From Newton to Hawking, scientists love wagers. Now Lewis Wolpert has bet Rupert Sheldrake a case of fine port that: "By 1 May 2029, given the genome of a fertilised egg of an animal or plant, we will be able to predict in at least one case all the details of the organism that develops from it, including any abnormalities." If the outcome isn't obvious, then the Royal Society will be asked to adjudicate.

Lewis Wolpert

I HAVE entered into this wager with Rupert Sheldrake because of my interest in the details of how embryos develop, and how our understanding of this process will progress. In my latest book, How We Live and Why We Die, I suggest that it will one day be possible to predict from an embryo's genome how it will develop, and I believe it is possible for this to happen in the next 20 years.

I am, in fact, being a little over-keen because 40 years is a more likely time frame for such a breakthrough. Cells and embryos are extremely complicated: for their size, embryonic cells are the most complex structures in the universe.

Animals develop from a single cell, a fertilised egg, which divides to produce cells that will form the embryo. How that egg develops into an embryo and newborn animal is controlled by genes in the chromosomes. These genes are passive: they do nothing, just provide the code for proteins. It is proteins that determine how cells behave. While the DNA in every cell contains the code for all the proteins in all the cells, it is the particular proteins produced in particular cells that determine how those cells behave.

Every cell of the embryo contains many copies of several thousand different proteins. These proteins have a plethora of functions: acting as enzymes to break down and build other molecules, providing structures for the cell, interacting with each other, and many more. The complexity of the interactions between millions of molecules is amazing.

As the proteins determine how the cells behave, it is their activity that causes the embryo to develop. Underlying this process, though, are the genes, as they control which proteins are made - including some proteins that activate specific genes. It is essential that there is this control over which cells continue to divide, and of mechanisms to pattern the embryo so that different cells develop into different structures, such as the brain or limbs.

There is a huge incentive to understand these processes and so be able to work out the development of an embryo given only its genome. This ability could pave the way for regenerative medicine by allowing scientists to program stem cells to become structures that could replace damaged parts of the body.

To win the bet, we will have to be able to predict the behaviour of almost all the cells in the embryo. In a small worm, say the nematode Caenorhabditis elegans, there are 959 cells, making it the ideal model to solve this problem. It is a major challenge, but advances in cell biology, systems biology and computing will take us there.

Rupert Sheldrake

LEWIS WOLPERT's faith in the predictive power of the genome is misplaced. Genes enable organisms to make proteins, but do not contain programs or blueprints, or explain the development of embryos.

The problems begin with proteins. Genes code for the linear sequences of amino acids in proteins, which then fold up into complex three-dimensional forms. Wolpert's wager presupposes that the folding of proteins can be computed from first principles, given the sequence of amino acids specified by the genes. So far, this has proved impossible. As in all bottom-up calculations, there is a combinatorial explosion. For example, by random folding, the amino-acid chain of the enzyme ribonuclease, a small protein, could adopt more than 1040 different shapes, which would take billions of years to explore. In fact, it folds into its habitual form in 2 minutes.

Even if we could solve protein-folding, the next stage would be to predict the structure of cells on the basis of the interactions of millions of proteins and other molecules. This would unleash a far worse combinatorial explosion, with more possible arrangements than all the atoms in the universe.

Random molecular permutations simply cannot explain how organisms work. Instead, cells, tissues and organs develop in a modular manner, shaped by morphogenetic fields, first recognised by developmental biologists in the 1920s. Wolpert himself acknowledges the importance of such fields. Among biologists, he is best known for "positional information", by which cells "know" where they are within the field of a developing organ, such as a limb. But he believes morphogenetic fields can be reduced to standard chemistry and physics. I disagree. I believe these fields have organizing abilities, or systems properties, that involve new scientific principles.

