马克·扎克伯格与普莉希拉·陈：人工智能如何助力治愈疾病

这个领域将因为人工智能产生巨大的推动力。

This is a a space that I mean, that there's just gonna be a huge amount of leverage with AI.

在开发工具方面，这个领域还有很大的提升空间。说起来挺不可思议的，都2025年了，生物学还没有类似元素周期表那样的基础框架。

It still seems like there could be a lot more effort in this space around building tools, and it's kind of this crazy thing that we're, you know, here in, you know, 2025, and there's not the kind of periodic table of elements equivalent for biology.

我们认为，这可能是你最需要打造的一套关键工具。

We think that this is, like, probably one of the most important sets of tools that you need to build.

我们最初提出要在本世纪末治愈并预防疾病时，大多数科学家根本没法正经看待我们。

When we first set out that the goal to cure and prevent disease by the end of the century, people like, honestly, most scientists couldn't look at us with a straight face.

他们

And They're

觉得我们疯了。

like, you're crazy.

是的。

Yes.

这确实如此，因为如果你只是把钱砸给全国每个实验室最有可能成功的项目，根本不可能实现这个目标。

And it was true because if you just decide to spend the money funding the next best grant for every single lab in the country, like, you there's no pathway to that being true.

我觉得生物领域的人看这事儿，会觉得野心大得离谱。

The biology folks, I think, looked at it as if it were crazy ambitious.

然后搞AI的人就觉得，哎，这挺没劲的。

And then the AI folks are like, well, that's kind of boring.

这本来就是自然而然就会发生的事。

That's just automatically gonna happen.

所以我感觉，好吧，就这样吧。

So I feel like it's like, okay.

中间还有一段需要填补的空白。

There's something in between there that needs to be bridged.

在我们的夏季重播系列中，我们将重温过去一年里我们最喜欢的一段对话。

For our summer replay series, we're bringing back one of our favorite conversations from the past year.

随着人工智能继续重塑几乎每一个行业，它最重要的应用之一可能就在科学领域本身。

As AI continues to reshape nearly every industry, one of its most important applications may be in science itself.

在这次对话中，马克·扎克伯格和医生。

In this conversation, Mark Zuckerberg and Doctor.

陈和陈启智基金会正致力于开发新的生物学工具，以加速科学发现。

Chan discuss the Chan Zuckerberg Initiative's long term effort to accelerate scientific discovery by building new tools for biology.

他们没有专注于单一疗法，而是投资于基础设施、数据集和人工智能系统，相信这些是研究人员加快疾病研究所需的关键支撑。

Rather than focusing on individual therapies, they're investing in the infrastructure, data sets, and AI systems they believe researchers will need to make faster progress against disease.

我们探讨了细胞图谱、虚拟细胞模型、生物枢纽，以及这样一个观点：医学的未来，或许更依赖于工具的改进，而非仅仅依赖新的疗法。

We explore Cell Atlas, virtual cell models, Biohub, and the idea that the future of medicine may depend as much on better tools as on better treatments.

马克、普莉西拉，欢迎来到a16z播客。

Mark, Priscilla, welcome to the a16z podcast.

谢谢你们的邀请。

Thanks for having us.

是的。

Yeah.

很高兴能来。

Great to be here.

很期待。

Excited.

好的。

Alright.

很高兴有你们来。

Excited to have you.

你们在做很了不起的事。

You're doing exciting stuff.

是的。

Yeah.

为此，差不多十年前，你们创立了陈·扎克伯格倡议，目标是到本世纪末治愈、预防和控制所有疾病。

To that end, almost a decade ago, you guys started the Chan Zuckerberg Initiative with the mission and intent to cure, prevent, manage all disease by the end of this century.

你们本可以投入时间和资源去实现很多其他使命。

There's a lot of missions that you guys could have poured your time and resources into.

跟我们说说，当时为什么选择了这个方向？

Take us behind the conversations of why you guys picked this one.

或许，普莉西拉，我们先听你说说你的故事？

Maybe, Priscilla, why don't we start with you and your side story?

当我谈到我们做基础科学研究的方式时，总让人感到意外。

It always surprises people when I talk about how we work in basic science research.

我接受的是儿科培训，大家总是以为这肯定跟临床治病直接相关。

I trained as a pediatrician and people always think, oh, it must be about medicine.

我选择从医，是因为我想真正改善人们的生活。

And for me, I went into medicine because I wanted to improve people's lives.

我想带来一些改变。

I wanted to make a difference.

我想有能力去帮助别人。

I wanted to be able to help others.

在加州大学旧金山分校当住院医的时候，我接触了很多病人，尤其是那些孩子和他们的家庭——我们根本搞不清他们到底得了什么病。

And I think training as a pediatrician at UCSF, I met a lot of patients and frankly like little kids and families for which we just had no idea what the problem was.

运气好的话，他们能说出某个特定基因的名字；更多时候，他们只能被归到一大类疾病里，然后我们拿到一份打印出来的PDF，上面写着‘目前我们知道的就这些’。

And they might have a specific gene that they could name if they were lucky, Or they could be grouped into a bunch of other diseases and there'd be a general sort of PDF they'd print out like this is what we know.

那时候，作为实习医生或住院医，我的任务就是从这几行有限的信息里，想办法推导出该怎么照顾这个病人。

And then it was my job as an intern or resident to try to translate like a few lines of information to how we were supposed to take care of the patient.

