AI Has Started Making Drug-Like Antibodies. The Hard Part Is Knowing What That Means.

Artificial intelligence can now help scientists create antibodies that look increasingly drug-like on a computer. That is a real technical shift, but it does not yet settle the bigger claim often made around AI biotech: that an algorithm can design a medicine ready for people without the long, expensive cycle of laboratory testing and refinement.

The reporting identifies a disagreement hidden inside the phrase "AI-designed." One camp counts an antibody as AI-designed when a computer proposes the basic sequence that scientists later improve into a clinical candidate. Another sets a much higher bar: the first design should be ready to move into the clinic with no further engineering. Those are not small differences in wording. They describe very different levels of scientific independence from the model.

By 2025, researchers had shown that AI could clear the first bar. But companies and researchers interviewed by STAT were still divided over whether new models could match or surpass established antibody-discovery methods at the harder task. Investors may value AI-native companies as if that future is already here, while even people building the systems remain cautious about what their outputs can replace.

For a public reader, the cleanest way to understand the field is to separate design from delivery. An AI system can make the search for a promising molecule faster and more imaginative; it cannot erase the need to show that a molecule works, is safe enough to test, can be manufactured, and helps real patients. The revolution will be measured in those handoffs, not in a striking computer-generated sequence alone.

Read the full story at STAT ↗

Retatrutide is investigational and is not FDA-approved. Other medicines referenced here are covered as third-party news for educational awareness only. Catalyst does not present this information as medical advice, treatment guidance, or a claim of safety, efficacy, or approval.