How digital DNA could help you make better health choicesWhat if you could know exactly how food or medication would impact your health -- before you put it in your body? Genomics researcher Jun Wang is working to develop digital doppelgangers for real people; they start with genetic code, but they'll also factor in other kinds of data as well, from food intake to sleep to data collected by a "smart toilet." With all of this valuable information, Wang hopes to create an engine that will change the way we think about health, both on an individual level and as a collective.
1,334,433 views April 2017~~July 2022 | Jun Wang • TEDJun Wang
Genomics researcher
At iCarbonX, Jun Wang aims to establish a big data platform for health management.
Today I'm here, actually, to pose you a question. What is life? It has been really puzzling me for more than 25 years, and will probably continue doing so for the next 25 years. This is the thesis I did when I was still in undergraduate school. While my colleagues still treated computers as big calculators, I started to teach computers to learn. I built digital lady beetles and tried to learn from real lady beetles, just to do one thing: search for food. And after very simple neural network -- genetic algorithms and so on -- look at the pattern. They're almost identical to real life. A very striking learning experience for a twenty-year-old.
00:55
Life is a learning program. When you look at all of this wonderful world, every species has its own learning program. The learning program is genome, and the code of that program is DNA. The different genomes of each species represent different survival strategies. They represent hundreds of millions of years of evolution. The interaction between every species' ancestor and the environment.
01:34
I was really fascinated about the world, about the DNA, about, you know, the language of life, the program of learning. So I decided to co-found the institute to read them. I read many of them. We probably read more than half of the prior animal genomes in the world. I mean, up to date.
01:57
We did learn a lot. We did sequence, also, one species many, many times ... human genome. We sequenced the first Asian. I sequenced it myself many, many times, just to take advantage of that platform. Look at all those repeating base pairs: ATCG. You don't understand anything there. But look at that one base pair. Those five letters, the AGGAA. These five SNPs represent a very specific haplotype in the Tibetan population around the gene called EPAS1. That gene has been proved -- it's highly selective -- it's the most significant signature of positive selection of Tibetans for the higher altitude adaptation. You know what? These five SNPs were the result of integration of Denisovans, or Denisovan-like individuals into humans.
02:52
This is the reason why we need to read those genomes. To understand history, to understand what kind of learning process the genome has been through for the millions of years. By reading a genome, it can give you a lot of information -- tells you the bugs in the genome -- I mean, birth defects, monogenetic disorders. Reading a drop of blood could tell you why you got a fever, or it tells you which medicine and dosage needs to be used when you're sick, especially for cancer.
03:23
A lot of things could be studied, but look at that: 30 years ago, we were still poor in China. Only .67 percent of the Chinese adult population had diabetes. Look at now: 11 percent. Genetics cannot change over 30 years -- only one generation. It must be something different. Diet? The environment? Lifestyle? Even identical twins could develop totally differently. It could be one becomes very obese, the other is not. One develops a cancer and the other does not. Not mentioning living in a very stressed environment.
04:07
I moved to Shenzhen 10 years ago ... for some reason, people may know. If the gene's under stress, it behaves totally differently. Life is a journey. A gene is just a starting point, not the end. You have this statistical risk of certain diseases when you are born. But every day you make different choices, and those choices will increase or decrease the risk of certain diseases. But do you know where you are on the curve? What's the past curve look like? What kind of decisions are you facing every day? And what kind of decision is the right one to make your own right curve over your life journey? What's that?
04:57
The only thing you cannot change, you cannot reverse back, is time. Probably not yet; maybe in the future.
05:04
(Laughter)
05:05
Well, you cannot change the decision you've made, but can we do something there? Can we actually try to run multiple options on me, and try to predict right on the consequence, and be able to make the right choice? After all, we are our choices.
05:26
These lady beetles came to me afterwards. 25 years ago, I made the digital lady beetles to try to simulate real lady beetles. Can I make a digital me ... to simulate me? I understand the neural network could become much more sophisticated and complicated there. Can I make that one, and try to run multiple options on that digital me -- to compute that? Then I could live in different universes, in parallel, at the same time. Then I would choose whatever is good for me.