对我来说，正是在那时，我真正意识到基础科学的力量，以及我们为何必须深耕基础研究，才能推动人类认知的边界。

And for me, that's when I really realized the power of basic science and how we need to work on basic science to advance the forefront of what's possible.

我觉得这就像一条希望的流水线。

I think of it as the pipeline of hope.

是的。

Yeah.

那你为什么会觉得自己能治愈所有疾病呢？

And why did you think you could cure all disease?

这目标也太激进了吧。

Because that's like a very aggressive goal.

你想来回答这个问题吗？

Do you want to answer that

嗯。

one?

对。

Yeah.

当然了，我们并不是说要直接治愈所有疾病，这点要说明白。

Well, I mean, we're not going to cure all diseases, to be clear.

我们的策略其实是帮助科学家和整个科研界去实现治愈所有疾病的目标。

I mean, the strategy is to help scientists and the scientific community cure all diseases.

所以核心策略就是加快基础科学的发展速度。

So the strategy is really one of accelerating the pace of basic science.

我们当时的观点是，纵观科学史，几乎所有重大突破都离不开新工具的诞生——这些工具让人能以全新的方式观察现象。

The theory that we had was if you look at the history of science, most major breakthroughs are basically preceded by the invention of a new tool to observe phenomena in a new way.

对吧？

Right?

所以我当时在想，

So I was

比如显微镜，能让人看到细菌，还有望远镜，能观测更远的宇宙。

thinking about things like the microscope, right, being able to observe bacteria or other fields, the telescope.

你懂的。

You know?

打个工程方面的比方，如果没有这些工具，就相当于写代码时没法单步调试，那真是以前的老办法了。

It's just to use an engineering example, without those kind of tools, it's kind of like you're coding without being able to step through the code and debug That's like the old days, when it's Yeah.

所以我们整个思路就是，帮着打造能加速整个领域进展的工具。

So our whole approach on this is basically let's help build tools that will accelerate the pace of the whole field.

我觉得这中间有个空白点，因为你看科学界的资金分配，绝大部分都来自政府和NIH的资助。

And I think that there's a niche that I think fits that because if you look at how funding works in science, the vast majority of funding comes from the government and NIH grants.

这些资金被拆分成一笔笔小额度的项目，通常只支持研究人员做些短期的课题。

It's parceled out into these relatively small grants that allow individual investigators to investigate usually pretty near term things.

而开发这类新工具——不管是成像技术，还是现在流行的虚拟细胞模型这类AI应用——往往周期长、成本高。

And the development of these kind of new types of tools, whether it's imaging or building now a lot of AI things like virtual cell models, are longer term oftentimes more expensive to develop.

想象一下，可能需要一亿到十亿美元，花上十年到十五年的时间来研发，然后把这些工具开放给科学界，推动整体进度。

So think about on the order of maybe $100,000,000 to $1,000,000,000 over a ten to fifteen year period, and then you try to unlock those tools and give them to the scientific community to accelerate the pace.

这就是我们的基本想法。

So that's kind of the theory.

没错，而且说实话，很多人在做工具的时候，其实很难因此获得应有的认可。

Right, and it seems like there's also something that is you don't really get credit for the tools in a lot of ways.

我的意思是，有公司一直在用你们的工具，而且用得特别满意，但我之前根本不知道还有这回事。

I mean, we have companies that use your tools, and they're very happy about it, but I didn't even know that that was the case.

所以这就是原因。

So That's

所以这才是慈善。

why it's philanthropy.

是的。

Yeah.

不过话说回来，

Well, is.

但大多数人做慈善，其实也图个名声。

But most people do philanthropy to get credit too.

我的意思是，这本来就是其中一部分。

I mean, that's kind of a part of it.

所以我想问，你有没有考虑过这一点，还是根本就没想过？

So I guess, did you think about that, or were you just like, no.

这件事一定能成，只要成功了，那就够了，对吧？

This is gonna work, and if it works, that's all we need?

我们特别专注的一件事，就是真正帮助每一位科学家提升能力，甚至扩展到初创企业创始人身上。

We're super focused on, like, actually making every scientist better and beyond science, like, startups, startup founders.

因为说实话，这事我们单打独斗是做不成的。

Because the point is we can't do this alone.

当初我们定下目标，要在本世纪末实现疾病治愈与预防，说实话，大多数科学家听后都直摇头。

And when we first set out the goal to cure and prevent disease by the end of the century, people like, honestly, most scientists couldn't look at us with a straight face.

他们觉得我们疯了。

They're like, you're crazy.

确实如此。

Yes.

当时确实如此，因为如果你只是把钱平均分给全国每个实验室，去资助他们下一个最好的项目，根本不可能实现这个目标。

And it was true because if you just decided to spend the money funding the next best grant for every single lab in the country, like you there's no pathway to that being true.

但如果你逼着大家认真思考：到底什么样的路径才最有可能实现这个目标？

But if you forced people to really think about this and like, okay, what is the most credible pathway to doing this?

那实现这条可行路径的障碍是什么？

And what are the barriers to that credible pathway?

后来我们终于找到了突破口。

Then we sort of got somewhere.

对吧？

Right?

