Our new JABG executive editor asked me to write an editorial on current and future needs for research in our area, from an industry point of view . This made me remember that I was involved into something similar 10 years ago. So, it is time for an update, and for some reflection—always interesting.

FABRE‐TP is a European Technology Platform dedicated to Farm Animal Breeding & REproduction. Such ETPs are industry‐led forums that develop R&D agendas. Those agendas are used at the EU and national levels to get support for funding (see tinyurl.com/u4wvxk9); that is, a think tank—just what we need for needs for research in our area .

In 2011, FABRE‐TP published its Strategic Research Agenda (SRA) (see tinyurl.com/wkqynrt), a 48‐page document that was put together out of input from 55 members of 13 expert panels, each panel focusing on themes such as food quality or robustness, on animal species from honeybees to horses or on technologies such as genetics or genomics.

For JABG readers, the most relevant part is in those technology chapters and the overlapping bits of the theme and species chapters. Time frames of five, 15 and 25 years were set for research needs, and that was 10 years ago—so let's see how far we have come with the 5‐year one and what progress has been made towards the 15‐year one; and if there is anything new, by now. I have omitted anything that is about policy rather than research—and filtered according to my own interests and prejudices.

In another dimension, the entries for research needs cover traits , methodology and issues .

Traits . Perhaps surprisingly, only one conglomerate of traits really features in the SRA: robustness and resilience , with a few extensions to animal behaviour and welfare traits. Practically every chapter lists it as a short‐term priority and often also as a medium‐term one. So, the species‐oriented people seem to be more or less happy with their current ability to influence the regular production and reproduction traits, and that is a very good thing in itself.

A few decades ago, we used to refer to these robustness traits as “secondary traits”—and by now, they have evolved to hot item #1 in livestock breeding: very clearly a main (and wide open!) field for research to focus on; and since then, we have seen an increasing interest in behavioural traits too.

Methodology . Of course, the “current ability to influence the regular traits” of above has much to do with the worldwide move to genomic prediction that was just past its implementation infancy in 2010 (we were all busy blending polygenic and genomic EBVs at that time: Single Step was still to come). So, again not surprisingly, the second hot item in the SRA centres around genomic technology, with entries ranging from “optimal implementation of genomic data into genetic evaluations” (from the Cattle panel, with much focus on across‐breed evaluation and data systems that were then too large for Single Step) to “fully annotated quality genome sequences with massive discovery of variants” which is still on everybody's wish list, and increasingly so.

We all know that GBLUP needs proper training of its markers more than anything else, so the other SRA's hot methodology is of course phenotyping . This is very much a matter of kitting out the proper equipment to measure things: that is not for JABG. But once the Big Data on milk, body weight, feed intake, activity, body temperature and whatever else is being streamed into our databases, we will need some very robust methodology to (a) handle and manage it and (b) convert it to meaningful information for genetic evaluation, for example by relating the volatility of within‐animal records over time to the resilience traits that form the hot issue of the near future—as we saw above. The also‐hot animal behaviour traits of above would benefit from the same.

And then there is the concept of what we used to call “biomarkers” a few decades ago, now known as metabolomics . There are chips available now that produce readings of a few hundred metabolite concentrations in the drop of blood applied to them. Here, we really need a good evaluation of the equivalent of the candidate gene approach versus anonymous DNA markers: Do we maximize the relevant information out of all these data by understanding what each metabolite actually does in the body, or is the black box a more hopeful approach?

Both these methods have to do with the accuracy of breeding value estimation. The next important part of the breeder's equation is the generation interval: always more a matter of proper logistics than of science. But the SRA mentions “schemes incorporating large‐scale genotyping at the embryo level” as a medium‐term priority, bringing us back to the concepts of velo‐ and whizzogenetics of the 1990s: apply GBLUP to extremely young animals, select them, and propagate them artificially. Later, the concept of Iterated Embryo Selection came up, which would involve stem cells for the same purpose. Question 1 would then be why those schemes never made it to implementation in livestock breeding practice, leading to question 2: And what could we do about that? For the JABG community specifically, how should such large‐scale genotyping schemes be organized, so as to maximize the efficiency of the breeding programmes they feature in?

