剂量悲剧与线性缩放的致命误区
在建立这种物理与生物学联系前，我们必须理解缩放定律最惨痛的一个历史教训。1960年代，在中央情报局（CIA）臭名昭著的 MKUltra计划（MKUltra: 旨在通过药物改变人类行为的机密研究）研究背景下，科学家试图探究麦角酰二乙胺（LSD: 一种强烈的致幻剂）对大象行为的影响。当时的研究人员做出了一个极其草率的假设：药物剂量安全范围与生物体质量成正比（Linear scaling: 假设剂量随体重呈线性增长）。他们已知猫的安全剂量约为0.3毫克，而大象的体重是猫的1000倍，于是便直接注射了近300毫克的巨大剂量。结果，大象在注射后数分钟内便抽搐倒地并死亡。这一灾难性结果表明，大分子化学物质的代谢速度并不取决于单纯的物理质量，而是直接受基础代谢率（Basal Metabolic Rate: 维持生命所需的最低能量消耗速度）控制，而代谢率的增长相比于体重来说，是非线性的。
Original English
In the 1960s, the CIA was working on a top secret project known as MKUltra, and one of their main goals was to find out how drugs like LSD could be used to change human behavior. And this is where elephants come in because elephants are normally quite docile, but sometimes they just snap. And the hypothesis was that this change in behavior might be triggered by the release of an LSD-like substance that naturally occurs in their brains. So if that's true, then administering LSD to a docile elephant might reproduce that behavior. So the real question was how much LSD should you give an elephant so that the dose is large enough to cause a psychological reaction, but not so large that it causes harm? Well, the researchers didn't know, LSD had never been given to an animal that big, but they did know that the safe dose in cats was around 0.3 milligrams. Now, since an elephant has around a thousand times the mass of a cat, they figured we'll give it a thousand times the dose. They received approval to perform the experiment on Tusko, an Indian elephant at the Lincoln Park Zoo in Oklahoma. And there they injected him with nearly 300 milligrams of LSD. But within five minutes, Tusko trumpeted, collapsed, fell heavily onto his right side, defecated, and went into status epilepticus. They administered a few other drugs in an attempt to revive him, but Tusko died shortly thereafter. The mistake they made was to assume that safe drug dosage scales linearly with mass. It does not.代谢极限与表面积定律的修正
为了探寻更准确的缩放规律，物理学家通常会将复杂的生物体简化，并引入面积定律（Surface Law: 认为代谢率应与表面积成正比，即质量的2/3次方）。这是因为所有的能量消耗最终都会转化为体热并从皮肤表面散发出去。如果大象的代谢率真与体重呈1:1的线性关系，其内部产生的热量将达到猫的1000倍，而其散热表面积（随半径平方增长）仅增长100倍，这将导致大象在极短时间内因无法散热而将自己“煮熟”。如果按2/3次方的表面积定律来计算，大象的代谢率仅为猫的100倍，所需的LSD安全剂量也应只有30毫克左右。然而，到了1932年，瑞士生物学家马克斯·克莱伯（Max Kleiber）通过对各种哺乳动物的实测数据进行双对数坐标绘图，推翻了传统的2/3次幂模型，确立了克莱伯定律（Kleiber's Law: 发现生物代谢率与体重呈3/4次幂关系）。在该定律下，大象的代谢热量约是猫的178倍（约45,000千卡），对应的LSD致死剂量应为53毫克左右。
Original English
There's a lot going on in your body. Your heart is pumping, that takes energy. You're moving food around. Digestion takes energy. Keeping your brain going takes energy, breathing. Everything that you do requires energy. For a cat, for a single day, they require roughly 250 kilocalories of energy, used by all of their trillions of cells. But there is nothing special about a cat's cells. If you take a cat's cells and an elephant's cells and put them under a microscope, you'll find they're a similar size, similar makeup, and they perform the same sorts of functions. The same is true for other animals. In other words, the building blocks of animals are always roughly the same. So an elephant that has a thousand times the mass has about a thousand times as many cells. So you'd expect it would need a thousand times as much energy. So 250,000 kilocalories per day. But that is where you run into problems. So what's the issue here? If I'm an organism and I'm burning up energy all day long, that energy is radiated out in the form of body heat. So we're constantly losing heat through our surface, through our skin, to the environment. To see why this matters, let's simplify the problem and do the standard physicist thing. Let's assume our animals are perfect spheres. Since the volume of an animal is proportional to its mass, our 3000 kilogram elephant has a volume a thousand times greater than our three kilogram cat. So its radius must be 10 times larger. That's because volume is proportional to radius cubed. But surface area only grows as radius squared. So the elephant's surface area only increases by a factor of 100. Generating a thousand times as much heat while only having a hundred times the