他们说，根本就没有共享的工具，大家也不合作做大型项目，更别提构建合适的数据集了。

They were like, well, like, there's no shared tools or we're not working on big projects and building the right datasets.

我们就说，行吧。

And we're like, okay.

那我们就从头做起，动手解决这个问题——于是就有了开发共享工具的想法，因为现在科学界根本没人做这个。

Well then we can start And doing something about so that's where the idea of building shared tools, because no one right now in the science

这真有意思。

Well, that's so interesting.

所以你们的意思是，你们要彻底治愈所有疾病？

So basically, you're like, we're going cure all disease.

他们说：不可能。

And they're like Can't.

这根本做不到。

That can't be done.

为什么做不到？

Why can't it be done?

因为咱们没有合适的工具。

Well, because we don't have the tools.

明白了。

Okay.

这个过程还挺有意思的。

That's a pretty cool sequence.

是的。

Yeah.

说起来还有个挺有意思的现象，搞生物学的人觉得这目标太疯狂了，而搞人工智能的人却觉得这事儿有点乏味。

Mean, there's also this funny thing where the biology folks, I think, looked at it as if it were crazy ambitious, and then the AI folks are like, Well, that's kind of boring.

这会自然而然地发生。

That's just automatically gonna happen.

我知道这没问题。

I know that's okay.

中间还缺一座桥，如果你能用现代的人工智能工具，来打造生物学家真正需要的工具的话。

There's something in between there that needs to be bridged, and if you can use the modern AI tools in order to build the types of tools that biologists need.

所以我们看待自己工作的核心就是——

So that's a big part of how we think about our work is-

人工智能可能是有史以来被同时高估和低估的技术。

AI has got be the most overestimated and underestimated technology ever, simultaneously.

是啊，就像早期的互联网一样，但我们把自己和生物枢纽的工作定位为前沿生物学加前沿技术，对吧？

Yeah, It's so well probably like the internet early on, but we think about ourselves and the work that we're doing at the Biohub as frontier biology paired with frontier Right?

有些实验室专注前沿人工智能，专门研发最先进的模型；同时也有不少生物研究机构在做非常领先的研究，比如开发新数据集或攻克特定难题。

So there are labs that do frontier AI that basically are building the most advanced models, and then there are lots of biological research organizations that effectively do very leading edge research to build, to either discover new datasets or looking to certain challenges.

但到目前为止，还没人尝试把这两者结合起来。

But so far, there hasn't been anyone who's tried to do both of those at once.

你看，我是说，哪怕像AlphaFold这样厉害的东西。

When you look at I mean, even something like AlphaFold, which is amazing.

它是基于一个几十年前就产生的公开数据集构建的。

It was built off of this dataset that was a public dataset that had been produced decades ago.

我觉得如果你能把这两者结合起来，就能为训练AI模型专门生成数据集，从而构建能执行特定功能的虚拟细胞。

What I think you have the opportunity to do if you do both of those together is produce specific datasets for the purpose of training AI models to build virtual cells that can do specific things.

没错。

Right.

所以，我觉得这是一个挺有意思的领域。

So, I think that that's a pretty interesting zone to be in.

在我们做过的所有事情中，其实我们刚开始成立 CZI 的时候，重点放在了好几个方向上。

Of all the things that we've worked on, actually when we started CZI, we actually focused on a number of areas.

后来我们发现，科学研究的回报远远高于其他领域，所以我们就一直加大投入，反复坚持，直到今天十年过去了，Biohub 实际上成了我们慈善事业的核心重点。

What we found is just that the science researchers had by far the biggest returns, so just doubled down on it over and over and over until now we're at the point that we're ten years in, and Biohub is really the main focus of our philanthropy at this point.

对，基本上就是这个方向。

But, yeah, that's basically the focus.

也许你们没给自己足够的肯定，总觉得那些都是小规模的科研工作。

Maybe you're not giving yourselves enough credit because you're sort of saying, well, there's bite sized science.

我们可不想这样。

We don't wanna do that.

还有百年尺度的科学，听起来时间跨度很长，但其实是可实现的、有雄心的目标。

There's century scale science, and that seemed like a long time horizon, but achievable, ambitious.

但你们其实找到了一些特别棒的中间地带——那些宏大的科学挑战。

But you've actually identified, which I think is really fantastic, grand scientific challenges that are right in between.

它们的时间跨度是十年到十五年，至少从你们的表达方式和激发科学界热情的方式来看是这样。

They're ten to fifteen year horizons, at least per kind of the way you communicate about them and the way you energize

对。

Yeah.

你们用这种方式去激发科学界对这些目标的热情。

The scientific community about them.

十年到十五年是个挺有意思的时间范围，有点像风险投资支持的公司所规划的周期，也像一个团队能持续合作的时长。

Ten to fifteen is kind of an interesting time horizon, sort of like similar to the time horizon of a venture backed company, similar to the time horizon on which a team can work together for that period of time.

你们是怎么得出这个时间范围的？

How did you get to that number?

那你们又是怎么考虑每个十年到十五年阶段要应对的挑战的呢？

And then how are you thinking about the challenges that you take on in each ten to fifteen year wave?

因为这些目标具体、可实现。

Because that's concrete, achievable.

你们公布这些挑战的方式，已经积累了不少信誉。

You build a lot of credibility around it the way that you've announced those challenges.

我倒想听听你们是怎么想的。

Well, I'm curious how you guys think about it.

