同构实验室探讨人工智能驱动的药物发现与医学的未来 - 第252期

你好，欢迎收听NVIDIA人工智能播客。

Hello, and welcome to the NVIDIA AI podcast.

我是您的主持人诺亚·克拉维茨。

I'm your host, Noah Kravitz.

今天我的嘉宾是马克斯·亚特尔伯格和谢尔盖·雅克宁。

My guests today are Max Yatterberg and Sergei Yaknin.

马克斯是Isomorphic Labs的首席人工智能官，谢尔盖是首席技术官。

Max is the chief AI officer at Isomorphic Labs, and Sergei is the chief technology officer.

Isomorphic Labs是一家致力于打造世界领先的人工智能药物设计引擎的公司，旨在变革药物研发，开启生物医学突破的新纪元。

Isomorphic Labs is the company building a world leading AI drug design engine to transform drug discovery and usher in a new era of biomedical breakthroughs.

我现在与马克斯和谢尔盖在圣何塞举行的GTC 25现场连线。

I'm here with Max and Sergei live at GTC 25 in San Jose.

两位先生，感谢你们在繁忙的一周中抽出时间参加我们的播客。

Gentlemen, thanks for taking time out of the week to join the podcast.

不。

No.

谢谢。

Thank you.

很高兴能来这里。

It's great to be here.

是的，很荣幸。

Yeah, it's a pleasure.

那么，你们能否先简单介绍一下自己、你们的背景，以及你们是如何来到Isomorphic Labs的？

So why don't we start with a little bit, if you would, about yourselves, your background, and how you wound up at Isomorphic Labs?

Max，你先说吗？

You want to go, Max?

好啊，当然。

Yeah, sure.

我已经在人工智能领域工作了大约十五年。

So I've actually been in field of AI for about fifteen years now.

这可是在人工智能还没火起来之前的事了。

This is long before it was cool.

那时候，我主要做计算机视觉，研究如何在早期的AlexNet基础上应用于ImageNet，如何扩展用于物体识别和文本识别。

Back then, I was working a lot on computer vision, so applying early days of AlexNet, how do we use this for ImageNet, how do we scale up for object recognition, text recognition.

读博士期间，我做出了当时世界上最好的图像识别模型。

During my PhD, had the best image recognition models in the world.

实际上，我们当时在这一领域创办了一家公司，最终被谷歌收购并并入DeepMind。

Actually, we had a company in this space that ultimately got acquired by Google to join DeepMind.

这就是我开始与丹尼斯合作的原因。

That's how I ended up starting to work with Dennis.

明白了。

Got you.

丹尼斯在DeepMind待了很长时间，研究生成模型的早期阶段，后来迷上了强化学习。

Dennis Asavas Spent a long time at DeepMind working on early days of generative models and then called the reinforcement learning bug.

他深入研究了AI在游戏领域的挑战性问题，试图击败《星际争霸》中的顶尖职业选手。

Worked a lot on these challenge domains of AI for games, trying to beat the top professionals at StarCraft

还有

and

去吧，所有这些之类的东西。

Go, all this sort of stuff.

但真正核心的是，深度学习是我热爱的东西。

But really at the core, deep learning was the thing I loved.

这是一种了不起的技术。

It's this amazing technology.

我希望看到它对世界产生真正根本性的影响，对人类带来积极影响。

And I want to see it have real fundamental impact on the world and positive impact on humanity.

因此，当我们开始看到AlphaFold的进展时，就出现了一个绝佳的机会来创立一家新公司——Isomorphic Labs。

And so as we started to see AlphaFold unfolding, there was this incredible opportunity to create this new company, Isomorphic Labs.

这是一个绝佳的机会，可以运用我们多年来开发的这套强大的深度学习和机器学习工具箱，真正尝试变革药物设计领域。

And this was just such a brilliant opportunity to apply all of this amazing deep learning and machine learning toolbox that we've been developing over time to really try and transform the drug design space.

这正是促使我加入Isomorphic并领导这一领域人工智能工作的原因。

That's what brought me over to isomorphic to really head up AI in this space.

对，所以这正是你所热爱的技术与持续推动前沿的交汇点，同时将其应用于对你有意义的事情上。

Right, so to hit that cross of the technology you love so much and continuing to push the frontiers, but applying it to something that's meaningful to you.

没错，没错。

Exactly, exactly.

我一直以来都既喜欢人工智能的应用，也喜欢它的基础理论。

And I've always loved the application of AI as well as the fundamentals.

看到它产生如此积极的影响，这对我来说至关重要。

And just to see it have that really positive impact, that's really key for me.

当然，这太棒了。

Absolutely, that's great.

谢尔盖？

Sergei?

是的，我也早在它还没流行的时候就投身其中了。

Yeah, I've also been at this since way before it was cool.

我想借用马克斯的话，我当时住在多伦多，就读于多伦多大学。

I want to say to borrow from Max's words, I was living in Toronto attending U of T.

我们有个叫杰夫·辛顿的人，可以说是这领域的奠基人之一，你可能听说过他。

We had this guy, Jeff Hinton, kind of Some godfather of may have heard of him.

那真的很棒。

It was really great.

我的意思是，以前在多伦多大学教课，我们都学习过神经网络。

I mean, used to teach courses at U of T and we all learned about neural networks.

但那时候的神经网络还不是深度神经网络。

And back then these were not deep neural networks.

那些只是非常浅层的几层网络，大家长期都在尝试解决手写数字识别问题。

These were very shallow couple of layers and everybody was trying to solve handwritten digit recognition problems for a very long time.

我进入行业后，在多家不同的科技公司工作，涉及金融科技或电信等行业。

I went into the industry and worked in a number of different technology companies, working in industries like fintech or telecommunications.

后来我曾在亚马逊工作了一段时间，负责建立他们的加拿大软件工程团队。

Ended up working at Amazon for a little while building their Canadian software engineering organization.

在行业工作了大约十年或十二年后，我开始对医疗健康领域产生了浓厚兴趣。

After about ten or twelve years working in the industry, I sort of got really interested in healthcare.

那是在2012年左右，2013年左右。

This was around twenty twelve, twenty thirteen.

人们开始常规性地生成DNA测序数据，我意识到人们在分析这些数据集时遇到了很大困难，因为它们非常庞大。

People started generating DNA sequencing data as a routine, and I realized that folks were really struggling to actually analyze these data sets because they were quite large.

于是，我抓住机会加入了一家癌症研究机构，帮助他们推进一项雄心勃勃的全球项目：统一全球的癌症DNA测序数据，通过云端统一处理，并向全球研究人员开放，让每个人都能利用这些数据。

And so I took the opportunity to join a cancer research institute to help them with this really ambitious global project of unifying the world's cancer DNA sequencing data and processing it uniformly on the cloud and making it available to researchers worldwide so that everybody could make use of this data.