06:02
I probably have the most comprehensive digital me on the planet. I've spent a lot of dollars on me, on myself. And the digital me told me I have a genetic risk of gout by all of those things there. You need different technology to do that. You need the proteins, genes, you need metabolized antibodies, you need to screen all your body about the bacterias and viruses covering you, or in you. You need to have all the smart devices there -- smart cars, smart house, smart tables, smart watch, smart phone to track all of your activities there. The environment is important -- everything's important -- and don't forget the smart toilet.
06:43
(Laughter)
06:44
It's such a waste, right? Every day, so much invaluable information just has been flushed into the water. And you need them. You need to measure all of them. You need to be able to measure everything around you and compute them.
07:00
And the digital me told me I have a genetic defect. I have a very high risk of gout. I don't feel anything now, I'm still healthy. But look at my uric acid level. It's double the normal range. And the digital me searched all the medicine books, and it tells me, "OK, you could drink burdock tea" -- I cannot even pronounce it right --
07:23
(Laughter)
07:24
That is from old Chinese wisdom. And I drank that tea for three months. My uric acid has now gone back to normal. I mean, it worked for me.
07:34
All those thousands of years of wisdom worked for me. I was lucky. But I'm probably not lucky for you. All of this existing knowledge in the world cannot possibly be efficient enough or personalized enough for yourself. The only way to make that digital me work ... is to learn from yourself. You have to ask a lot of questions about yourself: "What if?" --
08:03
I'm being jet-lagged now here. You don't probably see it, but I do. What if I eat less? When I took metformin, supposedly to live longer? What if I climb Mt. Everest? It's not that easy. Or run a marathon? What if I drink a bottle of mao-tai, which is a Chinese liquor, and I get really drunk? I was doing a video rehearsal last time with the folks here, when I was drunk, and I totally delivered a different speech.
08:30
(Laughter)
08:33
What if I work less, right? I have been less stressed, right? So that probably never happened to me, I was really stressed every day, but I hope I could be less stressed. These early studies told us, even with the same banana, we have totally different glucose-level reactions over different individuals.
08:52
How about me? What is the right breakfast for me? I need to do two weeks of controlled experiments, of testing all kinds of different food ingredients on me, and check my body's reaction. And I don't know the precise nutrition for me, for myself.
09:11
Then I wanted to search all the Chinese old wisdom about how I can live longer, and healthier. I did it. Some of them are really unachievable. I did this once last October, by not eating for seven days. I did a fast for seven days with six partners of mine. Look at those people. One smile. You know why he smiled? He cheated.
09:37
(Laughter)
09:38
He drank one cup of coffee at night, and we caught it from the data.
09:43
(Laughter)
09:44
We measured everything from the data.
09:46
We were able to track them, and we could really see -- for example, my immune system, just to give you a little hint there. My immune system changed dramatically over 24 hours there. And my antibody regulates my proteins for that dramatic change. And everybody was doing that. Even if we're essentially totally different at the very beginning. And that probably will be an interesting treatment in the future for cancer and things like that.
10:14
It becomes very, very interesting. But something you probably don't want to try, like drinking fecal water from a healthier individual, which will make you feel healthier. This is from old Chinese wisdom. Look at that, right? Like 1,700 years ago, it's already there, in the book. But I still hate the smell.
10:34
(Laughter)
10:35
I want to find out the true way to do it, maybe find a combination of cocktails of bacterias and drink it, it probably will make me better. So I'm trying to do that.
10:44
Even though I'm trying this hard, it's so difficult to test out all possible conditions. It's not possible to do all kinds of experiments at all ... but we do have seven billion learning programs on this planet. Seven billion. And every program is running in different conditions and doing different experiments. Can we all measure them?
11:12
Seven years ago, I wrote an essay in "Science" to celebrate the human genome's 10-year anniversary. I said, "Sequence yourself, for one and for all." But now I'm going to say, "Digitalize yourself for one and for all." When we make this digital me into a digital we, when we try to form an internet of life, when people can learn from each other, when people can learn from their experience, their data, when people can really form a digital me by themselves and we learn from it, the digital we will be totally different with a digital me.
11:59
But it can only come from the digital me. And this is what I try to propose here. Join me -- become we, and everybody should build up their own digital me, because only by that will you learn more about you, about me, about us ... about the question I just posed at the very beginning: "What is life?"
12:30
Thank you.