对我们来说，当我们看十年到十五年这个时间尺度上的重大挑战时，关键是要让人一看就觉得：有路可走。

But for us, when we looked at the grand challenges from the ten to fifteen year time horizon, it needs to be like, when you look at it, you're like, I see a path.

不一定要所有问题都解决了，我们才去接手。

Not everything needs to be solved for us to take it on.

事实上，如果所有问题都解决了，那反而觉得这应该

In fact, if everything's solved, then that feels like that should

而且目标还得足够有野心。

just And then one's ambitious enough.

对。

Yeah.

我们确实有一定的风险承受能力，对吧？

Like we have some risk appetite, right?

所以我们想找那些有明确实现路径、有靠谱负责人能推动的事情。

So we want things where we're like, there's a credible pathway, someone who is at the helm who can do this.

同时也要有足够的不确定性，让我们觉得值得去承担这个风险。

And there's enough ambiguity where we feel like we could take on that risk.

如果我们真做成了，回报可能还会超过预期。

And if we do it, like the returns could be higher than even expected.

我们在生物枢纽项目里是这样规划的：我们有三个生物枢纽。

And the way we model that in the Biohubs is we have three Biohubs.

一个在旧金山，一个在芝加哥，一个在纽约。

We have one in San Francisco, one in Chicago, one in New York.

纽约的团队专注于细胞工程。

The one in New York works on cell engineering.

我们能不能改造细胞，让它们进去探测信号、读取信息，或者执行特定操作？

Can we engineer cells to go in and detect signals, read it out, or to take certain actions?

芝加哥这边，我们在构建组织，研究组织内细胞之间的通讯。

In Chicago, we're building tissues and looking at cell communications within tissues.

而旧金山的团队则专注于深层成像和转录组学。

And then in San Francisco, we're looking at deep imaging and transcriptomics.

这些地点的选择都不是随意的。

And that work, the locations are not by accident.

我们还会考虑合作的大学，因为有人会来生物枢纽做研究，跨领域、跨学科，不受传统实验室的限制。

We also look at the partner universities because we have folks who come to the Biohubs to do this work, collaborative, inter interdisciplinary, and sort of unconstrained by the traditional lab.

同时，我们也依托这些高校的实验室来支持我们的工作。

But we also build off of the labs at these academic institutes that support the work.

所以我们是这样来选定重大挑战和落地地点的。

And so that's how we sort of choose the grand challenge the locations.

后来，大型语言模型和人工智能的加入特别有意思，因为我们本来就在开发工具来收集有价值的数据、构建数据集。

And then the sort of layering in the large language models and AI coming into the picture has been so interesting because we were already building tools to measure interesting data, building the datasets.

但那时候我们还不太清楚该怎么用这些数据。

But we didn't really know what to do with them yet.

等大型语言模型一出现，我们才恍然大悟：原来这些数据都能被解读出来了。

And large language models coming onto the scene were like, wow, we can make sense of all of this now.

我想知道，在治疗领域，你怎么定义成功？

I'm curious what you view success as in the therapeutic realm.

我们经常思考如何理解生物学，有时会投资那些想探索全新生物领域、攻克未知病因疾病的初创公司。

So, you know, we think a lot about understanding biology, and sometimes we bet on startups that want to unlock completely new biological areas, diseases where we don't know what's going wrong.

还有一类人则说，好吧，

And then there's another group of folks who kind of say, Hey, okay.

既然我们已经搞清楚了问题出在哪，那就去解决它吧。

Now that we understand what's going wrong, let's fix it.

用药物来干预。

Let's come in with a drug.

让我们引入一种全新的化学方法，一种新型的抗体。

Let's come in with a new type of chemistry, a new type of antibody.

你觉得在十年、二十年，甚至五十年后，CZI生物枢纽在推动新药研发方面，成功的标志会是什么？

What do you think success for the CZI Biohub looks like ten, twenty, fifty years from now in terms of the new medicines that you've enabled?

我们希望看到一个庞大的社群崛起，他们正在重新定义精准医疗的实践方式。

We want there to be like an explosion of a community who are building just the new wave of what it means to be deploying precision medicine.

无论是罕见病还是常见病，真正关键的都是每个人的个体生物学差异，而我们过去却把它们笼统地归为一类。

Like I think for rare diseases and common diseases alike, you're really talking about individual biology that we sort of lump together.

我们往往并不清楚这些差异具体是怎么发生的，对吧？

And we often don't know how it happens, right?

我们知道你有某种基因突变，最糟糕的情况是你携带了一个意义不明的变异。

We know that you have this mutation or the worst nightmare is you have a variant of unknown significance.

这到底意味着什么？

What does that even mean?

那个让人头疼的意义不明变异。

The horrible VUS.

是啊，确实可怕。

Yes, horrible.

你跟别人说你有点了解，但又说不清这到底意味着什么。

And you're like, you tell someone you kind of know something, but we don't know what it means.

但如果我们看看现在分析变异和单细胞转录组的方法，已经开始能说清楚：这个变异实际上影响了哪些下游细胞。

But if you look at the way we've been able to look at variants and look at single cell transcriptomics, we're starting to be able to say, Okay, this variant actually impacts this set of downstream cells.

