Skip to content


entry signage


大体而言,一个科学理论需要做出”惊奇“而可以验证(证伪)的预测,并通过实验的检验。这是科学哲学的基本素养。”fit evidence in“是一种逻辑诡辩,暂时不予讨论。


顺便说一下,DeepL把”Genetic Entropy“ is BS中的”BS“翻译为胡说八道(大概它以为这是Bullshit的缩写),我是不赞成的。BS = Biblical Science,不接受反驳。

“Genetic Entropy” is BS: A Summary

基因熵 “是一种胡说八道:摘要



The idea of “genetic entropy” is one of a very few “scientific” ideas to come from creationists. It’s the idea that humanity must be very young because harmful mutations are accumulating at a rate that will ultimately lead to our extinction, and so we, as a species, can’t be any older than a few thousand years. Therefore, creation. John Sanford proposed and tried to support this concept in his book “Genetic Entropy & The Mystery of the Genome,” which is…wow it’s bad. EDIT: If you want to read “Genetic Entropy,” you can find it here (pdf). It’s a quick read, and probably worth the time if you want to be familiar with the argument. Might as well get it from the source. “基因熵”的概念是来自创造论者的极少数 “科学”概念之一。它认为人类必定非常年轻,因为有害突变的积累速度最终会导致我们的灭绝,所以我们作为一个物种,不可能比几千年前更老。因此,创造科学家约翰-桑福德(John Sanford)在他的《退化论:基因熵与基因组之谜》(Genetic Entropy & The Mystery of the Genome)一书中提出并试图支持这一概念,这本书是……哇,真糟糕。编辑:如果你想读《退化论》一书,你可以在这里找到它(pdf)。这本书很通俗易懂,如果你想熟悉这个论点,花时间浏览一下,了解来龙去脉还是值得的。

Everything about the genetic entropy argument is wrong, including the term itself. But it comes up over and over and over, including here, repeatedly, I think because it’s one of the few sciencey-sounding creationist arguments out there. So join me as we quickly cover each reason why “genetic entropy” is BS. 关于基因熵论的一切都是错的,包括这个词本身在内。但是它却一次又一次地被人提及,包括在这里也反复出现。我想,这是因为它是少数几个听起来很科学的创造论者“原创“论点之一。所以,请和我一起,快速地了解一下”基因熵 “错误的原因。

I’m going to do this in two parts. First we’ll have a bunch of quick points, and after, I’ll elaborate on the ones that merit a longer explanation. Each point will be labeled “P1”, “P2”, etc., as will each longer explanation. So if you want to find the long version, just control-f the P# for that point. 我将分两部分进行。首先,我们将有快速给出一组论点,之后,我将对那些值得仔细解释的观点进行阐述。每个要点将被标记为 “P1″、”P2 “等,每个较长的解释也是如此。因此,如果你想找到长的版本,只要用control-f调出搜索,并输入编号P#即可。

P1: “Genetic entropy” is a made-up term invented by creationists to describe a concept that already existed: Error catastrophe. Even before it’s a vaguely scientific idea, the term “genetic entropy” is an attempt at branding, to make a process seem more dangerous or inevitable through changing the name. I’m going to use the term “error catastrophe” from here on when we’re talking about the actual population genetics phenomenon, and “genetic entropy” when talking about the silly creationist idea.
P1: “基因熵 “是创造论者捏造的术语,用来描述一个已经存在的概念——错误灾难(Error catastrophe)。甚至在它成为一个模糊的科学概念之前,”基因熵 “这个术语就是一种宣称手段,使徒通过改变名称,使一个过程看起来更加危险或不可避免。从现在开始,当我们谈论实际的群体遗传学现象时,我将使用 “错误灾难 “一词,而当谈论愚蠢的创造论思想时,我将使用 “基因熵”。

P2: Error catastrophe has never been observed or documented in nature or experimentally. In order to conclusively demonstrate error catastrophe, you must show these two things: That harmful mutations accumulate in a population over generations, and that these mutations cause a terminal decline in fitness, meaning that they cause the average reproductive output to fall below 1, meaning the population is shrinking, and will ultimately go extinct.

This has never been demonstrated. There have been attempts to induce error catastrophe experimentally, and Sanford claims that H1N1 experienced error catastrophe during the 20th century, but all of these attempts have been unsuccessful and Sanford is wrong about H1N1 in every way possible.