12:31
(Applause)
12:37
Chris Anderson: One quick question for you. I mean, the work is amazing. I suspect one question people have is, as we look forward to these amazing technical possibilities of personalized medicine, in the near-term it feels like they're only going to be affordable for a few people, right? It costs many dollars to do all the sequencing and so forth. Is this going to lead to a kind of, you know, increasing inequality? Or do you have this vision that the knowledge that you get from the pioneers can actually be pretty quickly disseminated to help a broader set of recipients?
13:15
Jun Wang: Well, good question. I'll tell you that seven years ago, when I co-founded BGI, and served as the CEO of the company there, the only goal there for me to do was to drive the sequencing cost down. It started from 100 million dollars per human genome. Now, it's a couple hundred dollars for a human genome. The only reason to do it is to get more people to benefit from it. So for the digital me, it's the same thing. Now, you probably need, you know, one million dollars to digitize a person. I think it has to be 100 dollars. It has to be free for many of those people that urgently need that.
13:52
So this is our goal. And it seems that with all this merging of the technology, I'm thinking that in the very near future, let's say three to five years, it will come to reality. And this is the whole idea of why I founded iCarbonX, my second company. It's really trying to get the cost down to a level where every individual could have the benefit.
14:15
CA: All right, so the dream is not elite health services for few, it's to really try and actually make overall health care much more cost effective --
14:22
JW: But we started from some early adopters, people believing ideas and so on, but eventually, it will become everybody's benefit.
14:32
CA: Well, Jun, I think it's got to be true to say you're one of the most amazing scientific minds on the planet, and it's an honor to have you.
14:39
JW: Thank you.
14:40
(Applause)
xiao gu, Translator
Yolanda Zhang, Reviewer
00:00
今天我在这里给大家抛出一个问题 什么是生命 这个问题困扰了我至少25年 可能在未来的25年内依然如此 这是我在念本科时做的论文 当我的同学们依然拿 电脑作为大型计算器使用