接着我们再观察表达出来的蛋白质，看看它和健康细胞有什么相似或不同之处。

And then we start looking at the proteins that get expressed and how it looks similar or different to what a healthy cell would look like.

然后你就能开始锁定目标，比如把这个作为治疗靶点。

Then you can start targeting, Okay, like let's look at that as a target.

你既知道要针对的靶点有多精准——基于把突变和蛋白表达联系起来的能力，也能预测可能的脱靶效应。

And you both know the specificity of the target you want to build based on the ability to connect mutation to protein expression, as well as to be able to predict off target effects.

会有什么副作用？

What are the side effects?

因为你也能知道这种药还会在身体的哪些其他地方起作用。

Because you also know where else that drug will be able to interact with the body.

这些情况其实很罕见，我觉得大多数疾病都应该被看作罕见病，因为每个人的生理状况都不一样。

So those are rare, I really think most diseases should be thought of as rare diseases because each one of our biology is different.

但现在我们都被归到一堆里去了，对吧？

And right now we just get lumped, right?

我们通常只按年龄、人口统计特征、族裔来分类，要是运气好，还能考虑到这些因素。

We get lumped based on age, demographics, ancestry, if we're lucky to have that level of understanding.

但真正来说，每个人的生理状况都是独特的。就拿高血压或抑郁症来说，我们往往只能靠试错，先拿个药试试看效果如何。

But truly, each one of our biology is different and say like if you look at hypertension or depression, like we kind of just go by trial and error and saying like, let's just try that drug and see what happens.

真正该做的，是根据每个人独特的生理特征，精准、快速地进行治疗。

But what should really happen is being able to precisely and accurately and quickly treat people by looking at individuals biology.

我们希望推动基础科学研究，如果有人能用上我们搭建的模型，来开发诊断工具和治疗方法，我们会非常高兴。

We want to enable the basic science, and we would be thrilled if people picked up the models that we build to be able to build the diagnostics, the therapeutics that need to come.

你们构建的数据集真的很出色。

You've built amazing data sets.

不得不说，你可能听不到初创公司、制药公司和研发团队的反馈，但这些声音确实存在，因为你坚持了开源。

Have to say, you may not hear the feedback from the startup community and the pharma community and the R and D community, but it's there because you've committed to open source.

所以并不是每个人都会发论文，但大家都在用这些工具。

And so people may not all be writing papers, but they are using those tools.

我们投资的一家初创公司正在研究特发性肺纤维化。

There's a startup in our portfolio working on idiopathic pulmonary fibrosis.

从名字就能看出这病有多棘手。

The name tells you how vexing the disease is.

它是特发性的。

It's idiopathic.

我们根本不知道它为什么会发生。

We don't know why it happens.

IPF就是这么得名的。

IPF is named that way.

他告诉我，他用你们的细胞基因图谱，分析了成千上万的单个细胞，对比患者和健康人的数据，专门锁定成纤维细胞，再深入查看它们的基因表达。

And so, you know, he was telling me that he used your cell by gene atlases to look at millions of single cells in patients with disease, without disease, try to pinpoint the fibroblasts, double click on the fibroblasts and their gene expression.

这太了不起了。

It's incredible.

然后试着用这些发现来指引方向，看看在这么一种病因不明、莫名其妙的疾病里，新的药物靶点可能在哪里。

And try to, you know, use that to inform, hey, where could I go after a new drug target in this disease that's fundamentally a strange clump of idiopathic origin?

我觉得有一大批创新者特别喜欢这些工具、可视化界面和查询系统，还有你们打造的这套软件方式，让这些数据变得无比易用。

So I think there's a huge group of innovators who love the tools, the visualizations, the query systems, and really the software approach that you've built to making that data incredibly accessible.

细胞基因这个项目其实是个意外。

Cell by Gene is almost an accident, though.

你再多讲讲吧。

Tell us more.

你是想让你来介绍一下细胞基因，还是我来开头？

So do you want to share a little bit about Cell by Gene, or do you want me to start?

我不确定你想了解哪部分，但总的来说，细胞图谱这项工作挺疯狂的——都2025年了，生物学还没个像元素周期表那样的标准框架。

I don't know which part you want to get into, but the Cell Atlas work overall, it's kind of this crazy thing that we're here in 2025, and there's not the periodic table of elements equivalent for biology.

当初推动这个项目的主要想法就是：我们怎么才能通过生物枢纽的研究和其他资助项目，把各种数据整合起来，统一成一个标准化的格式？

That was a lot of the inspiration of it was, all right, how do we both, through work that we're gonna do in the Biohub and through other grants, be able to pull together and standardize a format where you can have all this data?

刚开始的时候，我们压根没想过要用这些数据来构建虚拟细胞模型。

When we were starting off, we didn't even necessarily have in mind that we were going to use that to build virtual cell models.

是的

Yeah.

我觉得随着人工智能的发展，这一点才逐渐清晰起来，这确实非常令人兴奋。

I think that that's sort of just come into focus as the AI work has advanced, but that's a very exciting thing.

我们 definitely 应该花不少时间聊聊虚拟细胞模型，但我还不太清楚你想重点讨论细胞图谱的哪部分。

We should definitely spend a bunch of time on the virtual cell models, but I'm not sure what you wanted get into on the Cell Atlas.

单细胞研究其实是我们十年前刚启动时首批资助的项目之一。

Well, the single cell work is was one of our first RFAs ten years ago we started.