surface area to radiate it away would end very poorly for the elephant. If this were the case, it would boil alive. So in 1838, French scientists proposed a different scaling law. Since metabolism generates heat and that heat is radiated through the surface, metabolic rate or B, should scale in proportion to the surface area, A, instead. This became known as the surface law. Plugging that in for R, we find that metabolic rate should scale with mass to the two-thirds. According to this scaling, an elephant that's a thousand times as heavy as a cat should only burn a hundred times as many calories. So 25,000 instead of 250,000. And the appropriate dose of LSD for Tusko would've been just 30 milligrams. These two wildly different predictions come from different assumptions about how metabolic rate scales as a function of mass. But notice in both cases, it's just proportional to mass raised to some power. These kinds of relationships are called power laws. For nearly a hundred years, biologists generally agreed that the two-thirds exponent of the surface law was the correct one for metabolic rate. But then in 1932, Swiss biologist, Max Kleiber, decided to put it to the test. He plotted them against their mass on a log log plot. And as expected, all the data did fall on a straight line. But the slope wasn't two-thirds. Instead, it was about three quarters. This became known as Kleiber's Law. It implies that if you double an animal's mass, metabolic rate goes up by about 1.68, an increase of 68% instead of the 59%, which would be predicted by the two-thirds scaling law. So according to Kleiber's Law, an elephant burns roughly 178 times as many calories as a cat, or about 45,000 kilocalories. And the actual LSD dose Tusko should have received was 53 milligrams.分形网络与WBE理论的数学证明
为什么生命代谢会遵循以1/4为基数的幂律，而不是基于传统几何面积的1/3基数？这一谜题由物理学家杰弗里·韦斯特（Geoffrey West）、生物学家詹姆斯·布朗（James Brown）以及布莱恩·恩奎斯特（Brian Enquist）在1997年共同提出的 WBE理论（WBE Theory: 解释生物学中1/4幂律的资源输运分形模型）所解答。该理论基于三个简洁的假设：第一，输送资源的血管网络必须填满体内空间；第二，血管的最细终端（毛细血管）在所有哺乳动物中尺寸一致；第三，演化已经优化了该输送网络以最小化能量反射和输运损耗。在数学上，根据豪斯多夫维数（Hausdorff Dimension: 描述分形几何复杂度的分维数）的定义，三维空间中高度分形的自相似皱褶表面，其有效几何维度可以从2维向3维逼近。由于生命体内毛细血管网的分形分布，生物体交换物质的实际“有效表面积”不再与长度平方成正比，而是与长度立方（即体积）呈正比。由此推导出的系统特征长度与体重的1/4次方成正比，进而完美地在数学上推导出代谢率与体重的3/4次方成正比，解释了克莱伯定律。
Original English
West, Brown and Enquist teamed up to try and find a compelling explanation for Kleiber's Law. They started by assuming three simple premises. The first premise is that the networks that distribute resources are space-filling, since they need to reach every cell in the body. The second premise is that the terminal units of those networks, the thinnest segments on the outer periphery of the delivery system, have the same width regardless of the size of the organism. That is the outermost blood vessels that carry nutrients to an elephant's skin cells are about as thick as the ones in a mouse. The elephant just has many more of them. And the third premise is that over time, evolution has driven these transport networks toward an efficient design. But because fuel needs to reach every part of the body, you would also need many different paths. And as you go to larger and larger organisms, the networks inside need to reach a larger and larger volume. If instead we had just one vessel up to this point and split it only when the paths needed to diverge, we could serve the two regions using a lot less vessel material and a lot less blood to fill the vessels. Of course, you can extend this logic for all the blood vessels in the body, and you end up with a much more efficient design of branching blood vessels. And as it turns out, this happens if the cross-sectional area of the vessels stays the same before and after the branching. If you keep repeating this pattern across the network, you end up with a branching self-similar fractal. And if you look at the actual shape of the circulatory system, it has this geometry. But how do you get from this to quarter power scaling laws? Well, mathematician Felix Hausdorff discovered that self-similar fractals have an interesting