00:26
我已经开始思考如何教电脑学习 我构建了一个数字化瓢虫 试图从自然界的瓢虫那里学习一件事 觅食 通过一个简单的神经网络
00:42
比如遗传算法等 来观察这个模型 它几乎和活生生的瓢虫没有区别 对20岁的我来说这是非常震撼的体验
00:55
生命本身就是一套学习程序 纵观大千世界 每个物种都有自己的学习程序 这个程序就是 基因组 程序的代码就是 DNA 每个物种的不同基因 代表了不同的生存策略 也代表了数亿年的进化演变 以及物种的祖先和环境的 相互作用
01:34
我十分着迷于这个世界 着迷于DNA 生命的语言 还有学习程序 所以我决定跟人合作建立 一个机构来解读它们 我做了很多相关研究 我们可能已经解读了世界上 超过一半动物的基因 到目前为止
01:57
我们收获了很多 我们对一个物种进行了多次基因测序 包括人类基因组 我们完成了第一个 亚洲人基因组的测序 我对自己的基因也进行了多次测序 为了充分利用这个平台 看看所有这些重复出现的碱基对 ATCG 你几乎无法从中读懂任何含义 但是再看一下这组字母 AGGAA 这五个字母 这五个SNP（单核苷酸多态性） 代表了一种非常特别的单形体 它们是位于藏族人身体中 一种叫做EPAS1的基因 这种基因已被证明是 高度选择性的结果 是藏族人对高海拔 适应性进行积极选择的 重要标志 这说明什么呢 这五个SNP极有可能是 灭绝的丹尼索瓦人 或与丹尼索瓦人 有亲缘关系的人种的DNA 与人类DNA结合的结果
02:52
这就是我们需要解读这些基因的原因 它可以让你了解历史 了解基因这套学习程序 在数百万年中经历了什么样的演变 通过解读基因可以得到很多信息 它能告诉你基因中的一些错误 例如出生缺陷 单基因遗传病 而仅仅需要一滴血 它就能解释你为什么会发烧 或者当你生病 特别是罹患癌症时 告诉你需要使用哪种药物和多少剂量
03:23
通过基因检测可以获得很多信息 不过看看这个数据 30年前 中国还很穷 当时只有0.67%的成年人患有糖尿病 而现在 这个数字是11％ 我们的基因不会在 30年内发生这么大的改变 在只有一代人的时间跨度内 肯定有别的影响因素 是饮食 环境 或者生活方式吗 即使是同卵双胞胎 都可以变得很不一样 一个很胖 一个很瘦 一个人得了癌症 另一个则没有 更不用说我们每天都生活在 一个充满压力的环境中
04:07
10年前我搬到了深圳 由于某些原因 大家可能都知道 一个人的基因在感受到压力时的 表现会完全不一样 人生是一场旅程 基因只是一个起点 而不是终点 你出生时就拥有某种疾病的患病风险 然而每天你都会做出不同的决策 这些决策会增加或 减少某种疾病的风险 但你知道自己在这个曲线上的位置吗 过去的曲线是什么样子的 你每天又面对着什么样的选择 在人生旅程的这张曲线图上 能让你处在正确位置上的 选择究竟是什么 那是什么
04:57
唯一你无法改变 无法逆转的事情 就是时间 目前是这样 不过这件事儿也说不好
05:04
（笑声）
05:05
既然不能改变已经做出的决定 那我们可以做些什么呢 我们能够同时运行不同的选择 对所期待的结果进行预测 再做出正确的选择吗 毕竟 每个人都是由自己的选择决定的
05:26
这些瓢虫后来启发了我 25年前 我构建了数字化的瓢虫 试图模拟自然界中的瓢虫 我是否可以同样 构建一个数字化的我 来模拟真实的我自己呢 我当然明白其中的神经网络可能会 变得更加复杂 那么我是否可以做一个 并尝试运行这个数字化的我 来计算出不同的选择结果呢 这样我就可以同时平行的生活在 不同的宇宙中 然后选择对我来说最合适的方案
06:02
我的生命数据可能是 这个星球上最全面的 我在自己身上可是做了不少投资 这个数字化的我告诉我 种种信息表明我有痛风方面的 遗传风险 需要不同的技术才能打造 这个数字化的我 你需要蛋白质 基因 代谢 抗体的数据 你需要筛查全身 搜集体内外所有细菌和病毒的数据 你需要各种智能设备 智能车 智能家居 智能桌子 智能手表 智能手机等等 来跟踪你所有的活动 要知道环境数据很重要 一切都很重要 另外别忘了智能马桶
06:43
（笑）
06:44
这简直就是一种浪费 是吧 每天这么多宝贵的信息 就这么被水冲掉了 你需要它们 需要测量这些数据 你应该测量并计算周围的 所有这些东西
07:00
而这个数字化的我告诉我 我有遗传缺陷 我患痛风的概率很大 我暂时还没什么感觉 看着挺健康的 但看看我的尿酸水平 是正常范围的两倍 而数字化的我搜索了 所有的医药典籍 告诉我 你可以喝牛蒡茶 这个词我甚至都不太会读
07:23
（笑）
07:24
这种茶来自中国的古老智慧 不过我喝了三个月的牛蒡茶 尿酸值就恢复正常了 这招儿对我还挺管用的
07:34
所有这些千年智慧 对我而言都是有用的 我很幸运 但可能对于你们来说就不一定了 世界上一切现有的知识 不可能对每个人都有效 不可能都是对症下药 能让这个数字化的我有效的唯一方法 就是从自己的身体中学习 你必须问自己很多问题 如果这么做会怎样
08:03