当时我们就想，嗯，这事儿应该是能做成的。

And we're like, okay.

我们确信这是有可能实现的。

We think this is possible.

我们当时就资助了相关方法的研究，目的是建立统一的标准流程。

We actually funded the methodology for it to to standardize how it was gonna be done.

那是十年前的事了。

So that was ten years ago.

然后我们资助了几家实验室，让他们开始构建这个数据集。

And then we seeded a few labs to start building out that dataset.

但我们意识到，细胞类型和组合方式可能有数百万甚至数十亿种。

But we're like, there are like millions or billions of different cell types and different permutations.

我们到底该怎么推进呢？

Like, how are we going to do this?

尤其是当这项技术还在迅速发展的时候。

And especially with like a burgeoning technique.

所以我们最终支持了几个团队，他们开始开展工作。

And so we ended up seeding a few groups, and they started doing work.

后来他们告诉我们，遇到了一个问题。

And then they told us they had a problem.

他们的工作流程卡住了，因为数据标注速度跟不上。

There was a bottleneck in their workflow because they couldn't annotate the data fast enough.

于是我们开发了 Cell by Gene 这个标注工具。

So we built Cell by Gene was an annotation tool.

这个就是最初的想法来源。

That's the original source of this.

所以我们开发了这个标注工具，方便做单细胞研究的人轻松标注数据。

So we built the annotation tool to make it easy for people who are doing single cell science to be able to annotate the data.

然后我们把收集到的数据公开出来，让大家都能共享。

And then we put the data that we collected publicly so people could share.

因为大家都用同一个标注工具，数据格式也就自然统一了。

But because everyone started using the same annotation tool, everyone was standardized then on the same data formats.

后来围绕这个工具逐渐形成了一个社群，大家希望互相分享，一起构建这个图谱。

And then there started being a community around the tool, and they wanted to share back and build the Atlas.

如今十年过去了，这个共享资源已经汇集了数百万个细胞数据，供整个科学界使用。

So now after ten years, there are millions of cells that have been built into this shared resource for the entire scientific community.

我们只资助了其中大约百分之七十五。

We only funded about 75% of it.

抱歉，我说错了。

Sorry, that's wrong.

我们只资助了其中的25%。

We've only funded 25% of it.

剩下的75%来自整个科研社群，他们觉得这很有用，而且提供了一种简单的方式，让大家能一起标准化、共同建设。

75% came from the broader community saying, this is useful and there's an easy way for us to standardize and build these together.

相同的元数据。

The same metadata.

是的。

Yeah.

没错。

That's right.

这其实挺有意思的，有点像你所说的网络效应。

It's like an interesting what you'd call a network effect.

对。

Yeah.

效应。

Effect.

是的

Yeah.

我正想说呢

Was gonna say,

听起来就像互联网一样

it sounds like the Internet.

添加注释，保持虚拟细胞模型

Up annotations, stay the virtual cell model.

对

Mhmm.

我们在刚开始做这项工作时，非常重要的一点是，所有参与者都要使用统一的格式

Well, it was very important when we were getting started with the work to have everyone who was doing it have a consistent format.

这样才方便使用和移植

So that way it could be used and portable.

等这种做法逐渐普及开来，其他人也就发现它很有价值了

And then once that kind of took off as as the way that it would get done, then other people just found it valuable.

是的。

Yeah.

即使是跟以前的数据库比如geo之类的相比，它们的标准化程度和质量控制也远远不够。

Even relative to prior databases like geo and and whatnot, they're just simply not as standardized or QC.

是的。

Yeah.

对。

Yeah.

咱们来聊聊虚拟细胞吧。

Let's get

进入虚拟细胞的话题。

into virtual cells.

你觉得最大的挑战是什么？

What what are the great challenges that you would focus on?

不妨说说它的前景和希望，以及目前我们面临哪些困难或进展到什么程度了。

Maybe talk about what is the promise or the hope and maybe some of the challenges or where we're at with it.

是的。

Yeah.

我的意思是，我们认为这将成为目前最重要的工具之一，主要是从蛋白质开始，逐步构建细胞内部不同结构，再到整个虚拟免疫系统，涵盖多个层级的体系。

I mean, we think that this is going be one of the most important tools at this point, is basically building up the hierarchy from proteins to just different structures within the cell to whole to, like, whole, like, in a virtual immune system or different levels of hierarchy.

我们早就知道，这套工具对科研人员提出假设将起到至关重要的作用。

And we knew that this is gonna end up being, like, a very important set of tools for people to effectively generate hypotheses for for different science work.

甚至在你真正开展完整实验之前，就能初步估算出实验可能的走向。

You know, even before you get to the point where you're really running full experiments in it, you can come up with some, estimate of how that might run.

它对普丽西拉刚才提到的精准医疗应用场景也会很有帮助，但我们觉得，这可能是你最需要构建的核心工具集之一。

It will be useful for some of the precision medicine type examples that Priscilla was talking about a few minutes ago, but we think that this is probably one of the most important sets of tools that you need to build.

这不是单一的东西，对吧？

It's not a single thing, right?

可以从多个角度来切入这个问题。

There's different angles to come at this from.

细胞图谱的数据有助于在细胞层面理解各种现象。

The Cell Atlas data is helpful for understanding things on a cellular level.