对我来说，这让我看到了技术如何真正推动医疗的发展。

And to me, this kind of opened the door to how technology actually enables healthcare.

这让我走上了一条全新的道路：在参与这个项目一段时间后，我决定攻读该领域的博士学位，于是来到欧洲，在德国的欧洲分子生物学实验室攻读博士。我当时想，作为一名不懂生物学的计算机科学家，我应该让自己置身于一个我最不了解的环境中。

And this set me down this whole path where after working on this project for a little while, I decided to do a PhD in the field, came to Europe, to Germany to do my PhD at the European Molecular Biology Laboratory, I kind of thought, you know, as a computer scientist who doesn't really understand biology, I should just immerse myself in an environment where I knew the least about what was going on.

因此，这段博士生涯成为了一段令人惊叹的开悟之旅，让我深入了解了分子生物学。

And so, you know, it was this incredible eye opening journey through this PhD learning about molecular biology.

从这段经历走出来后，我决定立即重返产业界。

Coming out of the other side of this, I actually decided to jump right back into the industry.

我只是觉得，这里才是创新速度最快的地方。

Just felt like this is where velocity was highest.

于是，我加入了一家名为Sophia Genetics的公司，我们从事分子诊断工作，即对癌症患者的DNA进行测序，以更好地帮助诊断他们所患癌症的类型。

And so I joined a company called Sophia Genetics where we did molecular diagnostics, which is sequencing DNA of cancer patients to try to better help diagnose the types of cancer that they had.

我们花了很大力气打造这家公司，但我在2021年左右注意到，利用人工智能进行药物发现的兴趣开始显著上升，先是不同的制药公司和硅谷初创企业开始谈论这项技术。

And we had a great time building out this company, but I noticed right around sort of 2021, 2022, there started to be this real uptick in the interest of using AI in drug discovery and it first started bubbling up with different pharma companies and different Silicon Valley startups sort of starting to speak about using this technology.

然后我听到了丹尼斯团队正在筹备Isomorphic Labs的消息。

And then I heard from Dennis's team who were looking to start isomorphic labs.

我完全被从零开始打造这家公司的想法所吸引。

And I was just completely enthralled by the idea of building this company from the ground up.

这还是Iso Labs正式公布之前的事。

This was before Iso Labs was even announced.

因此，这是一个机会，让我能与马克斯和其他几位同仁一起成为这个组织的创始团队成员。

And so it was an opportunity to be part of this founding team of this organization together with Max and a couple of other folks.

当我了解到所有让这家公司成为可能的因素后，我几乎立刻就做出了决定：是的，我当然想参与这个项目。

And as soon as I heard about all of the different factors that were going make this company possible, it just became a really easy decision to say, yeah, sure, I want to work on this.

我想全身心投入，给这个项目一个机会，从此义无反顾。

I want to really spend my time to give this a shot and never look back.

太棒了。

Amazing.

我们来聊聊这家公司吧。

So let's talk about the company a little bit.

Isomorphic Lab的使命是用AI优先的方法，从第一性原理重新思考药物发现。

The Isomorphic Lab's mission is to reimagine drug discovery from first principles with an AI first approach.

你们打算挑战传统药物发现中的哪些基本假设？

What fundamental assumptions about traditional drug discovery are you looking to challenge?

我认为传统药物发现中一个关键的问题是，你会在某个特定的药物设计问题上投入极大的努力。

I think one of the key things that you see in traditional drug discovery is you work really, really hard on a particular drug design problem.

你针对某种特定疾病、特定适应症，这意味着你通常会锁定某个特定的蛋白质作为药物作用的靶点。

So you're going after a particular disease, particular indication, that means probably a particular protein, the target that you want to try and modulate with your drug.

团队会投入巨大精力去解决这一个具体问题。

And teams work really, really hard on solving that one problem.

而这个设计过程可能需要数年时间。

And that design process can take years.

这非常困难。

It's really difficult.

然后你就得进入临床阶段。

Then you have to go into the clinic.

但当你完成这部分设计工作后，这些成果无法迁移到下一个药物设计问题上。

But after you've done that piece of design work, none of that transfers to the next drug design problem.

所以你面对的是下一个疾病、下一个靶点、下一个蛋白质。

So you've got the next disease, the next target, the next protein.

而在这里，我们可以借助机器学习和人工智能，以非常不同的方式来思考问题，因为我们正在构建能够覆盖整个蛋白质组、整个蛋白质宇宙、整个化学领域的通用模型，比如AlphaFold，这意味着我们拥有能够针对某一靶点、某一蛋白质、某一疾病的技术和方法，并且这些方法同样适用于其他情况。

And this is where we can think about things very, very differently with machine learning, with AI, because we're building models, models like AlphaFold that generalize across the whole of the proteome, the whole of the protein universe, the whole of chemistry, which means that we have techniques and we can think of ways of doing drug design against these models that work for one target, one protein, one disease over here.

同样的模型、同样的技术，也可以用于解决另一个问题。

And exactly the same models, the same techniques can work for another one over here.

因此，我们得到了一个高度可泛化的药物设计引擎。

And so we get a very generalizable drug design engine.

这甚至改变了公司的DNA。

And so then that changes even the DNA of the company.

我们不再仅仅关注单一疾病，甚至不再局限于单一疾病领域，而是真正思考如何解决所有疾病，以及如何围绕这一目标构建一家公司。

Instead of just thinking about a single disease, even a single disease area, really think about how can we be solving all disease and how can we build a company around that.

也许还有一个需要讨论的假设，就是你必须不断进行湿实验，以让自己扎根于某种可测量的科学现实。

Maybe one more assumption to talk about is this assumption that you need to continuously do wet lab experiments to ground yourself in some measurable scientific reality.

当前的传统药物设计依赖于设计-合成-测试循环，作为化学家，你会假设：如果我合成这个特定分子，它可能会与某个蛋白靶点结合，并具备这些特性。

Of current conventional drug design relies on these design make test cycles where as a chemist, you're going to hypothesize that, well, if I take this particular molecule, it might bind to a protein target, it might have these properties.

然后我会合成它并验证这个假设。

And then I'm going to make it and test that assumption.

但大多数时候，我会发现它实际上并不具备这些特性。

And most of the time I will learn that actually it doesn't have those properties.

如果我们真的能很好地预测这类性质，那这个问题现在就已经解决了。

If we were really great at predicting properties like that, well, would be a solved problem right now.