property. Hausdorff assigned these space-filling fractal curves a value of 2.0, corresponding to their dimensionality. The same ideas apply to a 2D surface. Repeat the right pattern of folds at smaller and smaller scales, and a 2D surface fills more and more of a 3D volume. In the mathematical limit, it becomes space filling with Hausdorff dimension of 3.0. That enables an organism for a given size to pack in more of these metabolic surface areas than would be expected. As a result, the Hausdorff dimension of the surface of the circulatory system is roughly three, meaning its surface area doesn't scale as its length squared, but it's length cubed. And since the metabolic rate hinges on how fast resources can be exchanged across the surface area, well, it must also scale like length cubed. Since every cell needs to be served by the network, this means that the volume around the network should be proportional to the animal's mass. And since volume is just surface area times length and surface area is proportional to length cubed, that means both volume and mass must be proportional to length to the fourth, which can be rewritten to show that length is proportional to mass raised to the one quarter. And if you plug that into the equation for the metabolic rate, you find that the metabolic rate must be proportional to mass to the three-quarters, exactly as Max Kleiber had found. West, Brown and Enquist published their work in 1997, and it soon came to be known as WBE Theory, after their initials.十亿次心跳与人类的寿命特权
在WBE理论框架下，许多生物特征的缩放指数都可以完美预测：例如大动脉半径随质量的3/8次方缩放，肺面积则随11/12次方缩放。在此基础上，我们可以推导出哺乳动物的心率随体重的负1/4次方变慢，而寿命则随体重的正1/4次方延长。有趣的是，寿命与心率在数学上互为倒数。当我们将两者的缩放公式相乘时，质量因子便完全抵消，只剩下一个常数。这意味着对于几乎所有哺乳动物来说，无论体型大小，在其生命周期内分配到的心跳总数都是约十亿次。例如，寿命仅1-2年的鼩鼱心率高达每分钟1200次，而活到70岁的大象心率只有每分钟30次，两者的总心跳数均维持在十亿次左右。人类是唯一的特例。在19世纪中叶确立病原菌学说并改善公共卫生条件后，人类的平均寿命大幅攀升，使得现代人类一生中能获得接近三十亿次的心跳。这正是科学与技术赋予人类“额外多活两世”的直接证据。
Original English
Take a mammal's heart rate, for instance. Heart rate is equal to the blood flow rate over the amount or volume of blood in every beat. The volume of blood per beat has been found to scale in direct proportion to an animal's mass. So that's just M. And the blood flow rate? Well, remember that metabolism is all about how nutrients get distributed around the body. So most biologists agree metabolic rate and blood flow rate are directly proportional to one another. So heart rate should scale as metabolic rate over mass. Swapping in the scaling law Kleiber had found, that gives us M to the minus one quarter, meaning bigger animals should have slower heartbeats than smaller ones. And this is exactly what we observe in nature. The world's smallest mammal, the Etruscan Shrew, has an extraordinary heart rate of 1200 beats per minute. That's 20 beats per second. Whereas the biggest land mammal, the African bush elephant, has a typical heart rate of only 30 beats per minute. And we can do something similar for lifespan. One of the leading theories is that an animal's lifespan is based on the accumulation of metabolic damage. So the rate at which an animal accumulates damage is its metabolic rate per unit of mass, and its lifespan should be the inverse of that rate. So lifespan is proportional to M over B, or substituting in Kleiber's Law, M to the one quarter. So lifespan should increase with mass. And you do see this in nature. A shrew only lives for one to two years in the wild while a mighty African elephant can live up to 70 years. So it's the, you know, live fast and burnout and die young, right? Or spend it frugally and live a really long life. Heart rate scales as B over M. And lifespan scales as M over B. They're inverses of each other. So when you multiply these two terms, they cancel out, leaving you with just a constant. So that suggests that no matter what mammal you're talking about, it should have roughly the same number of heartbeats. The shrew's 1200 beats per minute multiplied by a lifespan of around 1.5 years gives you around 950 million heartbeats. Meanwhile, an elephant's 30 beats per minute multiplied by a lifespan of