抗性育種研究方法探討-赤霉病

其實不管做哪個性狀，以下幾個名詞算是過去二三十年以及未來若干年在植物育種領域所能看到的大部分文章了,這裡先跟大家簡單總結一下。另外對於幾種方法的評論其實也主要來源於下面解讀的文章，加上小編的一些拙見。

最傳統也是目前最有效的育種方法：Phenotypic selection

後來從動物研究領域引入了四種結合遺傳學的方法：

鑒定優良性狀遺傳位點的方法：QTL mapping和GWAS

優良性狀實際應用的方法：MAS和GS

QTL mapping做了幾十年，GWAS也火了十來年了，但感覺對育種上也就對像rust這種單基因抗病的有點作用。不過隨著小麥序列的釋放，讓我們對這四種方法拭目以待。

目前感覺植物上未來最有前途的方法：High-throughput phenotyping。

接下來進入今天正文：最近關於赤霉病的文章還是挺多的！小編過去三周已經給大家介紹了幾篇文章，有兩篇也都是最新發表的。剛剛過去的這一周又有一篇相關文章發表，雖然所在雜誌的影響因子不是很高，但實驗的內容還是非常紮實的。小編今天也正好借這篇文章來談談目前赤霉病抗性育種研究的方法！

文章題目如下，第一作者是Thomas Miedaner，通訊作者是Tobias Wurschum，單位是成立於1818年的德國霍恩海姆大學（University of Hohenheim），在農業研究領域算是非常好的一所大學了。另外，我們上次介紹的Hermann Buertmayr也是本文的作者之一。

文章概括：

本文的主體是用GWAS的方法來鑒定一個durum germplasm中的抗性位點。但最有趣的部分是文章利用所鑒定到的主效QTL以及所有QTL來評估MAS，GS，和Phenotypic Selection三種方法在實際抗性育種中的應用價值。最後的結論大家一定會覺得有點意思，但應該也不會驚訝。

材料：

A highly diverse panel of 184 durum wheat lines.

In brief, this diversity panel comprised 170 winter and 14spring types, including old as well as modern cultivars with economic importance as well as current breeding lines.

Six groups according to their geographic origin: Austria and Germany (Group 1), France (Group 2), Canada and United States (Group 3), Italy and Spain (Group 4), Hungary (Group 5) and Bulgaria, Romania, Russia, Slovakia,Turkey and Ukraine (Group 6).

田間設計：

小編前段時間講過一些field design的內容，目前也還在繼續學習當中，這篇文章用到的是a-lattice design，好像在國內和美國用的並不是很多。

In total, five environments (location-by-year combinations) were tested.

The genotypes were sown mechanically in ana-lattice design(Schutz & Cockerham, 1962) with three replications in HOH and OLI, and two replications in TUL, as observation plots of one row with 1.25 m length.

接種方法：

田間Spray

Disease Evaluation：

本文採用的方法在genetic study中好像並不常見，但是在育種項目中還是蠻多的：一個數值（1-9）綜合考慮Incidence和Severity，這在育種中算是簡單有效的一種方法。但是對於genetic study來說，FHB的incidence和severity經常是由不同遺傳位點控制的，所以拿這樣一個數值作為唯一的phenotyping來進行GWAS算是本文的一個弱點吧。另外，DON的含量其實對於育種有更大的價值，本文也沒有關於這項的數據。

FHB was rated from 1 to 9, where 1 indicates no visible symptoms within a plot, and 2–9 represents 95%, respectively, of all spikelets of a plot showing symptoms

This rating reflects the number of infected spikes per plot(type I resistance) as well as the number of infected spikelets per spike (typeII resistance) in a single score.

GS和Prediction ability：如果有做相關方面的可以詳細閱讀一下這部分的內容：

The genomic prediction was made by Ridge Regression-BLUP with the R package「rrBLUP」

To compare the predictive ability of MAS and genomic selection, we used the four markers explaining more than 5% of the genotypic variance for MAS, while genomic prediction was based on all genome-wide markers.

The predictive ability of both approaches was estimated asPearson』s correlation coefficientbetween the predicted and observed trait values of 20% of the lines, with the prediction being based on effect estimatesfrom the remaining 80%of the lines. Resampling was repeated 1000 times.

All calculations were made with the open-source programming language and statistical software R (R Core Team 2013) and the statistical software packageASReml-R 3.0to solve the mixed models.

結果：

對於表型數據，遺傳力算是FHB當中非常高的了，應該是由於大部分都是感病。

對於GWAS結果，感覺並不是很好，檢測到的幾個抗病位點都只是單一marker，而沒有出現helicopter的多markers的結果，說明本文的marker密度還是不夠的。

最後，比較MAS和GS兩種方法的predictive ability。It must be noted here that the predictive ability of MAS also exploits relatedness between lines and not only QTL effects (Gowda et al., 2014)。

Boxplots for the comparison of marker-assisted selection using the same four QTL and genomic prediction (GP) using all markers. Genotypes with the tall Rht-B1a allele were excluded from both analyses

討論

Marker-assisted selection (MAS) based on the four strongest QTL yielded a predictive ability of 0.65, while genomic selection (GS) taking into account all genome-wide markers only marginally improved the predictive ability to 0.70.對於本文用到的這個群體來說，如果marker的密度在加大一些的話，相信MAS和GS的predition要能增加很多！

The predictive ability of phenotypic selection, being roughly estimated as the square root of the heritability, reached 0.92 and was thus by far the highest value in this comparison.

Phenotypic selection certainly takes longer than genomic selection, but on the other hand, population sizes are also not so limited as reliablehigh-throughput FHB phenotyping platformshave been developed in breeding companies (E. Ebmeyer, personal communication).

Furthermore, recurrent selection can be acceleratedusing early generations for FHB evaluation, such as F1:2 bulks after recombination of the best entries (Miedaner et al., 2009).

Nevertheless, the rather high predictive ability of genomic prediction suggests that this approach might be a valuable genomic tool to assist FHB resistance breeding if the linesare already genotyped with markers,for example, to utilize genomic prediction for yield or quality traits.

喜歡這篇文章嗎？立刻分享出去讓更多人知道吧！