P3: The process through which genetic entropy supposedly occur is inherently contradictory. Either neutral mutations are not selected against and therefore accumulate, or harmful mutations are selected against, and therefore don’t accumulate. Mutations cannot simultaneously hurt fitness and not be selected against.
P3: 所谓的基因熵发生过程本质上是矛盾的。要么选择不排除中性突变,因此中性突变累积;要么选择排斥有害突变,因此有害突变不累积。突变不可能同时损害适应性,又不被选择所排除。

P4: As deleterious mutations build up, the percentage of possible subsequent mutations that are harmful decreases, and the percentage of possible beneficial mutations increases. The simplest illustration is to look at a single site. Say a C mutates to a T and that this is harmful. Well now that harmful C–>T mutation is off the table, and a new beneficial T–>C mutation is possible. So over time, as harmful mutations accumulate, beneficial mutations become more likely.

P5: (Somewhat related to P4) A higher mutation rate provides more chances to find beneficial mutations, so even though more harmful mutations will occur, they are more likely to be selected out by novel beneficial genotypes that are found and selected for. This is slightly different from P4, which was about the proportion of mutations; this is just raw numbers. More mutations means more beneficial mutations.

P6: Sanford is dishonest. His work surrounding “genetic entropy” is riddled with glaring inaccuracies that are either deliberate misrepresentations, or the result of such egregious ignorance that it qualifies as dishonesty.

Two of the most glaring examples are his misrepresentation of a distribution of fitness effects produced by Motoo Kimura, and his portrayal of H1N1 fitness over time.
P6:桑福德是个不诚实的人。他围绕 “基因熵 “所做的工作遍布明显的误导和不准确。这些不准确之处要么是故意歪曲,要么是由于极度无知而导致的不诚实行为。

其中最明显的两个例子是他对Motoo Kimura提出的适应效果分布的歪曲,以及他对H1N1适应效果随时间变化的错误描述。

Below this point you’ll find more details for some of the above points.

P2: Error catastrophe has never been observed, experimentally nor in nature. There have been a number of attempts at inducing error catastrophe experimentally, but none have been successful. Some work from Crotty et al. is notable in that they claimed to have induced error catastrophe, but actually only maybe documented lethal mutagenesis, a broader term that refers to any situation in which a large number of mutations cause death or extinction. Their single round of mutagenic treatment of infectious genomes necessarily could not involve mutation accumulation over generations, and so while mutations my have caused the fitness decline, it isn’t wasn’t through error catastrophe. It’s also possible the observed fitness costs were due to something else entirely, since the mutagen they used has many effects.

J.J. Bull and his team have also worked extensively on this question, and outline their work and the associated challenges here. In short, they were not able to demonstrate terminal fitness decline due to mutation accumulation over generations, and in one series of experiments actually observed fitness gains during mutagenic treatment of bacteriophages.

P2:无论是在实验中还是在自然界中,都没有观察到错误灾难的发生。在实验中曾有过一些诱导错误灾难的尝试,但都没有成功。Crotty等人的工作值得注意——他们声称诱发了错误灾难,但实际上可能只是记录了致命的诱变(lethal mutagenesis)。这是一个更广泛的术语,指的是大量突变导致死亡或灭绝的情况。他们对感染性基因组的单轮诱变处理不可能涉及历代的突变积累,因此,虽然突变可能造成了适应度下降,但这并不是错误灾难的效果。他们观察到的适应代价完全可能由于其他原因造成的,因为他们使用的诱变剂有许多副作用

J.J. Bull和他的团队也对这个问题进行了广泛的研究,并在此概述了他们的工作和相关挑战。简而言之,他们无法证明突变的世代积累会导致的最终适应度下降,甚至在一系列的实验中,实际上观察到对噬菌体进行诱变处理期间适应度的增加

You’ll notice that all of that work involves bacteriophages and mutagenic treatment. What about humans? Well, phages are the ideal targets for lethal mutagenesis, especially RNA and single-stranded DNA (ssDNA) phages. These organisms have mutation and substitution rates orders of magnitude higher than double-stranded DNA viruses and cellular organisms (pdf). They also have small, dense genome, meaning that there are very few intergenic regions, most of which contain regulatory elements, and even some of the reading frames are overlapping and offset, which means there are regions with no wobble sites.