因此，当我们拥有像马克斯所说的、与实验结果一样可靠的通用模型时，我们可以将它们结合起来运行，并直接将输出结果视为真理，这是一种极其解放药物设计方式。

And so what we can do when we have these general models, as Max says, that are as good as experiment, we can actually run many of them in combination and we can just take the output as truth, which is an incredibly liberating way to do drug design.

我们的药物设计师经常提到，他们现在可以对这些分子做出更大、更大胆的改动，而这在传统方式下是无法实现的，因为你对预测结果缺乏信心。

Our drug designers often talk about the fact that they can make bigger and bigger and bolder changes to these molecules, which is just not how you operate conventionally since you don't have a lot of confidence in the predictions you're making.

传统上，你只能做微小的改动，因为你不想毁掉已经构建起来的东西。

Conventionally, you make small changes because you don't want to destroy what you've already built.

你正在试图一次又一次地做出微小的增量改变。

You're trying to eke out little incremental changes one by one.

而当你能信任这些模型的预测时，你就可以自由地设计了。

Whereas when you can trust these models' predictions, you just design freely.

这非常棒，并引导我们明确了公司的北极星目标：我们希望将模型推进到可以仅通过一轮设计就完成整个药物设计流程的阶段——即全程在计算机中进行设计，仅在最后一步通过实验进行验证。

This has been really great and leads us to our North Star as a company where we're aiming to get our models to the stage where we can run an entire drug design program in this single design round where we basically continue designing in silico and only validate once at the very end using experiment.

但这里有个疯狂的地方。

But here's the crazy thing.

即使我们创造了这些完美预测实验结果的惊人模型，我们已经有了不少重大突破。

So even if we create these amazing models that are perfectly predictive of this experimental technique, we so have this perfect model, we've had some significant breakthroughs in this.

然而，潜在药物的空间大约是10的60次方。

Still, the space of potential drugs is something like 10 to the 60.

这是一个巨大的空间，即使你拥有一个完美预测的模型，也无法穷尽搜索每一个设计。

That's such a huge space that even if you have this perfectly predictive model, you can't exhaustively search every single design.

因此，你还需要比这些预测模型更进一步的东西。

So you need something more even than just these predictive models.

你需要真正的生成模型或搜索代理，能够实际探索化学空间并提出这些设计，从而应对这种高达10的60次方的巨大空间。

You need actual generative models or ways to search agents that can actually explore chemical space and come up with these designs to be able to navigate this massive 10 to the 60 space.

这与传统生物技术或制药行业早期筛选先导化合物的方式完全不同；在那里，你只是取一个分子库，可能是百万级，幸运的话达到千万级，甚至十亿级，然后通过实验进行筛选，但这仅仅触及了10的60次方空间的表面。

And that's completely different to the way that traditional biotech or pharma would do this early hit discovery, where there, what you're doing is you're taking a library of molecules, you know, maybe it's a million, maybe it's 10,000,000, if you're lucky, it's even to the billion scale, and you're screening them experimentally, that barely chips away at the surface of that 10 to 60 space.

而这正是我们能够开始实现的、对整个化学空间进行全面探索的巨大潜力所在。

And that's really the difference in potential, exploration of the full chemical space that we can start to do.

你能描述或解释一下，当你谈到生物学是一个信息处理系统时，这究竟意味着什么吗？

Can you describe or kind of explain what it means when you talk about biology as an information processing system?

请深入解析一下这个概念，然后结合你构建模型的方法来说明。

Kind of unpack that but then relative to your approach to model building?

是的，当我们思考细胞时，细胞处于一个充满各种化学信号和挑战的环境中，细胞需要处理这些信息并做出反应。

Yeah, when we think about cells, cells exist in this environment that throws information via different chemical signals at them, different challenges and the cell needs to be able to process that and do something with it.

细胞的目标是生存和增殖。

Cell is trying to survive and is trying to proliferate.

因此，我们看到细胞的行为方式及其利用自身工具的方式，与我们所做的事情有很多相似之处。

And so we see a lot of parallels between how cells act and they use the tools that they have at their disposal.

细胞所拥有的工具库实际上都编码在其基因组中。

The arsenal that the cell has at its disposal is really encoded in its genome.

细胞所能做的一切都编码在这个基因组里，细胞随后会制造蛋白质来实现它所需的功能。

Anything a cell can do is encoded in this genome, which then the cell makes proteins to actually do what it needs to do.

因此，我们认为这是一个绝佳的机会，可以真正建模这些细胞过程，深入理解细胞如何应对其环境，并在机器学习模型中复现这些机制，从而在AI模拟的细胞中实现体外模拟。

And so we see this as an ultimate opportunity for us to really model those cellular processes, to really understand how the cell deals with its environment and be able to replicate some of this in a machine learning model that would allow us to do that in silico in this AI analog of a cell.

当这一切最终实现，未来愿景完全落地时，会是什么样子？

So when this is all realized, the vision kind of down the road comes to full fruition, What does it look like?

在您所描述的这种全新药物发现范式下，我们的世界会变成什么样？

What does our world look like in this new paradigm of drug discovery that you've just been detailing?

它将如何影响我们照顾自身的方式、医学、医疗保健、预防性护理，方方面面？

How does it affect the way that, you know, we take care of ourselves, medicine, healthcare, preventative care, everything.

这个愿景具体是什么样子？

What does the vision look like?

让我试着描绘一个愿景，在我看来，这个过程实际上会分阶段进行，我们现在还处于早期阶段，仍在努力摒弃旧一代的工具——正如马克斯提到的那些不可扩展的流程，并构建真正技术驱动的方法。

Well let me try to paint one vision and in my mind it's actually there's going to be stages of this in some sense and we're at an early stage where we're still trying to kind of shed some of the older generation tools, you know, these non scalable processes as Max has talked about, and build a real technology approach.

技术是可扩展的。

Technology is scalable.

这正是它颠覆众多行业的方式。

This is what it's done to disrupt so many industries.

因此，第一步是探讨如何构建这些模型，使我们能够实现可扩展性，采用通用方法——我不再需要一支针对特定疾病的庞大科学家团队，而是拥有一个通用模型，让我能够做到这一点。

And so this first step is about how can we build these models that will bring scalability to this, that will bring these general approaches where I don't have to have a whole army of disease specific scientists to focus on a particular disease instead I have a general model that allows me to do this.

因此，在我看来，第一阶段就是我们如何以技术领先的方式实现这一点？

And so to me, stage one is just how do we do this in a tech forward way?