The Genetics panel of the SRA has an entry “metagenomic sequencing of microbial communities, e.g., in the gastrointestinal tract,” but none of the other panels seem to have been aware of the microbiome and how it influences practically everything, as was widely discovered a few years later. I would suggest that for the JABG community, the main focal issue should now be to evaluate the microbiome's composition as a novel breeding goal trait versus a novel element of the mixed model equations in MGBLUP, and in the latter case, to design the next‐generation Single Step approach.

Another noticeably absent methodology in this 2010 text is genome editing , of course. But let's leave that to other journals, for now.

Issues . As long as animal breeding has been seriously practiced, we have been worrying (usually very silently) about the selection‐induced erosion of the genetic variation that we all make a living off: the third element of the breeder's equation. The SRA has the same worries, particularly in relation to the genomic prediction methodology of above that it is so enthusiastic about at the same time. The livestock breeding literature holds a large handful of studies that show no erosion, and a very few others that show the opposite. But the worry remains, so this is a fertile ground for research, particularly in this era of whole‐genome sequencing. For example, can we quantify Eitan & Soller's selection‐induced genetic variation by now and get it under control? More proactively, is there a safe way to increase recombination rates and introduce some negative erosion?

The same holds for genetic antagonisms ; this is one of my personal hobby horses and the SRA is repeatedly worried about them too. Unfavourable genetic correlations between traits make it hard to deliver the breeding goal, and this may well be one of the most important limitations to all genetic improvement in livestock. The point is to align G (the genetic covariance matrix: what the population is capable of) with b (the index weighting factors: what we want the population to do). Walsh & Lynch (2018) posed the terrifying question: Is there genetic variation in the direction of selection? Some populations will have a G eigenstructure that fits b more easily than other populations do. This should be a crucial element in the design of the breeding goal for a population or, more interestingly, the other way around—especially for crossbreeding systems where several lines with their variable strengths and weaknesses can compensate each other. It assumes availability of many lines, as in poultry breeding and increasingly so in pig breeding. There is lots of fundamental research to do here.

And the final part: the SRA mentions the need for animal geneticists to collaborate much more intensively with animal nutritionists . Even genetically improved livestock does not generate animal protein out of thin air. Nutrition must remain in sync with ΔG; otherwise, genetic improvement is commercially futile because the added value will not be expressed. Multidisciplinary research is always highly praised, but it is very rarely practised. We need some significant groundbreaking here. Just the same towards epidemiology, of course.

Enough about the FABRE‐TP SRA. In his Thank you! editorial of late 2019, Asko Mäki‐Tanila, our former executive editor, wrote: animal breeding research has advanced in steps of some 10‐20 years apart: BLUP, REML, MOET and genomic EBV. It is now almost 20 years from the latter [i.e., Meuwissen et al., 2001] and time for another groundbreaking article . Hein van der Steen and I argued something similar during an FAO meeting in Rome in 2009 (tinyurl.com/wblaoxs, page 6).

See the figure below: looking back in time, the hot methodology for livestock breeding value estimation seems to alternate between (a) biology‐driven techniques (i.e., one has to understand how the animal works, or how relationships among relatives work) and (b) statistics‐driven techniques, mostly of the black‐box variety that amateur breeders all over the world are so sceptical about.

And we seem to be crossing a new watershed: as Hein and I forecasted quite accidentally, now in 2020 the livestock breeding sector is in for a decade (or two) of what John Woolliams once dubbed quantomics and what comes back to the “fully annotated quality genome sequences with massive discovery of variants” of above. So, the upcoming challenge for the JABG community will be to make those discovered variants (particularly the causal ones) work more effectively than today's black box does; apply that to longitudinal Big Data phenotypes, focusing on robustness and resilience and behaviour traits, and your research should be in business. And then go and write that groundbreaking article for Asko.