This means that deleterious mutations should be a higher percentage of the mutation spectrum compared to, say, the human genome. So mutations happening faster plus more likely to be harmful equals ideal targets for error catastrophe.
你会注意到,所有这些工作都涉及噬菌体和诱变处理。那人类呢?好吧,噬菌体是致命诱变的理想目标,特别是RNA和单链DNA(ssDNA)噬菌体。这些生物体的突变和替代率比双链DNA病毒和细胞生物体高几个数量级(pdf)。它们的基因组也小而密集,这意味着基因间隔区域非常少,其中大部分包含调控元素,甚至有些阅读框(reading frame)是重叠和偏移的,这意味着有些区域没有摇摆位点(wobble sites)。

这意味着,与人类基因组相比,噬菌体de 有害突变在突变谱系中的比例应该更高。因此,噬菌体的突变发生得更快,更可能是有害突变——错误灾难的理想目标。

In contrast, the human genome is only about 10% functional (<2% exons, 1% regulatory, some RNA genes, a few percent structural and spacers; stuff with documented functions adds up to a bit south of 10%). It’s possible up to 15% or so has a selected function, but given what we know about the rest, any more than that is very unlikely. So the percentage of possible mutations that are harmful is far lower in the human genome compared to the viral genomes. And we have lower mutation and substitution rates.

All of that just means we’re very unlikely to experience error catastrophe, while the viruses are the ideal candidates. And if the viruses aren’t susceptible to it, then the human genome sure as hell isn’t.

But what of H1N1? Isn’t that a documented case of error catastrophe. That’s what Sanford claims, after all.

Except yeah wow that H1N1 paper is terrible. Like, it’s my favorite bad paper, because they manage to get everything wrong. Here’s a short list of the errors the authors commit:

They ignored neutral mutations.他们忽视了中性变异。

They claimed H1N1 went extinct. It didn’t. Strains cycle in frequency. It’s called strain replacement.他们宣称H1N1灭绝了。实际上不是这么回事。菌种循环的频率很高。这就是所谓的菌种替换(strain replacement)。

They conflated intra- and inter-host selection, and in doing so categorize a bunch of mutations as harmful when they were probably adaptive. 他们把宿主内部和宿主之间的选择混为一谈,于是把一堆突变都归为有害,而这些突变可能是增加适应性的。

They treated codon bias as a strong indicator of fitness. It isn’t. 他们把密码子偏向(codon bias)作为适应性的有力指标。其实不然。

Translational selection (i.e. selection to match host codon preferences) doesn’t seem to do much in RNA viruses. 翻译选择(Translational selection。即选择与宿主密码子偏好匹配)对RNA病毒似乎没有什么影响

They ignored host-specific constraints based on immune response, specifically how mammals use CpG dinucleotides to recognize foreign DNA/RNA and trigger an immune response. In doing so, they categorized changes in codon bias as deleterious when they were almost certainly adaptive. 他们忽略了基于免疫反应的宿主特定限制,特别是哺乳动物如何使用CpG二核苷酸来识别外来DNA/RNA并触发免疫反应的机制。于是,他们把密码子偏向的变化归为有害的,实际上这些变化几乎肯定是适应性的。

They conflated virulence (how sick a virus makes you) with fitness (viral reproductive success). Not the same thing. And sometimes inversely correlated. 他们把毒性(virulence,病毒使你生病的程度)和适应性(病毒的繁殖成功率)混为一谈。这不是一回事。而且有时还成反比关系。

Related, in using virulence as a proxy for fitness, they ignored the major advances in medicine from 1918 to the 2000s, including the introduction of antibiotics, which is kind of a big deal, since back then and still today, most serious influenza cases and deaths are due to secondary pneumonia infections.

So no, we’ve never documented an instance of error catastrophe. Not in the lab. Not in H1N1.


P3: “Genetic entropy” supposedly works like this: Mutations that are only a little bit harmful (dubbed “very slightly deleterious mutations” or VSDMs) occur, and because they are only a teensy bit bad, they cannot be selected out of the population. So they accumulate, and at some point, they build up to the point where they are harmful, and at that point it’s too late; everybody is burdened by the harmful mutations, has low fitness, and the population ultimately goes extinct.
P3:”基因熵 “的工作机制据说是这样的。只有一点点害处的突变(被称为 “非常轻微的有害突变 “或VSDMs)出现了,由于它们只有一点点害处,所以无法从群体中选择出来。因此,它们可以不断积累,到了某一时刻,积累到有害的地步,到那时一切都晚了;每个人都被有害的突变所累,适应性低下,人类最终灭绝了。

Here are all of the options for how this doesn’t work.

