, some issues were still existed, including the insufficient number of genotyped individuals in majority of core breeding farms and the inappropriate implementation processes has restricted its wide application in practice.【】In combination with the actual situation of domestic pig breeding, the “two-step” strategy for genomic selection was proposed in this study, that is, the off-test evaluation and the early-stage prediction. Off-test evaluation referred to the genetic evaluation of replacement pigs by SSGBLUP after off-test, and early-stage prediction was carried out when the number of chips reached a certain scale. 【】In this study, the 50 k chip datasets of three breeds consisting of Duroc, Landrace, and Yorkshire were used as the base group to simulate the large-scale population of different breeds, respectively. The four generations were simulated: the first three generations were treated as the base population, and the fourth generation as the test population, two traits with medium and low heritability was simulated for each individual. The estimated breeding values of SSGBLUP and traditional BLUP model for different traits were calculated by the pig genomic selection platform based on the HIBLUP software. The predictive performance of early-stage was evaluated according to whether the individual’s testing records have influence their genomic estimated breeding values (GEBV) in test population. 【】The results showed that the predictive performance of off-test evaluation and early-stage for traits with medium heritability were better than those with low heritability. The selection accuracy of SSGBLUP was better than traditional BLUP. Moreover, with the increase of the number of chips and the expansion of the population size, the prediction accuracy was higher. The early-stage predictive performance of SSGBLUP was better than that of traditional BLUP, the early-stage prediction could be carried out when the number of genotyped pigs reached about 2 000, and castrating the last 30% individuals according to GEBV could ensure that the top 1% excellent individuals would not be mistakenly eliminated. And the prediction accuracy performance was increasing with the increased number of genotyped pigs. 【】The “two-step” strategy pretty was conformed to the state of domestic breeding program, and was easy to implement and promote the pig breeding in China. When the number of genotyped pigs was small, off-test evaluation could be carried out to improve the accuracy of selection, as well as efficiency, to a certain extent; when the number of genotyped pigs was large, early-stage prediction could be performed by castrating the pigs on the lower rank of GEBV, which could increase the amount of testing for more excellent pigs, and could also strength the selection intensity and accelerate the genetic gain. The “two-step” strategy was in line with the actual requirements of genomic selection in pig industry. The implementation of this strategy could further promote the application of genomic selection and speed up the genetic gain in pig breeding.
genomic selection; pig; two-step; off-test evaluation; early-stage prediction
10.3864/j.issn.0578-1752.2021.21.016
2020-10-19;
2021-02-04
國(guó)家自然科學(xué)基金面上項(xiàng)目(32072725)、國(guó)家生豬產(chǎn)業(yè)體系(CARS-35)、中央高?;究蒲袠I(yè)務(wù)費(fèi)專項(xiàng)資金資助項(xiàng)目(662020DKPYCFA006007)、豬基因組選擇育種創(chuàng)新群體(2020CFA006)
唐振雙,E-mail:zst@webmail.hzau.edu.cn。通信作者趙書紅,E-mail:shzhao@mail.hzau.edu.cn。通信作者邱小田,E-mail:23753846@QQ.com
(責(zé)任編輯 林鑒非)