但更进一步的阶段将使我们更接近于：如何以一种能够说服监管机构的方式实现这一目标，让他们相信这些模型能够以极高的准确率预测结果，也许未来我们不再需要花费五到七年进行临床试验，因为我们能从数学上证明我们设计的分子一定会有效。

But further stages actually get us much closer to how can we do this in a way that will start convincing regulatory bodies that these models are predicting things with such a high degree of accuracy that maybe we don't need to spend five or seven years in clinical trials in the future because we can prove mathematically that the molecule we design is going to work.

因此，这是一个非常重要的下一阶段，因为我认为这将彻底改变整个行业的工作方式。

And so that's a really important next stage because I think it's going to really change how the entire industry works.

当我们现在思考这个行业时，FDA平均每年批准约50种新药，这个数字可能已经维持了三十年之久，因此，即使我们拥有极其强大的模型，能够设计出更多药物，目前仍存在这些限制。

When we think about this industry now, the FDA approves about 50 drugs per year on average and so it's been that number for probably three decades And so there's a real limitation to how many drugs, even if we had incredibly powerful models that would allow us to design many, many more drugs, we currently have these limitations.

因此，作为人类，共同克服这一限制将成为一项关键成就。

And so being able to overcome that together as humanity is going to be a key accomplishment.

但这开启了一个我们经常谈论却远未实现的精准医疗未来：我们应当能够极其精确地诊断出一个人体内的状况。

But what this opens up is this future of a precision medicine that we talk about often, but actually are nowhere near today, where we should be able to very, very precisely diagnose what is going on with somebody.

想象一下，他们患有一种特定类型的癌症。

Imagine they have a particular kind of cancer.

癌症有一百万种不同类型，取决于分子特征，取决于患者基因组内的突变。

There's a million different types of cancer, depending on the molecular signature, depending on the mutations within the genome of the patient.

然后我们应当能够设计出一种高度定制的化合物组合，最适合该患者或这一小群患者。

And we should be able to then design a very bespoke combination of compounds that would be best for that patient or for that very small group of patients.

因此，我们需要遵循这些阶段才能达到目标。

And so we need to follow kind of the stages to get there.

但对我来说，这是路上的下一个里程碑。

But to me, that's the next waypoint on the road.

理想情况下，最后一个里程碑是，我们不应等到人们生病后再去更好地治疗他们，而应提前预防疾病。

Then ideally the last waypoint is rather than waiting for somebody to get sick so that we can cure them better, we should be getting ahead of that disease.

因此，我们应当努力在健康个体体内的某种生物标志物开始偏离正常方向时就检测到，并设计干预措施，从一开始就防止他们生病。

So we should be actually trying to detect when a certain biomarker in a healthy individual starts going in the wrong direction so that we can design interventions that will prevent them from getting sick in the first place.

我认为这将彻底改变这个行业。

And I think that will completely change this industry.

是的，绝对如此。

Yeah, absolutely.

你所采取的这种方法，我想知道传统势力如何看待你，以及他们对这种做法的反应。

The approach that you're taking, I'm wondering kind of how the old guard, you know, looks at you and kind of responds to this approach.

同时也想知道这如何影响人才招聘和团队建设。

And then also wondering how it shapes recruiting and team building.

当你招募人员加入团队时，开放的心态、愿意换一个角度看问题，这些有多重要？

And if when you're, you know, bringing people onto the team, how much of a as an open mindset, a willingness to look through the glass the other way, what have you.

在组建团队时，这一点在你的策略中占多大比重？

How much does that go into your approach to building out your team?

是的，关于人工智能在药物设计中的应用，我们实际上已经看到第一批公司和制药公司开始涉足这个领域，时间大概在五到七年前。

Yeah, so AI for drug design, we've actually seen maybe a first wave of companies and pharma companies actually start to dip their toes into this space for maybe five, six, seven years now.

但我认为这与我们所构建的属于不同的浪潮。

But I think that's a slightly different wave to what we've been building.

这一波人工智能在药物设计中的应用，本质上是探讨如何将这些机器学习工具融入传统的药物设计流程？

This first wave of AI for drug design has really been, how do we just use some of these machine learning tools in the traditional drug design processes?

这通常表现为构建针对特定靶点的本地模型——你拥有一些数据，并针对这些数据拟合小型本地模型，以帮助指导下一轮实验设计和测试。

And that often comes out as building these local models where you're doing drug design around a particular target, you've got some data and you're fitting small local models to that to help inform you on that next experimental design make test round.

这与我们所构建的模式截然不同，我们的模式是创建非常通用的模型，这些模型实际上可以应用于空间中的任何不同部分。

And that's a very, very different paradigm to what we have been building, which is let's create very general models, models that we actually can apply to any different part of space.

而且，我想，在我们刚开始的时候，很多人并不认为这是正确的方向。

And, you know, a lot of people, I think, at least when we were starting out, didn't think that was probably the right way to go.

但我们一再看到，我们确实能够构建这些通用模型，不仅限于像AlphaFold这样的模型，还包括许多其他预测能力和生成能力。

But what we've just seen again and again is we can actually build these general models, not just the AlphaFalls of this world, but many other sort of predictive capabilities, generative capabilities.

我们看到了将这些模型应用于行业中最棘手问题的潜力。

We see the potential to apply these on literally the hardest problems in the industry.

例如，我们与诺华的合作，我想这并不是什么秘密，他们把行业中一些最困难的靶点交给了我们。

So for example, our collaboration with Novartis, I think it's no secret that they've thrown some of the hardest targets in the industry to us.

这些都是一些人们已经研究了十年之久的药物设计难题。

These are sort of drug design problems that people have spent ten years there.

你能举个例子吗？

Can you give an example?

我们目前无法给出确切的例子，但这些确实是他们已经研究了十多年却毫无进展的靶点。

We can't give an exact example at the moment, but these are genuinely targets they've been working on for over ten years, not making progress.

化学家们找到我们的团队，说这些靶点根本不可能攻克。

The targets that chemists come up to our leads and say, this is impossible.

别试了。

Don't try it.

但与此同时，我们仅用几个月的时间就在这些靶点上取得了突破，创造了全新的化学物质，找到了调节这些靶点的全新方法，这让这些化学家们震惊不已。

And at the same time, we see that in just months, we're able to make traction on this creating novel chemical matter, finding novel ways to modulate these targets, which are literally blowing the minds of these chemists.

所以我认为，整个行业要真正理解这里发生的一切，还有很长的路要走。

So I think there's still a long road for the whole of the industry to really fully understand what's happening here.

这种理解正在发生。

This understanding is happening.

在我看来，五年后，人工智能在药物设计中的应用将遍及整个行业。

The way I see it in five years' time, AI for drug design is going be across the entire industry.

没有人工智能的药物设计，就像试图在没有数学的情况下进行任何科学工作。

Doing drug design without AI is going to be like trying to do any type of science without maths.