One, you could have a bunch of neutral mutations. Neutral because they have no effect on reproductive output. That’s what neutral means. They accumulate, but there are no fitness effects. So the population doesn’t go extinct – no error catastrophe.



Or you could have a bunch of harmful mutations. Individually, each with have a small effect on fitness. Individuals who by chance have these mutations have lower fitness, meaning these mutations experience negative selection. Maybe they are selected out of the population. Maybe they persist at low frequency. Either way, the population doesn’t go extinct, since there are always more fit individuals (who don’t have any of the bad mutations) present to outcompete those who do. So no error catastrophe.

Or, option three, everyone experiences a bunch of mutations all at once. All in one generation, every member of a population gets slammed with a bunch of harmful mutations, and fitness declines precipitously. The average reproductive output falls below 1, and the population goes extinct. This is also not error catastrophe. Error catastrophe requires mutations to accumulate over generations. This all happened in a single generation. It’s lethal mutagenesis, a broader process in which a bunch of mutations cause death or extinction, but it isn’t the more specific error catastrophe.

But we can do a better job making the creationist case for them. Here’s the strongest version of this argument that creationists can make. It’s not that the mutations are neutral, having no fitness effect, and then at some threshold become harmful, and now cause a fitness decline population-wide. It’s that they are neutral alone, but together, they experience epistasis, which just means that two or more mutations interact to have an effect that is different from any of them alone.


So you can’t select out individual mutations (since they’re neutral), which accumulate in every member of the population over many generations. But subsequent mutations interact (that’s the epistasis), reducing fitness across the board. 因此,你不能选择出个别的突变(因为它们是中性的),只能任凭这些突变在种群的每个成员中经过许多代积累。但随后的突变会相互影响(这就是外显性),全面降低适应性。

But that still doesn’t work. It just pushed back the threshold for when selection happens. Instead of having some optimal baseline that can tolerate a bunch of mutations, we have a much more fragile baseline, wherein any one of a number of mutations causes a fitness decline. 但这仍然不起作用。它只是推后了选择发生的选择门槛。我们不再有一些可以容忍大量突变的最优基线,而是有一个更脆弱的基线,其中任何一个突变都会导致适应度下降。

But as soon as that happens in an individual, those mutations are selected against (because they hurt fitness due to the epistatic effects). So like above, you’d need everyone to get hit all in a single generation. And a one-generation fitness decline isn’t error catastrophe.

So even the best version of this argument fails.



P4 and P5: I’m going to cover these together, since they’re pretty similar and generally work the same way.


Basically, when you have bunch of mutations, two things operate that make error catastrophe less likely than you would expect.


First, the distribution of fitness effects changes as mutations occur. When a deleterious mutation occurs, at least one deleterious mutation (the one that just occurred) is removed from the universe of possible deleterious mutations, and at least one beneficial mutation is added (the back mutation). But there are also additional beneficial mutations that may be possible now, but weren’t before, due to epistasis with that new harmful mutation. These can recover the fitness cost of that mutation, or even work together with it to recover fitness above the initial baseline. These types of mutations are called compensatory mutations, and while Sanford discusses epistasis causing harmful mutations to stack, he does not adequately weigh the effects in the other direction, as I’ve described here.


Related, when you have a ton of mutations, you’re just more likely to find the good ones. We actually have evidence that a number of organisms have been selected to maintain higher-than-expected mutations rates, probably due to the advantage this provides. My favorite example is a ssDNA bacteriophage called phiX174. It infects E. coli, but lacks the “check me” sequences that its host uses to correct errors in its own genome. By artificially inserting those sequences into the phage genome, its mutation rate can be substantially decreased. Available evidence says that selection maintains the higher mutation rate. We also see that during mutagenic treatment, viruses can actually become more fit, contrary to expectations.

与此相关,当你有许多的突变时,更有可能找到好的突变。实际上有证据表明,一些生物体刻意选择维持高于预期的突变率,可能是因为高突变提供了优势。我最喜欢的例子是一种叫做phiX174的ssDNA噬菌体。它感染大肠杆菌,但缺乏其宿主用来纠正自身基因组错误的 “自检”序列。通过人为地将这些序列插入噬菌体的基因组中,它的突变率可以大大降低。现有证据表明,这种噬菌体有意选择维持了较高的突变率。我们还看到,与预期相反,在诱变处理过程中,病毒实际上可以变得更适应

So as mutations occur, beneficial mutations become more likely, and more beneficial mutations will be found. Both processes undercut the notion of “genetic entropy”. 因此,随着突变的发生,有益的突变变得更有可能,更多的有益突变将被发现。这两个过程都削弱了 “基因熵 “的概念。

P6: John Sanford is a liar. There’s really isn’t a diplomatic way to say it. He’s a dishonest hack who misrepresents ideas and data. I’ve covered this before, but I’ll do it again here, for completeness.