我们现在做的最重要的一件事之一，就是这家叫进化尺度的公司，他们原本有一批在Meta做蛋白质折叠模型的研究员，现在加入了生物枢纽。这个团队的负责人亚历克斯·里夫斯，将全面负责整个科学项目。这其实挺有意思的，因为这意味着人工智能和生物学的融合，是由一个真正懂生物学的人工智能专家来领导，而不是一个懂点AI的生物学家——这多少反映了我们对这两者重要性的判断。

One of the most important things that we're doing right now, there's this great company, Evolutionary Scale, who actually had a bunch of researchers who'd formerly worked at Meta on protein folding models, is joining Biohub, and Alex Reeves, the leader of it, is actually going to be the head of the whole science program, which is actually interesting when you think about it, where it's like you have AI and biology coming together, and really it's like an AI person who understands biology is running it rather than a biologist who has some understanding of AI, I think, just speaks a little bit to where we think the relative weight of these things is.

就像普里西拉说的，各个生物枢纽的分工不同，纽约这边专注细胞工程，目标是让细胞能记录身体内部的各种变化，并把数据共享出来，再用来构建模型。

We basically view, like Priscilla was saying, with the different Biohubs, New York doing cellular engineering will basically make it so that you can have cells that can record different things that are going on around the body and share that data, then you can build that into models.

芝加哥的生物枢纽则专注于记录炎症反应，通过研究这些现象，形成另一套独特的数据集。

The Chicago Biohub being able to record inflammation, and and basically study that in order to kind of help understand, like, that's a that's a different dataset.

我们还有成像研究所，刚刚训练了第一批模型，这是首批能理解细胞在不同状态下的空间分布特征的模型。

We have the Imaging Institute, which is we just trained our our first set of models around that, which are the first, like, spatial models around understanding, like, the way that that kind of cells look in different states.

长远来看，就像工业界做语言模型那样，一开始是各种专项能力，慢慢训练整合，越来越通用——我们也是这个思路。

And, eventually, just like you have this analogy on the industry side around language models where you have different capabilities, then over time you train them into models and it gets more and more general, that's kind of the idea here.

我们会围绕重大的生物学难题来建设生物枢纽。

We'll build the Biohubs around grand biological challenges.

每个生物枢纽都会开发工具，用来生成全新的数据集。

The Biohubs will build tools that will generate novel datasets.

然后我们基于这些数据构建模型，最终把这些模型整合成一个越来越全面的虚拟细胞图景。这套系统对科研人员有用，也希望能帮到那些在药物研发领域的初创公司和企业——虽然这不是我们直接负责的部分，但毫无疑问，这是整个体系中非常关键的一环。

We will build models based on those, and then eventually combine the models into an increasingly general view of a virtual cell that will be useful both for scientists and hopefully startups and companies that are working on finding drugs, which is not our part of the whole thing, but I think is obviously a really important part of what needs to happen.

是的

Yeah.

你们在做投资时，总是会考虑风险。

And you you guys think about risk all the time in terms of when you make investments.

我觉得，用虚拟细胞模型来做虚拟生物学的最大优势，就是能大胆尝试那些更有风险的想法。

Like, I think the promise of being able to do virtual biology using a virtual cell model is you can actually take on riskier ideas.

现在，申请科研经费很难，湿实验又贵又慢。

Right now, like grant funding can be hard to come by, and the wet lab work is expensive and slow.

这不只是钱的问题，时间也很关键。

And it's not just money, it's also time.

所以你只能选那些看起来成功率较高的项目，才能维持你的科研生涯。

And so you have to choose something that you think is gonna have some likelihood of success to keep your lab career going.

这就让研究人员倾向于只冒一点小风险，毕竟得保证有一定成功率，才能拿到终身教职、发论文，或者完成其他目标。

And so it naturally lends people to take on like some risk, but not a lot of risk because they need to make sure that they are hitting like a certain percentage of the time to make tenure or publish or whatever they need to do.

但如果你有了一个能高度模拟生物过程的虚拟细胞模型，就可以先在电脑上测试和调整，提出那些在实验室里因成本太高、耗时太长而不敢碰的问题，先在计算机里验证一下有没有希望，再决定是否投入大量时间和资金去做湿实验。

But if you had a virtual cell model where you could simulate really high quality biology, you could actually then start testing and tinkering on the computational side and like ask riskier questions, things that would have been expensive and costly in terms of time and resources to do in the lab, and actually see if there is promise doing the experiments in silico before you make the time and money investment in the wet lab.

你有没有

Do you

觉得它有点像模式生物？

think of it kind of like a model organism?

是的。

Yeah.

就像新一代的果蝇？

Like it's the new fruit fly?

对。

Yeah.

我本来想问，细胞这么复杂，你觉得这个模型能做到多精确？

I was going to ask, given the complexity of a cell, how accurate do you think you'll get the model to?

我的意思是，就算你能完美还原一个细胞，虚拟细胞要精确到什么程度才算真正有用？

Mean, just assuming maybe you get it to a perfectly accurate representation of a cell, but like how accurate to be useful would the virtual cell have to be?

我觉得它肯定会不断迭代，越来越精准，因为现在我们还只停留在转录组层面。

I think it will obviously iterate and get better and better because right now we're still just talking about transcriptomics.