它将成为科学的基本组成部分，尤其是这一领域的科学。

It's just going be a fundamental part of science and particularly this science.

我听着你们两位谈论这个话题。

And I'm listening to the two of

你们在谈论它。

you talk about it.

从某种角度看，这听起来很简单，但这种方式却蕴含了你们所提到的所有力量。

It sounds so simple in a way, but in a way that brings all the power that you're talking about behind it.

但我想，如果你已经以另一种方式做了五十年的职业生涯，或者类似的情况。

But I would imagine if you've been doing it the other way for, you know, a fifty year career or what have you.

我不知道，仅仅是心理上的抵触可能就会成为一个问题。

I don't know, just the mental resistance to it might be an issue.

是的。

Yeah.

而且别误会。

And and don't get me wrong.

说这些话听起来真的很简单。

It sounds really easy to say these things.

但实际上，这些都是非常非常困难的建模问题。

In reality, these are really, really hard modeling problems.

这些都是所谓的圣杯级建模问题，人们已经研究了几十年。

These are holy grail modeling problems, things that people have been working on for decades.

对。

Right.

但我在想，人们对于改变做事方式的抵触情绪。

But I mean, I'm imagining just the resistance that people have to change new ways of doing things.

如果你和你的团队长期以来一直致力于解决这些问题，突然出现了一种新技术，能在极短时间内解决它们。

If you've been you and your team have collectively been pounding on these problems and then there's this new technology that solves it in such a short time.

我认为，各方都存在一种健康的怀疑态度。

There's a healthy degree of skepticism going all around I want to say.

我认为，如果你身处这些制药公司之一，这实际上可能非常具有挑战性，因为它们已经根据传统的药物设计方式形成了特定的组织结构。

I think actually it can be really challenging if you're inside one of these pharma companies because they have coalesced on a particular structure as a result of how you would normally do drug design.

它们通常会按疾病领域进行组织。

They will organize themselves often by disease.

你会有一个专门专注于肿瘤学的部门，另一个则专注于眼科。

You'll have a whole area that is just focusing on oncology and another that is focusing on ophthalmology.

而我们所开发的这些模型，却跨越了所有这些领域。

And so you have these whole structures and the types of models that we're creating, they span across all of this.

所以，想象一下，在一个按这些疾病领域划分的公司内部，如何启动这样的倡议。

And so imagine trying to nucleate this kind of initiative inside a company that is organized by these disease areas.

你该把它们放在哪里？

Where would you put them?

如果把它们放在外面，它们又如何真正渗透进这个结构并推动变革？

And if you put them outside, how would they actually permeate that structure and be able to institute change?

因此，我认为大型制药公司内部存在相当大的障碍，而这在某种程度上对ISO来说反而是好事。

So I think there's significant barriers within big pharma itself, which in a way is great for ISO.

但真正令人鼓舞的是，看到一些加入ISO本身的化学家和生物学家眼中发生的转变，其中许多人最初带着相当程度的怀疑加入。

But what's been really heartening actually is seeing some of the transformation in the eyes of some of the chemists and biologists that have joined ISO itself, many of whom have also joined with a healthy degree of skepticism.

随着时间推移，我们通过数据和这些模型切实有效的证据，说服他们接受了这些方法。

Over time, we've convinced with data, with proof of the working of these models to embrace these.

我们有一个称为‘ISO方式’的东西，本质上就是在这种AI优先的模式下如何进行药物设计。

And we have something that we call the ISO way, which is essentially how do you do drug design in this AI first approach?

我们的所有化学家都参与了这一浪潮，与构建技术的科学家和工程师们共同创造这种‘ISO方式’。

And all of our chemists are part of this wave of how do we actually invent this ISO way together with all of the scientists and engineers that are building the technology.

这带来了巨大的变革。

And this has been really transformative.

某种程度上，我秉持着一种经典的亚马逊原则：从客户出发逆向思考，你需要真正理解你的客户是谁，然后从那里逆向推导，这将指导你的产品设计。

In some sense, I have this kind of old school Amazon principle of working from your customer backwards, where you want to really understand who your customer is, and then you want to work backwards from that, and that will guide you in your product design.

而在这里，我们拥有一个理想的情境：客户就是我们的同事，那些在Isomorphic Labs与我们并肩工作的伙伴们。

And we have this dream situation here where actually the customer, our peers, our buddies that are working together with us at isomorphic labs.

因此，与之相反，ISO从设计之初就将所有团队深度融合在一起。

And so at ISO, by contrast, by construction, we have actually meshed all of our teams.

我们所有团队都坐在同一个开放空间里。

All of our teams are sitting together in common spaces.

并不是技术团队坐在这里，化学团队坐在那里。

It's not like a technology team sits over here, a chemistry team sits over here.

我们交错而坐，紧密协作地推进这些项目。

We're sitting interspersed and we're working very collaboratively on these projects.

这双向受益，因为我们从这些长期从事本领域的专家那里获得了宝贵的价值和洞见，他们帮助我们确保模型真正扎根于他们的专业知识。

That cuts both ways in the sense that we're getting incredible value and insight from these long term career domain experts that are helping us make sure that our models are really grounded in the knowledge of their trade.

但同样，在进行药物发现项目时，我们也以这种技术驱动的方式进行。

But similarly, when we're doing drug discovery projects, we are doing that in this tech forward way.

我觉得这种大熔炉极大地改变了化学家们的工作方式，正在帮助塑造下一代化学——以一种完全新颖的方式进行药物设计，而这一切每天都在Isomorphic Labs被发明出来。

So I feel like this melting pot has been really amazing in transforming how chemists do their jobs and are helping evolve this next generation of chemistry that is doing drug design in a completely new way and it's being invented at ISO on a daily basis.

我们的嘉宾是来自Isomorphic Labs的马克斯·亚特伯格和谢尔盖·雅克宁，我们正在讨论他们革命性的药物设计方法、构建世界模型以及采取技术优先的路径。

Our guests are Max Yatteberg and Sergey Yaknin from Isomorphic Labs and we're talking about their really revolutionary approach to drug design, building world models and kind of taking a tech forward approach.

但谢尔盖，正如你所说，协作才是其核心所在。

But Sergei, as you were saying, the collaboration, you know, really at the core of it.

但我想问问你关于技术的事情，特别是AlphaFold。

But I want to ask you about the technology and ask you about AlphaFold.

AlphaFold 3是当前的版本。

AlphaFold three is the current version.

据我了解，它在预测生物分子结构方面是一个重大突破。

And in my understanding, it was a big breakthrough in predicting biomolecular structures.