我现在有时差反应 你们可能看不出来 但我确实有 如果我少吃点呢 如果我服用二甲双胍 是否就可以活得更长呢 如果我去爬珠穆朗玛峰呢 可能没那么容易 或者去跑马拉松呢 如果我喝一瓶茅台酒 一种中国的烈性酒 真的喝醉了会怎么样 上次在这里跟工作人员 进行了一次录像彩排 我那会儿喝醉了 结果我的演讲内容完全跑偏了
08:30
（笑）
08:33
如果我的工作量少一点呢 我的压力是否就减轻了 这种情况从来没有发生在我身上 我每天压力都很大 但我希望我的压力能够小一点 这些早期研究告诉我们 即使吃同样的一根香蕉 不同个体的血糖水平反应 都可能完全不同
08:52
那我呢 一顿适合我的早餐应该吃些什么 我需要做两个星期的对照实验 测试各种不同的食物成分 并检查我身体的反应 我不知道对我来说精确的营养 到底应该包含什么
09:11
我想搜遍所有中国的古老智慧 找到让我活得更久 更健康的秘诀 我确实付诸行动了 不过其中有一些并不现实 我在去年十月份尝试了一次 七天不吃饭 我与我的六个伙伴一起进行了 为期七天的绝食体验 看看他们 有一个人笑了 知道为什么吗 他作弊了
09:36
（笑）
09:38
晚上他偷偷喝了一杯咖啡 我们直接从他的数据中发现了
09:43
（笑）
09:44
我们差不多解读出了 数据中的所有内容
09:46
我们能够追踪这些数据 而且切切实实的看到了 我的免疫系统的变化 举个例子 给大家一点直观的信息 我的免疫系统在24小时内 发生了巨大的变化 而基于这个巨大变化 我的抗体开始对我体内的 蛋白质进行调节 所有参与者都是如此 尽管每个人的免疫系统 天生各不相同 这很可能是将来治疗 癌症或类似疾病的 一个有意思的方法
10:14
而且正在变得越来越有趣 但有些方法你可能未必想尝试 比如饮用健康人的粪水 虽然这会让你感觉更健康 这也是来自古老中国的智慧 不妨看一下 1700年前 这种方法就已经被记录在册了 但我还是没法接受那个味道
10:34
（笑）
10:35
我想找出另一种方式 或许我们可以用混合了 益生菌的鸡尾酒来替代 会让我感觉好一点 所以我打算试试
10:44
尽管我依然在努力尝试 但是要测试出所有可能的 方法依然是非常困难的 针对每个个体做 各种实验也并不现实 但是我们在这个星球上 有70亿个学习程序 70亿 每个学习程序都在 不同的条件下运行 并进行不同的实验 我们可以测量所有这些个体吗
11:12
七年前 我在《科学》杂志上 发表了一篇文章 为了庆祝人类基因组计划10周年 我当时说过 测序你自己 为了个体 也为了全人类 但是现在我要说的是 数字化你自己 为了个人 也为了全人类 当我们把这个 数字化的我 变成了 数字化的我们 当我们试着构建 一个数字化生命网络 当人们可以互相学习 学习彼此的经验 彼此的数据 当人们真的可以自主 打造一个数字化的我 让我们得以中进行学习 这个 数字化的我们 将与 数字化的我 完全不同
11:59
但它只能来自 数字化的我 这就是我在这里提出的建议 加入 我 成为 我们 每个人都应该建立数字化的自我 因为只有这样 你才能更加了解你自己 了解 我 了解 我们 了解我最开始提出的那个问题 什么是生命
12:30
谢谢
12:31
（鼓掌）
12:37
克里斯·安德森（CA）： 我想快速的问一个问题 这项工作无疑相当出色 我猜大家可能还有一个问题 因为我们都期待着这些神奇的 个性化医疗技术成为可能 而在短期内 感觉还只有 少数人才能负担得起 是吗 仅仅是做基因测序就需要花很多钱 这是否会导致在某种程度上 增加了不平等 或者您是否有这样的构想 从这些早期的志愿者 身上获取的知识 可以被快速的复制推广 从而帮助更广泛的群体呢
13:15
王俊：很好的问题 十七年前 当我与人合伙创立华大基因 并担任这家公司CEO的时候 我唯一的目标就是 推动基因测序成本的下降 从最开始要花1亿美元完成 对一个人全基因组的测序 到现在只需要几百美元 我这么做的唯一原因 就是希望让更多的人从中受益 而 数字化的我 也是由此诞生的 目前你可能需要 一百万美元去数字化一个生命个体 但我认为在未来 必须要降到100美元 甚至是免费 尤其对于那些迫切需要 这项技术的人来说 这是必须的
13:52
所以这就是我们的目标 当这一切技术都能够融合之后 我认为在不远的将来 或许三到五年 一切就会成为现实 这就是为什么我创立了碳云智能 我的第二家公司 其目的就是要把成本降低到 让每个人都可以从中受益的水平
14:15
CA：好的 所以您的梦想 不是让它为少数的精英服务 而是真正要使 医疗健康服务更具普世价值
14:22
王俊：的确如此 但我们需要从 一些早期的先行者开始 从更加相信这个想法的一些人开始 但最终它将能够让每个人受益
14:32
CA：王俊 我不得不说 你是这个星球上 最令人叹服的科学家之一 真的很荣幸能邀请到你
14:39
王俊：谢谢
14:40
（鼓掌）
https://www.ted.com/talks/jun_wang_how_digital_dna_could_help_you_make_better_health_choices/transcript