around 65 years gives you a little over a billion heartbeats. This is why nearly every mammal from a tiny field mouse to a gazelle, from a cheetah to a hippopotamus, they all get around a billion heartbeats between the day they're born and the day they die. But there is one major outlier, one mammal that gets significantly more than a billion heartbeats. And that is us, humans. We are the lucky ones. Three centuries ago, humans were much closer to the standard value of one billion heartbeats. But around the mid 1800s, germ theory and better sanitation methods became widespread, causing a stark decrease in the number of child mortalities and deaths from disease. So life expectancy began to climb up. And with it, the average number of heartbeats in a lifetime. We have systematically been increasing the number of heartbeats we get in our lifetime to the point where now the average human gets nearly three billion heartbeats before they die. I don't know if there's a better argument for science and technology than this. It has literally given the average human more than a full extra life.城市缩放定律：亚线性节约与超线性爆发
当杰弗里·韦斯特将这种生物网络规模缩放的物理视角引入人类社会时，他发现了极具启发性的现象：城市同样具有规模缩放效应，但其内部表现出了截然不同的双重特征。一方面，在基础设施网络层面，城市表现出类似于生物体代谢率的亚线性缩放（Sublinear scaling: 幂律指数小于1）。如道路长度、电缆数量和加油站需求量的缩放指数大约为0.85。这意味着当城市人口翻倍时，基础设施只需要增加约74%，带来了显著的“绿色节能”和规模经济效益。另一方面，在社会经济活动层面，城市展现了独特的超线性缩放（Superlinear scaling: 幂律指数大于1）。如国内生产总值（GDP）、专利发明数、平均工资，以及负面因素如犯罪率和疾病流行率，其缩放指数普遍在1.15左右。这意味着人口规模每翻一倍，人均产出和人均犯罪都会额外提升15%。城市就像是一个由人类社会网络交互所驱动的、自我加速运行的超线性引擎，这也定量地解释了为什么大城市的生活节奏（甚至包括人们在街头步行的物理速度）会明显快于小城镇。
Original English
Luis Bettencourt and others have looked at scaling laws in cities. They looked at how different properties like the amount of crime scale as the population of the city increases, and what they found is that if you plot serious crimes on a log log plot, the data clusters around a straight line with a slope of 1.15, meaning crime grows faster than linear or super linear. So for every doubling of a city's population, you get around 2.2 times as many criminal cases or around 120% more crime as opposed to the 100% you might naively expect. To make matters worse, researchers found that the same general pattern holds for the amount of wastewater and even the number of AIDS cases. In 2006, Dirk Helbing, Christian Kuhnert, and Geoffrey West looked at how the number of gas stations scales as a function of population. If a city is twice as big, does it need twice as many gas stations? They plotted the data on a log log plot and found a straight line. But the exponent wasn't one, it was about 0.8. This means that for every doubling, you only need around 74% more gas stations, which is a decent savings. The amount of roads and electrical cables also scale in roughly the same way. The rough figure that West gives in his book is that they all have scaling exponents of around 0.85. When it's shared resources, yes, cities can be surprisingly green. The argument is that cities can be even greener than you might think than living out in the middle of nowhere. But cities have even bigger benefits. Things like total wages, GDP, and the number of patents all scale superlinearly, with exponents that cluster somewhere around 1.15, meaning that for every doubling in size, you get around 120% more of each. All of this becomes especially significant when you compare a small town of say 50,000 people to a city 100 times its size because infrastructure needs only need to go up by a factor of about 50 while total wages, GDP, patents and inventions, they all go up by a factor of 200. So cities far from being detrimental to the world, they might actually be one of our best inventions and an indirect driver of a lot of scientific and technological progress. Perhaps this is also why people often say that life in the city feels faster. A feeling that seems justified because researchers looked at how fast people walk in cities of different sizes. And they found that people literally do walk faster in larger cities.📌 文中提及的人物和组织
人物: Geoffrey West, Max Kleiber, Brian Enquist, James Brown
媒体/书籍: Scale