你能谈谈吗？甚至可以回溯到最初，简单解释一下AlphaFold是什么，然后讲讲我们是如何发展到AlphaFold 3的，以及它有多重要？

Can you talk about and even go back to the beginning and kind of maybe just briefly explain what AlphaFold is, and then talk about how we got to AlphaFold three and how important it is?

是的。

Yes.

所以，如果回到过去，AlphaFold最初是DeepMind的一次黑客马拉松项目。

So maybe going back in time, AlphaFold actually started as a hackathon project in DeepMind.

那是一个为期两周的黑客马拉松项目。

This was like a two week hackathon project.

我们能用神经网络来预测蛋白质结构吗？

Can we throw a confnet on protein structure prediction?

但神奇的是，那里出现了生命的迹象。

And crazily, there was like signs of life there.

这逐渐发展成了AlphaFold项目。

And that snowballed into AlphaFold, the project.

AlphaFold一版在准确性上有了巨大提升。

AlphaFold one was a big step up in terms of accuracy.

但AlphaFold二版，也就是2020年的版本，首次实现了神经网络对蛋白质结构预测达到实验级别的精度。

But AlphaFold two, this was in 2020, was that first moment that people started to see experimental level accuracy of protein structure prediction from a neural network.

最终，AlphaFold2在去年荣获了诺贝尔奖。

And ultimately, AFFold2 went on to win the Nobel Prize just last year.

但AlphaFold2仅能预测蛋白质的结构，以及蛋白质与其他蛋白质相互作用时的结构。

But AFFold2 just predicts the structure of proteins and proteins coming into contact with other proteins.

但除了蛋白质之外，还有许多其他类型的生物分子，当我们考虑药物设计时，它们尤其重要。

But there's lots of other different types of biomolecules in addition to proteins that are particularly very important when we think about designing drugs.

蛋白质是分子机器的一部分，这些机器通过与其他蛋白质相互作用来工作，同时也与DNA、RNA和小分子等物质相互作用。

Proteins are part of these molecular machines that work by interacting with other proteins, but also things like DNA, RNA, small molecules.

这些小分子可能是你摄入的咖啡因，也可能是我们服用的药物。

And these small molecules could be things like like caffeine that you consume or it could be drugs that we consume as well.

所以，如果我们想研发一种药物，那么什么是药物呢？

And so really, if we want to create a drug, and what is a drug?

药物是一种能够介入并调节这些分子机器的物质。

A drug is something that comes in and modulates these molecular machines.

因此，我们希望真正能够理性地设计它。

And so we want to actually design that really rationally.

我们希望能够理解这种蛋白质与这些小分子，以及可能还有DNA共同作用时形成的分子机器的结构。

We want to be able to understand the structure of this protein with these small molecules, maybe also together with DNA as these molecular machines form.

这意味着我们需要超越AlphaFold2的全新能力，这也促使我们创造了它。

And so that meant we needed a completely new capability beyond AlphaFold2, and that led us to the creation.

这是与Isomorphic Labs和谷歌DeepMind合作完成的一项工作，即AlphaFold3。

This was a piece of work with isomorphic labs and Google DeepMind of AlphaFold3.

这对我们的研究来说是一个重大突破。

And this was a big breakthrough for us.

这是去年发布的，这是我们首次能够以前所未有的精度预测所有这些分子结合在一起的结构。

This this came out last year, and this was the first time that we could predict the structure of all of these molecules coming together at unprecedented accuracy.

现在，这一成果与其他模型结合，使我们的化学家能够对这些分子设计进行修改，并在短短一秒钟内看到结果。

And this now is the thing that together with other models allows our chemists to make changes to these molecule designs and literally in a second see the result of that.

这彻底改变了工作方式——传统上，如果你修改了分子设计并想了解结构如何变化，最快也需要数月才能获得结构数据。

That's a completely different way of working where traditionally, if you made a change to a molecule design and you wanted to see how that changed the structure, it would take literally months to get that structure back at best.

这完全改变了游戏规则。

That completely changes the game

当然，绝对如此。

for Absolutely, our sure.

我不想打断你，但我不想低估这一点：获得诺贝尔奖的是AlphaFold 3吗？

And and I didn't wanna interrupt you, but I don't wanna undersell the fact that AlphaFold it was AlphaFold three that won the Nobel Prize?

获得诺贝尔奖的是AlphaFold 2。

AlphaFold two won the Nobel Prize.

是的。

Yeah.

是的。

Yes.

好的。

Okay.

我不想打断你，但我也不能让这句话就这么过去。

I didn't wanna I didn't wanna cut in, but I didn't wanna let that go by.

那么，你们现在正在构建更多的模型吗？

So are you now building additional models?

你们在调整AlphaFold吗？

Are you tuning AlphaFold?

从这里开始，流程是怎样的？

What's the process like from here?

是的，我们的工作远远超出了AlphaFold 3的能力范围。

Yeah, we go quite a bit beyond the capabilities of AlphaFold three.

如果从整体药物设计的角度来看，要制造出一种最终能成为你从药店买到的药片的分子，你需要解决沿途的许多挑战。

If one thinks about the overall drug design problem, you need to solve quite a lot of challenges on the way to making a molecule that's going to be in the pill that you're going to buy from drugstore.

这不仅涉及确定分子如何与其目标相互作用，还涉及它在体内的实际行为。

And this has to do with ascertaining not only how is the molecule going to interact with its intended target, but actually how it's going to behave in the body.

我们希望这些分子能够顺利到达它们需要到达的地方。

We want these molecules to both make their way to where they need to.

我们希望它们能在体内停留足够长的时间，并安全地分解和排出体外。

We want them to stay around for as long as they need to, and we want them to safely break down and exit the body.

因此，在考虑药物设计时，我们需要同时解决这一系列挑战。

And so when we think about drug design, we need to solve this whole series of challenges and solve them simultaneously.

当我们思考需要构建的不同模型时，我们从这个非凡的结构预测问题开始，然后继续解决其他问题，比如预测我潜在的药物与该靶点结合的强度如何。

And so when we think about the different models that we need to build, we start from this incredible structure prediction problem and then we go on to solving other problems such as predicting if my potential drug binds to this target, how strong will it bind?

我希望它结合得足够强，但又不能太强，最好不是永久性的。

I want it to bind strongly, but not too strongly, potentially not forever.

然后我希望它能够进入细胞，并在需要时顺利排出。

And then I want it to be able to go inside a cell and I want it to be able to exit when it needs.

因此，我们所有的模型本质上都在解决这一系列广泛的问题。

And so all of our models are basically solving this whole wide variety of problems.

