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

AI Antibody Design vs Human Design: A Conceptual Overview

1Why Antibody Design Matters2How Humans Design Antibodies3How AI Learns to Design Antibodies4Comparing AI and Human Design5Evidence, Limits, and Open Questions
How AI Learns to Design Antibodies

Generating Candidate Sequences

2 / 4
Watch the model box on the left. It holds the learned patterns from the training data. As the animation runs, it emits candidate sequences one after another. Notice two things. First, the candidates are not copies of training sequences; they differ at particular positions, and those differences are the actual design proposals. Second, the candidates still look like antibodies, because the model only samples from the region of sequence space it learned. That is the strength: it stays where antibodies are viable. Now look at what the animation does not show. Nothing here tests binding to a target. The model is producing plausible antibodies, and plausible is not the same as functional. That gap is why the next stages exist.
0:00 / 0:00

A generative model for antibodies is trained to produce sequences that look like the ones it was trained on. In practice it learns a probability distribution over antibody sequences — roughly, which amino acids tend to appear at which positions and in which combinations — and then samples new sequences from that distribution.

The sampling step is what makes the output useful. If the model only reproduced training sequences exactly, it would add nothing over a database search. Instead it produces variants: sequences that follow the same statistical patterns but differ in specific positions. Those differences are the design proposals. Because the model was trained on antibodies that fold and function, its samples tend to stay in the region of sequence space where antibodies are viable, which is exactly the region that random mutation would mostly miss.

The limitation is equally direct. The model has learned what antibodies look like, not what binds your particular target. Its samples are plausible antibodies, and plausibility is not the same as binding. This is why generation is only the first stage: a model that proposes a thousand reasonable sequences has not told you which one, if any, will work.

上一页2 / 4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