我们新的JABG执行主编要求我从行业的角度撰写一篇有关我们领域当前和未来研究需求的社论。这使我想起十年前我也参与了类似的工作。因此，是时候进行更新并进行一些反思了-总是很有趣。

FABRE-TP是致力于畜禽养殖和繁殖的欧洲技术平台。这样的ETP是由行业主导的论坛，用于制定研发议程。这些议程在欧盟和国家一级用于获得资金支持（请参见tinyurl.com/u4wvxk9）；也就是一个智囊团-正是我们在本地区研究需求所需要的。

2011年，FABRE-TP发布了其战略研究议程（SRA）（请参见tinyurl.com/wkqynrt），这份长达48页的文件汇集了13个专家小组的55名成员的意见，每个小组的主题都涉及从蜜蜂到马的动物物种或遗传学或基因组学等技术的食品质量或坚固性。

对于JABG读者而言，最相关的部分是这些技术章节以及主题和物种章节的重叠部分。为研究需要设定了五年，十五年和二十五年的时间框架，那是十年前的事。因此，让我们看看五年五年来有多远，朝十五年取得了什么进展；如果有什么新鲜的东西，到现在。我忽略了所有与政策有关的内容，而不是与研究有关的内容，而是根据自己的兴趣和偏见进行了过滤。

在另一方面，研究需求的条目包括特征，方法论和问题。

特性。也许令人惊讶的是，SRA中只有一个特征集团真正具有特色：健壮性和韧性，对动物行为和福利性状有一些扩展。实际上，每一章都将其列为短期优先事项，通常也将其列为中期优先事项。因此，以物种为导向的人们似乎对他们目前影响常规生产和繁殖特征的能力或多或少感到满意，这本身就是一件好事。

几十年前，我们曾经将这些鲁棒性特征称为“次要特征”，到现在，它们已发展成为畜牧业中的热门项目＃1：很明显，这是一个主要的（广泛开放的！）研究领域上; 从那时起，我们也对行为特征越来越感兴趣。

方法论。当然，上述的“当前影响常规性状的能力”与2010年全球实现基因组预测刚刚起步阶段时有很大关系（当时我们都在忙于将多基因和基因组EBV融合在一起：步骤仍要来）。因此，同样毫不奇怪，SRA中的第二个热门话题围绕基因组技术，条目范围从“将基因组数据的最佳实现方式转化为遗传评估”（来自Cattle小组，重点是跨品种评估和数据系统，对于单步执行而言，它们太大了，无法“完全注释高质量的基因组序列，并大量发现了变体”，这一点仍然在每个人的心愿单上，而且越来越多。

我们都知道，GBLUP比其他任何事情都需要对其标记进行适当的培训，因此，另一个SRA的热门方法当然是表型分析。这很大程度上是要配备合适的设备来进行测量：这不是JABG的问题。但是，一旦将有关牛奶，体重，采食量，活动，体温以及其他任何方面的大数据传输到我们的数据库中，我们将需要一些非常强大的方法来（a）处理和管理它，并将（b）转换为如上所见，例如，通过将动物内记录的随时间变化的波动性与构成不久的将来的热点问题的弹性特征相关联，可以为遗传评估提供有意义的信息。上面的动物行为特征也将从中受益。

然后是几十年前我们称为“生物标志物”的概念，现在称为代谢组学。现在有可用的芯片，这些芯片可在施加到它们的血液中产生数百种代谢物浓度的读数。在这里，我们确实需要对候选基因方法与匿名DNA标记的等效性进行良好的评估：我们是通过了解每种代谢物在体内的实际作用来最大化所有这些数据中的相关信息，还是黑匣子？充满希望的方法？

这两种方法都与繁殖价值估计的准确性有关。育种方程式的下一个重要部分是世代间隔：总是更多地是适当的后勤问题而不是科学问题。但是SRA提到“将在胚胎水平整合大规模基因分型的方案”作为中期优先事项，这使我们重新回到了速度和毛病遗传学的概念1990年代：将GBLUP应用于年幼的动物，选择它们，然后人工繁殖。后来，出现了“迭代胚胎选择”的概念，该概念涉及干细胞以实现相同的目的。那么，问题1就是为什么那些计划从未在牲畜育种实践中付诸实施，导致问题2：我们对此可以做些什么？特别是对于JABG社区，应如何组织这种大规模的基因分型方案，以最大程度地发挥其繁殖计划的效率？

