跳到主要内容
Learn Motion
探索课程产品能力价格

下载 Learn Motion

用手机扫描二维码,下载 App。

App Store

扫码从 App Store 下载 Learn Motion打开商店页面

Google Play

扫码从 Google Play 下载 Learn Motion打开商店页面
登录
Learn Motion

How AI Designs a Drug Molecule from Scratch

1Why Designing a Drug Molecule Is Hard2Turning Molecules into Something a Machine Can Read3How Generative Models Propose New Molecules4Scoring and Filtering the Candidates5Testing, Learning, and Improving the Design
Why Designing a Drug Molecule Is Hard

The Catalogue Nobody Can Read

2 / 3
Two numbers carry this page. The small block is what chemistry has actually made and recorded — about one hundred million compounds. The enormous field behind it is the estimated drug-like space — around ten to the sixtieth. The point is not the exact figures but the ratio: the explored part is a speck. That is why you cannot search your way through it. Any method that only looks at compounds that already exist is stuck on that speck, and the molecule a new target needs may be somewhere far outside it.
0:00 / 0:00

A drug-like molecule is built from a small alphabet — mostly carbon, hydrogen, nitrogen, oxygen, and a few others — assembled into a framework with attached groups. Because the same atoms can be connected in enormously many ways, the count of possible molecules explodes very quickly.

Estimates of the number of molecules that could plausibly satisfy basic drug-like rules land around \(10^{60}\). The number of compounds that have ever been synthesized and recorded is roughly \(10^{8}\). The gap between those two numbers is the whole difficulty: the explored region is a rounding error next to the space that exists.

Put the ratio in perspective. If every recorded compound were a single grain of sand, the unexplored drug-like space would be more grains than there are atoms in a large fraction of the observable universe. No amount of faster testing closes a gap of that shape, because testing is linear and the space is not.

This is what makes drug discovery a search problem rather than a lookup problem. There is no index to consult. Any method that only examines molecules someone already made is confined to a tiny island, and the molecule that fits a new target well may sit somewhere in the ocean around it.

上一页2 / 3下一页

Learn Motion

生成课程,认真学会。

运营方:Wuhan Daoyin Technology Co., Ltd.

联系我们: [email protected]
隐私政策用户服务协议

© 2026 Learn Motion