我们正在探索更多观察细胞的方法，同时精度也在不断提升。

We're expanding into different ways of looking at the cell, But you get more and more accuracy.

但我觉得没必要做到百分之百准确才能有用，只要能在前期稍微降低一点风险就够了。

But I don't think it needs to be 100% accurate to be useful because you just want to be able to de risk the idea on the front end a little bit.

风险越低，效率自然就越高，这很明显。

And the more and more you de risk it, the more efficient it gets, obviously.

哪怕只是获得一个大致的方向性信号，就已经很有用了。

But it'll be useful if you even get directional signal.

是的，我同意。

And yes, I do.

我们确实把它当作一种模式生物，只不过它能更真实地反映人体的状况。

We do think about it as a model organism, but in a way that has fidelity to the human body.

我们不希望——

Don't want to-

所有模型都是不完美的。

All models are wrong.

有些还是很有用的。

Some are useful.

对。

Yes.

没错。

Yes.

希望它在某些方面能派上用场。

Hopefully, it has utility on certain aspects.

正是如此。

Exactly.

就像语言模型一样，我们会专门加入某些功能。

And just like the language models, you build in specific capability.

比如我们最近发表的一个模型叫变异体转换器。

So it's not so for example, one of the models that we're we're publishing is is variant former.

对吧？

Right?

本质上，它是通过大量成对数据来训练的。

Basically, makes it so that it's trained on a bunch of effectively pairs.

如果你对一个细胞在某个位置进行CRISPR编辑，就能看到另一端会出现什么结果。

If you have a cell, you apply CRISPR to it in a place, you see what comes out at the other side.

它基本上能做出这样的预测。

It basically is able to make that kind of a prediction.

比如，当你对一个细胞做了某种编辑，大概率会发生什么？

Like, if you have this edit that you're doing to a cell, what is likely going to happen?

另一个模型是扩散模型。

Another one of the models is it's this diffusion model.

你可以描述一种想让它模拟的细胞类型，它就会生成一个该细胞的合成模型。

Basically, you can describe a type of cell that you would like it to simulate, then it will just produce a kind of synthetic model of of of the cell.

这其实挺有意思的，就像之前Priscilla提到的，每个人都不一样，细胞也各有差异，你得能模拟这些罕见的组合。

Again, I mean, it's kind of interesting because to Priscilla's point before about how everyone is different and and, like and different cells have have kind of you know, so you wanna be able to simulate these kind of rare configurations.

至少能有个合成版本来想象它长什么样，这很有价值，之后还能拿它来做测试。

Having at least a synthetic version of what that could look like is interesting, and then you can test against that.

我觉得这个冷冻电镜模型挺有意思的，因为它能体现空间信息。

The cryo model I think is interesting because it's spatial.

它让你意识到，其实你可以用各种不同的模型来看待不同类型的问题，然后随着时间推移，逐步让它们变得更通用。

It gives you a sense of there are all these different models that you can have that allow you to basically look at different kinds of things, then you just train them in to be increasingly general over time.

真厉害。

Wow.

非常有意思。

Very interesting.

这些建模技术主要是大语言模型，还是说有专门的推理模型？

Is the modeling technology basically LLMs, or is there a reasoning model?

是说只是……

Is it just a Oh,

这个其实也很有意思，没错。

that's actually a Yeah, no, that's a fascinating one too.

因为有一个新模型，虽然还处于非常早期的阶段，但它是第一个针对生物学的推理模型。

Because one of the new models I think this one is very early, but it's basically the first reasoning model over biology.

所以意思是，你实际上拥有这些模型，它们以不同方式模拟世界运行的机制，而你希望的不只是让它们输出相关性，比如发现了什么，而是真正能推演事物如何演变、为什么会发生。

So the idea is that, yeah, you you you effectively have these models that that kind of simulate world models in different ways, and then you want it to be able to not just be able to spit out correlations, right, in terms of, like, what it's found, but actually be able to kind of reason through how things would evolve and why things would happen.

我知道这还很早期，但从概念上讲，它确实很有意思，我觉得这明显会是模型未来发展的重要方向。

I know it's quite early, but it's but it is interesting conceptually as what I think is clearly going to be an important direction in terms of how these models evolve.

对。

Yeah.

因为这正是我在想的。

Because that's what I was thinking.

如果它行不通，下一个问题就是为什么。

If it doesn't work, the next question you have is why.

是的。

Yeah.

我觉得在推理方面，关键在于你的假设是否站得住脚。

I think what you find in reasoning, Because it's the a merit to your hypothesis.

当然，当然。

Sure, sure.

是的，我的意思是，我以为你在说，如果推理模型失效了，那原因是什么？

Yeah, mean the Yeah, I thought you saying if the reasoning model doesn't work, why?

我的意思是，我觉得这确实——没错，确实是这样。

I mean, I think that- Yeah, well, yeah.

对。

Yeah.

没错，这正是

Yeah, that's

你思考问题的方式。

kind of your way of thinking.

用语言模型来类比的话，那就是你需要更完善的全局模型或更好的预训练模型，才能让推理能力变强。

That's kind mean, the language model analogy for that would be you need better kind of world models or better pre trained models in order to get the reasoning to be good.

但这确实也是。

But it's yeah.

你只是不断往里加功能，我觉得这其中可能还存在一个先后顺序。

You just you build more you build more capabilities into it, I think that there's probably an order too.