然后，对我们来说，一个关键挑战是如何让所有这些部分在概念上协同工作，从而构建一个完整的药物设计引擎，实现端到端的流程。

And then a key challenge for us is to actually how to make all of them work in concept altogether so that we can have this holistic drug design engine that allows us to do this end to end.

基础设施和计算能力在这个过程中扮演了多大的角色？

How big of a part does infrastructure and compute for that matter play in this whole process?

在这个过程中，你获得合适基础设施和计算资源的途径，是否帮助你比以往更快地推进？我假设这确实加速了整个过程的发展，你能谈谈基础设施的重要性吗？

And you know along the way how is your access to the right kinds of infrastructure and compute you know help you advance I'm assuming faster than you would have otherwise but you know help the whole process evolve like can you talk a little bit about the importance of the infrastructure?

说实话，这带来了巨大的影响，在某种程度上，这决定了一家公司能有多大的雄心。

Yeah, it makes a huge difference to be honest with you and in some sense this really sets the stage for how ambitious you can be as a company.

在当前人工智能无处不在的世界里，对资源、图形加速器的竞争非常激烈，如果你无法真正获得或有效掌控这些大型机器学习集群，你最多只能使用别人创建的模型，或许能根据你的特定用例进行调整，但你无法开展基础性研究。

In the current world of AI being everywhere, there's huge contention for resources, for graphics accelerators, and if you're not able to really have access to or be effective at commandeering these large machine learning fleets, you're at best going to be able to make use of models somebody else created, and you may be able to adjust these models to your particular use case, but you're not going to be able to do foundational research.

Isomorphic Labs 引以为豪的是，真正推动了药物发现领域基础人工智能研究的进展。

What Isomorphic Labs prides itself on is actually really moving the needle on this foundational AI research for drug discovery.

我们能够创造出全新的、通用且可扩展的方法。

We're able to create completely new methods that are general and that are scalable.

而这依赖于对海量基础设施的访问能力。

And this is predicated on having access to massive amounts of infrastructure.

因此，我们非常幸运能与谷歌云合作。

And so we're very lucky to be partnering with Google Cloud on this.

我们与他们紧密合作，托管我们的整个平台，包括机器学习研究和生产推理平台。

And we work very closely with them on hosting our entire platform, our machine learning research and production inference platform.

因此，这是我们的工作中的重要组成部分。

And so this is a huge part of what we do.

我们如何确保这支庞大的机器学习研究和工程团队能够最大限度地减少阻力，实现最高的研究效率，以便他们能够尝试大量想法？

How do we make sure that this massive machine learning research and engineering team can have the least amount of resistance, the highest research velocity possible so that they can try lots and lots of ideas?

因为机器学习研究是一门非常实证的科学。

Because ML research is a very empirical science.

我们不断尝试和训练模型。

We end up trying and training models over and over again.

因此，能够获得这些硬件并高效地使用它们，至关重要。

And so access to this hardware and the ability to do it efficiently is really, really important.

其中一个令人兴奋的进展是，我们开始在模型开发中看到与之相似的扩展规律。

And one of the really exciting things is to start seeing that actually we're seeing really analogous scaling laws in our model development.

我们在大语言模型领域看到的这些现象，现在也开始出现在生物分子系统的建模和对生化世界的理解中。

The sort of things you might see in the LLM space we're starting to see in modeling biomolecular systems and understanding this biochemical world.

这真的非常令人兴奋。

And that's really, really exciting.

实际上，无论是训练方面，模型该有多大，

Both actually on the training side, just how large our models should be.

还是在推理方面，如何利用推理时的计算资源进行扩展，也都非常令人兴奋。

But also really excitingly on the inference side as well, how we can scale with inference time compute.

因此，通过增加计算资源，这些模型的准确性和有效性有着巨大的提升空间。

So there's a huge amount of opportunity to actually scale the accuracy and the efficacy of these models with the amount of compute.

对，没错。

Right, right.

那数据呢？

And what about data?

数据处理、合成数据，你们的数据策略是什么？

Data wrangling, synthetic data, what's your approach to data?

我不知道，也许最佳实践？

And I don't know, best practices maybe?

是的，天哪，我们简直可以开始聊数据了。

Yeah, gosh, we could just start talking about data.

但至少我们先开始吧。

But let's make a start at least.

数据对任何机器学习问题都非常重要。

Data is very important for any machine learning problem.

它在科学领域尤其重要，而且在科学中很难处理，原因很简单：你无法仅凭肉眼判断它是否合理、是否优质。

It is especially important in science and it is especially difficult to work with in science for the simple reason that you can't just eyeball it and verify yourself as a human that it makes sense, that it's any good.

我们知道，在传统的自然语言处理或图像领域，我们可以看着模型生成的图像说：不行，这太糟糕了。

You know we're very used to data in the traditional natural language processing or imaging space where I can look at an image that was generated by a model and be like, no, that's garbage.

哦，是的，这真棒。

Oh yeah, that's really great.

除非结构非常糟糕，否则你无法通过观察就判断出：哦，这是个很棒的结构。

You cannot look at a structure unless it's really terrible and be able to say, Oh yeah, that's a great structure.

因此，我们在处理这类数据时必须非常谨慎。

And so we need to be very careful with how we work with this type of data.

说实话，数据永远都不够用。

And well, there's never enough of it, truth be told.

当然，也有一些优秀的公开资源。

There are some great resources of course that are publicly available.

蛋白质数据银行使得AlphaFold成为可能。

Protein Data Bank made AlphaFold possible.

令人惊叹的是，这个资源几十年来一直向科学家开放，这充分证明了，只要拥有正确的算法进步，就能利用每个人长期以来都能获取的数据，创造出极其新颖而强大的系统，这真的非常重要。

The amazing thing about that is that's a resource that's been available to scientists for decades and there's a real proof point in that that actually with the right algorithmic advancements you can create incredibly new powerful systems using data everybody has had access to forever and so that's really really important.

但当我们思考整个药物设计领域以及我们需要建模的复杂系统时，想想看，这里有一系列不同层次的问题，到目前为止我们主要讨论了分子建模和化学方面，但人类所经历的疾病呢？

But when we think about the entire drug design landscape and the complexity of the systems that we need to model, if you think about, you know, there's this kind of stack of different problems and we've been talking so far a lot about the molecular modeling, the chemistry side of this, but diseases experienced by humans.

因此，当我们向上层推进时，我们不仅要思考分子，还要思考细胞，不仅要思考细胞，还要思考组织，不仅要思考组织，还要思考器官，然后是人类，甚至人类与周围环境的互动。

So when we go up the stack, we reason not just about molecules, but about cells, and not just about cells, but about tissues, and not just tissues, but organs, and then humans, and then even how humans interact with the environment around them.