The Human Genome Project has itself set back the hopes it engendered. First, our genome contains only between 20,000 and 25,000 genes, far fewer than the 100,000 expected. In contrast, sea urchins have about 26,000, and rice plants 38,000. Moreover, our genome differs very little from the chimpanzee's genome, the sequencing of which was completed in 2005. As Svante P??bo, director of the Chimpanzee Genome Project, commented: "We cannot see in this why we are so different from chimpanzees."

Second, in practice, the predictive value of human genomes turns out to be low. Everyone knows tall parents tend to have tall children, and recent studies on the genomes of 30,000 people identified about 50 genes associated with being tall or short. Yet together these genes accounted for only about 5 per cent of the inheritance of height. This is not the only example of "missing heritability". Steve Jones, professor of genetics at University College London says that "hubris has been replaced with concern", and he suggests the present approach is "throwing good money after bad".

Wolpert is not alone in believing in the predictive value of the genome. Governments, venture capitalists and medical charities have bet and are still betting billions of dollars on it. More than a case of fine port is at stake.

A brief history of wagers

Scientific wagers date back to Greece in the 5th or 6th century BC and were often a rhetorical device for thinking about a subject. In their current form, they can also help stimulate fresh thinking.

One of the famous wagers of the more modern era was announced by Christopher Wren in 1684. He would give a book worth 40 shillings to anyone who could deduce Kepler's laws from the inverse-square law. Isaac Newton took this seriously and his deliberations eventually became his Principia - but too late to claim the prize.

In 1959, physicist Richard Feynman bet $1000 that it was impossible to build a motor no bigger than 1/64 of an inch on each side. He lost: electrical engineer Bill McLellan succeeded. Feynman was said to be disappointed because he hoped his bet would stimulate new technology, but McLellan's motor used existing techniques.

从牛顿到霍金，科学家们都爱打赌。如今Lewis Wolpert跟Rupert Sheldrake打赌说："到2029年5月1日，只需一颗受精卵，无论动物还是植物，我们就能预测出至少在一种情况下这颗受精卵成长过程的全部细节，包括所有异常情况。"如果结果并不明显，Lewis Wolpert就会接受英国皇家学会的审判。

Lewis Wolpert的独白：

之所以跟Rupert Sheldrake打赌是因为我对胚胎成长的过程很感兴趣，并且希望能对其有更深入的了解。在我最近出版的《How We Live and Why We Die》中，我认为总有一天人们能从胚胎的基因中预测出它成长的过程，我也相信这一设想会在未来的20年内实现。

实际上，我可能过于心急了，40年时间对实现这一突破似乎更有可能。因为细胞和胚胎结构极其复杂：单从尺寸上来讲，胚胎干细胞是宇宙中最复杂的结构。

动物们从一颗受精卵衍化而来，受精卵产生组成胚胎的细胞。染色体中的基因控制着卵子变成胚胎和新生动物的过程。但是这些基因十分懒惰：它们什么也不做，只为蛋白质提供编码。因此是蛋白质决定了细胞的行为。而细胞中的DNA包含所有细胞蛋白质的编码，只有个别细胞产生的特殊蛋白质才决定细胞行为。

胚胎中的每一个细胞都包含上千种不同蛋白质的复制品。这些蛋白质功能过剩：它们会像酶一样分解物质，或形成其它分子，或为细胞赋予结构，有些还会与其它蛋白质进行互动等等。数百万蛋白质分子同时进行活动的复杂状态令人吃惊。

蛋白质决定细胞行为，蛋白质的活动促使胚胎发展。但是这一过程的始作俑者是基因，包括某些需要蛋白质激活的基因，因为它们控制蛋白质的形成。基因的控制必不可少，只有它们决定哪些细胞继续，这样不同的细胞才会成长为不同的结构，如大脑和四肢。

只有了解基因，才能从一颗受精卵中判断胚胎的发展。科学家们还可以将研究结果应用到再生医学上去，用干细胞培育器官来替换身体内的坏死部分。

要想赢得这场战斗的胜利，我们必须能够预测胚胎中所有细胞的行为。以某种小型土壤线虫为例，它有959个细胞，是解决这一问题的理想模型。很显然，这是一项巨大的挑战，但是细胞生物学，系统生物学和计算机技术的发展会帮助我们将梦想变成现实。