SRA的遗传学小组有一个条目“微生物群落的基因组测序，例如在胃肠道中”，但是其他小组似乎都没有意识到微生物组及其对几乎所有事物的影响，这一点被广泛发现。几年后。我建议，对于JABG社区而言，现在的主要重点应该是评估微生物组的组成，将其作为新的育种目标特征与MGBLUP中混合模型方程式的新元素进行比较，在后一种情况下，设计下一个-生成单步方法。

当然，在2010年这本书中，另一个明显缺乏的方法是基因组编辑。但是，现在暂时将其留给其他期刊。

问题。只要认真地进行动物育种，我们就一直（通常非常沉默地）担心选择诱导的遗传变异侵蚀我们都以谋生为生：育种方程的第三个元素。SRA也有同样的担忧，特别是对于上述的基因组预测方法而言，它充满了热情。牲畜育种文献中有大量研究没有显示出侵蚀，而其他研究则相反。但是这种担忧仍然存在，因此这是进行研究的沃土，尤其是在这个全基因组测序时代。例如，我们现在能否量化Eitan＆Soller的选择诱发的遗传变异并使其得到控制？更积极地，是否有一种安全的方法来提高重组率并引入一些负面侵蚀？

这同样适用于遗传对立; 这是我个人的爱好之一，SRA也一再担心它们。性状之间不利的遗传相关性使其难以实现育种目标，这很可能是牲畜所有遗传改良的最重要限制之一。关键是使G（遗传协方差矩阵：总体能力）与b（指数加权因子：我们希望总体做的事情）对齐。Walsh＆Lynch（2018）提出了一个可怕的问题：选择方向是否存在遗传变异？一些人口将有一个摹特征结构适合b比其他人群更容易做到。这应该是设计种群育种目标的关键要素，或者更有趣的是，这是相反的方式-特别是对于杂交系统，其中具有优势和劣势的几条线可以相互补偿。它假定有许多品系可用，如家禽育种，而猪育种也越来越多。这里有很多基础研究要做。

最后一部分：SRA提到了动物遗传学家需要与动物营养学家更深入地合作。即使经过基因改良的牲畜也不会凭空产生动物蛋白。营养必须与ΔG保持同步；否则，遗传改良在商业上是徒劳的，因为不会表现出增加的价值。跨学科研究始终受到高度赞扬，但很少实践。我们在这里需要一些重大的突破。当然，流行病学也是如此。

FABRE-TP SRA足够了。在他的谢谢！我们前执行主编AskoMäki-Tanila在2019年末的社论中写道：动物育种研究相距约10-20年，包括BLUP，REML，MOET和基因组EBV的进展。距后者已经近20年了（即Meuwissen等，2001），并且是另一篇开创性文章的时间。我和海因·凡德·史汀（Hein van der Steen）在2009年于罗马举行的粮农组织会议上提出了类似的观点（tinyurl.com/wblaoxs，第6页）。

参见下图：回顾过去，估算牲畜育种价值的热门方法似乎在（a）生物驱动技术（即必须了解动物的工作方式或亲戚之间的关系如何工作）与（ b）统计驱动的技术，大多数是全世界业余繁殖者对此持怀疑态度的黑盒品种。

而且，我们似乎正在跨越一个新的分水岭：正如海因和我非常偶然地预测的那样，现在到2020年，畜牧业的发展将是约翰·伍拉姆斯曾经称之为量子组学的十年（或两年），然后又回到了“完全注释的”领域。大量发现上述变异的高质量基因组序列。因此，对于JABG社区而言，即将到来的挑战将是使那些发现的变体（尤其是因果变体）比今天的黑匣子更有效地工作。将其应用于纵向大数据表型，重点放在健壮性，弹性和行为特征上，您的研究应该在商业中。然后去为Asko写开创性的文章。