因此，如果我们想全面解决这个问题，就需要在每一层堆栈上都考虑数据。

So if we want to solve this problem holistically, we need to be thinking about data at each layer of the stack.

而这些数据集在今天根本不存在。

And those data sets simply don't exist today.

因此，我们需要应对的一个大问题是：如何创建合适的数据集，并且严格遵循原则。

So a big problem that we need to contend with is how can we create the right data sets and be very very principled about it.

确实，业界流传着一种说法，认为制药公司拥有数据但分享得不够。

There is you know a bit of a meme out there around the fact that maybe pharma has the data and haven't shared enough.

我们始终欢迎来自合作伙伴或其他组织的更多数据共享。

And we will always welcome more sharing from our partners or from other organizations.

但事实是，许多这类数据并不适合机器学习使用，因为它们是在过去的药物发现项目或其他科学项目中生成的。

But the truth is a lot of that data is not machine learning ready in the sense that it has been generated as part of past projects that were drug discovery projects or other types of scientific projects.

这些数据缺乏足够的多样性，无法用于训练我们希望构建的通用模型。

And that data does not have the necessary diversity to actually train the types of general models that we're interested in creating.

因此，尽管这些数据可能对其他用途有意义，也可能有助于提升特定子领域的问题表现。

And so while that data might be interesting for other purposes, it might be interesting in improving performance on a particular subspace of the problem.

但大多数情况下，它们对这些通用模型的帮助有限，不如人们期望的那么大。

Most of the time it's not going to be helpful or as helpful as one might hope for these general models.

因此，我们在Isomorphic实验室面临的一个问题是：如何以一种有原则的方式生成这些数据集，以推动模型的持续进步。

So something that we contend with at isomorphic labs is how do we generate these data sets in a principled manner that will enable continued progress of the models.

好处在于，我们可以借助当前的模型来指导我们在庞大的数据空间中寻找方向，生成新的数据集。

The great thing about it is we can lean on our current generation models to help advise us where to look in this massive data space, where to generate new data sets.

因此，我们形成了一个非常良性的循环。

And so we have this really virtuous cycle.

如果你已经拥有一个顶尖的前沿模型，你就能获得关于下一步重点方向的绝佳建议。

If you have already an excellent state of the art model, you're going to get great state of the art advice on where to focus next.

这帮助我们通过持续生成数据、训练模型的新版本，不断巩固我们的优势。

And so this helps us continue actually improving on the edge that we have by continuing to generate data, train new versions of the model.

这确实是一个非常良性的循环。

And this is a really virtuous cycle.

是的。

Yeah.

你在对话中已经提到过这一点，但为了明确一下，药物发现的未来，是否将是端到端的AI？

You've talked about this a little bit in the conversation, but kind of to put a point on it, the future of drug discovery, is it AI end to end?

那是什么？

What is it?

我不该假设这些，我应该问你们。

I shouldn't be positing things, I should ask you guys.

你对未来有什么看法？你之前提到过，谢尔盖，你谈到了预防医学和精准医疗。

What do you see down the road and how will it ultimately and you talked about this a little bit, think Sergei you were talking about preventative care and precision medicine.

最终的目标是什么？我不知道，人类的终点会是什么？

What's the, I don't know, what's the end game for humanity?

这种全新的药物发现范式最终能帮助人类成为什么样的存在？

What is it that this new paradigm of drug discovery could ultimately help humans become?

当我思考我们的未来时，真正的北极星是能够在计算机上完成所有药物设计，也就是我们所说的“计算机模拟”，从而消除当前存在的实验瓶颈。

When I think about where we're going, it really is this north star of being able to do all of drug design on a computer, what we call in silico, and to remove that experimental bottleneck that currently exists.

每当你需要走出实验室，实际合成分子并进行测试时，都会增加大量的成本和时间。

Anytime you have to go out into the real world, actually make some molecules and then test it out, this just adds so much overhead, so much time.

最终，我们确实看到了一条可行的路径，能够真正解决许多这些问题，实现这些模型的实验精度，并创建能够探索这一领域的智能体和生成模型。

Ultimately, we do see a track to genuinely solving a lot of these problems, getting to experimental accuracy for some of these models, creating agents and generative models that can explore this space.

这带来的结果是，无论任何靶点出现，无论我们想要调控哪种蛋白质，我们都能非常迅速地设计出化学物质，如果送到实验室，就会看到积极的结果。

Where this leads is, okay, we can get any target coming our way, any protein that we want to start modulating, and we can very, very quickly come up with chemical matter that if we did go to a lab, we'd see positive results there.

在近期，随着我们沿着这条路径前进，这意味着我们能够开拓新的靶点。

Near And term, as we go along that path, what that does is it means that it opens up new targets.

这些长期以来的靶点，或者说是行业里真正极具挑战性的难题，人们常称之为难以攻克的靶点。

So these sort of age old targets or industry, really, really challenging things that people use the term intractable.

我们尽量避免使用这个词。

We we try and not use that term.

就称它们为具有挑战性的靶点。

Just call them challenging.

具有挑战性的靶点。

Challenging.

但确实，这打开了新的疾病领域，也开拓了过去人们无法触及的靶点，而现在我们能够做到了。

But, yeah, you know, opens up disease areas, opens up targets that previously people couldn't approach, but now we can.

是的。

Yeah.

我们开始看到这一点。

And we start seeing that.

当你越来越多地在计算机上进行体外工作，而不进入实验室，减少设计-制造-测试循环时，你的速度会更快。

When you're doing more and more of this work in silico on a computer without going into the lab and you're reducing those design make test cycles, you're doing things faster.

这意味着从我们决定要针对某个目标，或者发现存在迫切的患者需求，到真正开始进入临床试验的时间会大大缩短。

And so that means just the time it takes from, hey, we want to go after this or we really see that there's this patient need to actually starting to get into patients starts to reduce.

同时，随着我们对这些分子如何发挥作用、它们如何与靶点相互作用、这些靶点本身如何与整个网络以及构成疾病的全部通路相互作用，以及这些分子如何与人体相互作用、如何被吸收、如何被代谢，我们有了更深入的理解。

Then also as we understand much more about how these molecules work, how they interact with these targets, how these targets themselves interact with the whole network, all of these pathways that constitute disease, how these molecules interact with the body, where they get taken up, how they get broken down.

我们对这一机制的理解已经深入得多。

We know so much more about how this works.

关于它们的安全性特征。

The safety profiles of this.

但当我们真的进入人体并开始测试这些分子时，它们的安全性要高得多。

But when we do go into people and start testing these molecules out, they're much safer.

它们的疗效也更高。

They have higher efficacy.