Rupert Sheldrake的独白：

Lewis Wolpert竟然寄希望于基因真是异想天开。基因促使组织制造蛋白质这的确没错，但是它们既没有计划，也不能解释胚胎们的发展。

一切问题的根源在于蛋白质。基因控制蛋白质中线性氨基酸类的编码，这些氨基酸再折叠形成复杂的立体结构。Wolpert认为只需特定基因的氨基酸就能判断蛋白质折叠的结果。迄今为止，这是根本不可能的。因为蛋白质折叠的可能性数不胜数。例如，通过随机折叠，核糖核酸酶（一种小型蛋白质）的氨基酸链能形成超过1040种不同的结构，单这一种蛋白质就需要数亿年的时间来探索。而实际上，氨基酸链折叠的过程只需两分钟。

即使我们能抓住蛋白质折叠的规律，下一步就是通过分析数百万蛋白质和其它分子之间的相互作用，来预测细胞的结构。这势必会引发另一次更大规模的信息爆炸，因为这一过程产生的可能性比宇宙中所有的原子数量还要多。

仅凭分析随机分子排列的规律不可能解释器官的形成。相反，早在20世纪20年代发育学家们就认识到细胞，组织和器官是按照特定的模式而生长，这种模式是由形态发生场所而决定的。Wolpert知道这些场所的重要性。在生物学家之中，他以知晓"位置信息"而闻名，位置信息就是细胞"知道"其在生长器官中的位置，比如四肢。但是他认为形态发生场所会被周围的化学或物理作用而削弱。这一点我不认同。我相信形态发生场所具有组织能力或者系统功能，对其的研究将会发现新的科学原理。

人类基因组计划就证明了这一预测很不现实。首先，我们的基因组只包含2万至2万5千个基因，与预期的10万相去甚远。相比较，海胆有2万6千个基因，而谷类植物的基因有3万8千个。此外，据2005年的研究显示人类与黑猩猩基因差别很小。黑猩猩基因组计划的负责人Svante P??bo曾说过："我们不能从黑猩猩的基因组中判断出为什么我们与黑猩猩不一样。"

其次，实际上，人类基因组的预测价值很低。每个人都知道高个家长容易有高个孩子，而最近对3万人的基因组进行鉴定后发现只有50个基因与人的高矮有关。这些基因加在一起只对身高遗传起到5%的作用。这并不是"失传现象"的唯一例证。伦敦大学学院的遗传学教授Steve Jones说过："骄傲已蒙蔽了忧虑的双眼。"他认为目前的研究方向是"赔了夫人又折兵".

Wolpert并不是唯一一个坚信人类基因组预测价值的人。政府部门，资本家们以及慈善机构都在上面下注，一掷千金。这样做的结果很危险。

科学家打赌简史

科学家打赌的历史可以追溯到公元前5、6世纪的希腊，那时打赌是一种用来刺激人们思考的手段。就现在来看，打赌仍旧可以激发人们的灵感。

现代最着名的打赌发生在1684年。Christopher Wren打赌如果有人能用平方反比定律推论开普勒定律，他就会将一本价值40先令的书送给这个人。Isaac Newton经过深思熟虑最终形成了他的着名理论，但对于领取奖赏为时已晚。

1959年，物理学家Richard Feynman打赌1000美元，预言不可能有人制造出边长不超过六十四分之一英尺的马达。最终电机工程师Bill McLellan抱得美元归。Feynman称这一结果令他很失望，他本希望这次能刺激人们进行技术创新，但是McLellan制造的马达仍使用现有技术。

1975年，Stephen Hawking与同伴宇宙学家Kip Thorne曾打赌天鹅座X-1是否含有黑洞，赌注是输家为赢家订阅杂志。结果Hawking认输，也恰好从这时起Hawking开始花费大量时间研究黑洞。