I’m only going to cover one particularly egregious example here, but see here for another I’m going to stick to the use of a distribution of mutation fitness effects from Motoo Kimura’s work, which Sanford modifies in “Genetic Entropy,” and uses to argue that beneficial mutations are too rare to undo the inevitable buildup of harmful mutations. 我在这里只讲一个特别恶劣的例子,另一个例子请看这里。我这里只讨论Motoo Kimura给出的突变适应效应分布。桑福德在《基因熵》一书中对其进行了修改,以支持有益突变太少,无法消除有害突变造成不可避免累积效应的说法。

Now first, Sanford claims to show a “corrected” distribution, since Kimura omitted beneficial mutations entirely from his. Except this “corrected” distribution is based on nothing. No data. No experiments. Nothing. It’s literally “I think this looks about right”. Ta-da! “Corrected”. Sure.

首先,桑福德声称自己展示了一个 “修正的”分布,因为木村在他的分布中完全省略了有益突变。只是这个”修正的”分布没有任何依据。没有数据。没有实验。什么都没有。这简直就是 “我认为这看起来差不多”。Ta-da! “归正了”……当然。

Second, Sanford justifies his distribution by claiming that Kimura omitted beneficial mutations because he knew they are so rare they don’t really matter anyway. He wrote:


In Kimura’s figure, he does not show any mutations to the right of zero – i.e. there are zero beneficial mutations shown. He obviously considered beneficial mutations so rare as to be outside of consideration.


Kimura’s rationale was the exact opposite of this. His distribution represents the parameters for a model demonstrating genetic drift (random changes in allele frequency). He wrote:


The situation becomes quite different if slightly advantageous mutations occur at a constant rate independent of environmental conditions. In this case, the evolutionary rate can become enormously higher in a species with a very large population size than in a species with a small population size, contrary to the observed pattern of evolution at the molecular level.


In other words, if you include beneficial mutations, they are selected for and take over the simulation, completely obscuring the role genetic drift plays. So because they occur too frequently and have too great an effect, they were omitted from consideration. 换句话说,如果把有益突变考虑在内,它们会被高频选择,破坏实验条件,完全掩盖遗传漂移所起的作用。因此,由于它们出现的频率太高,影响太大,所以只能省略,不予考虑。

Okay, let’s give Sanford the benefit of the doubt on the first go. Maybe, despite writing a book that leans heavily on Kimura’s work, and using one of Kimura’s figures, Sanford never actually read Kimura’s work, and honestly didn’t realize hat Kimura’s rationale was the exact opposite of what Sanford claims. Seems improbable, but let’s say it was an honest mistake.


The above passage (and the broader context) were specifically pointed out to Sanford, but he persisted in his claim that he was accurately representing Kimura’s work. He wrote:


Kimura himself, were he alive, would gladly attest to the fact that beneficial mutations are the rarest type。 木村本人如果还活着,会很高兴地承认有益突变是罕见的。

The interesting thing with that line is that it’s a slight hedge compared to the earlier statement. This indicates two things. First, that Sanford knows he’s wrong about Kimura’s rationale, and second, that he wants to continue to portray Kimura as agreeing with him, even though he clearly knows better.


There’s more in the link at the top of this section, but this is sufficient to establish that Sanford is a liar. 本节顶部的链接中还有更多内容,但我们这点讨论已经足以证明桑福德是个骗子。

So that’s…I won’t say everything, because this is a deep well, but that’s a reasonable rundown of why nobody should take “genetic entropy” seriously. 所以这就是……好吧,我不说了,因为水太深了——这就是为什么你不应该认真对待 “基因熵”的简单理由。

Creationists, if you want to beat the genetic entropy drum, you need to deal with each one of these points. (Okay maybe not P6, unless you want to defend Sanford.) So if and when you respond, specifically state which point you dispute and why. Be specific. Cite evidence.



版权所有:Eddy Zhang

This image has an empty alt attribute; its file name is kuawen-640x334.jpg

This blog is readers’ supprting. If you are ever moved to support this ministry or my family…

这个博客是由普通读者支持的。若您或您的教会愿意支持跨文翻译的事工,请使用 或以下二维码。