Science & Tech · Science & Environment · 4 hrs ago
Google method speeds protein structure prediction 8.6-fold
Google researchers developed a faster way to predict protein structures, a step that can help design molecules to attach to specific targets.
The method focuses calculations on the part of a target protein where a designed molecule is expected to bind, rather than processing the whole protein.
It made predictions 8.6 times faster without reducing accuracy in the researchers’ tests.
They used the approach to design proteins that bind to FAP, a molecule associated with solid tumors, and laboratory tests found some designs attached tightly.
The proteins are early-stage research, not proven cancer treatments, and their safety and effectiveness have not been established.
Google plans to combine robotic control with molecule discovery, while keeping human oversight and biosecurity screening in place.
Google researchers developed a method that speeds protein structure prediction by 8.6 times.
The technique focuses predictions on the local region where a designed protein is expected to attach.
Researchers used it to design proteins that bind to fibroblast activation protein alpha, a molecule associated with solid tumors.
The strongest tested designs achieved single-digit nanomolar binding affinity in laboratory experiments.
The protein binders have not been shown to safely or effectively treat cancer.
- Who
- Google researchers.
- What
- They developed a method that speeds protein structure prediction by 8.6 times and used it to design proteins that bind to FAP.
- When
- Published October 16, 2026. The research date is not stated.
- Where
- Not stated.
- Why
- To reduce the time and computational resources needed to test biological designs and potentially support the discovery of molecules for targeted cancer therapies.
This story does not have two clearly opposing sides.
Google researchers used the method to design candidate proteins that bind to FAP.
Researchers tested candidates in the laboratory using yeast surface display and fluorescence-activated cell sorting.
- Prediction speed
- 8.6 times faster
- Target molecule
- Fibroblast activation protein alpha (FAP)
- Binding affinity
- Single-digit nanomolar for the strongest experimentally tested designs
- DNA aptamer hit rate
- 33.5% for myoglobin and 42.1% for alpha-thrombin by the third round
- Alpha-thrombin specificity
- 98.5%