My life & career
The public benefit is potentially profound, but predicted proteins must still be translated into validated therapies and products.
YOTTABIT / THE BIG STORIES / Biology
What if scientists could explore 200 million protein shapes?
Proteins perform essential jobs inside living things. Their shapes affect what they can do. The AlphaFold database has made more than 200 million predicted structures available for research.
THE BIG PICTURE
That's like giving researchers an enormous new collection of molecular clues. It could help scientists explore diseases, design proteins and prioritize ideas that once would have been difficult to investigate.
A predicted shape is a starting point, not a medicine. Researchers still need to prove whether it works in real biology.
THE YOTTABIT PERSPECTIVE
What does it mean to me?
Big change is fascinating. Its implications are what matter.
My business
Teams that learn to pair computational search with wet-lab validation can move through more ideas efficiently.
My industry
Drug development, enzymes, crop science and industrial biology gain new starting points for experimentation.
FROM JIM CARROLL’S WORK
Jim’s hyper-science writing anticipated how rapid information sharing and simulation could change discovery cycles.
Meet the futurist behind YottaBit ↗WHAT COULD I DO MONDAY MORNING?
Identify the experiment that must follow a promising computational prediction.
CURIOUS FOR MORE?
Explore the story behind the story.
For those who want the deeper explanation, the research is always available. The fascinating possibilities are only the beginning.
Research, explanations and sources EXPLORE THE DETAILS +
Predictions at scale
The AlphaFold Protein Structure Database makes more than 200 million structure predictions accessible. That extraordinary reference library helps researchers ask questions and prioritize experiments, but the predicted shape of a protein is not the same as proof of its biological function, interactions or therapeutic value.
A new scientific workflow
Computational models can narrow the space of plausible mechanisms and make it easier to design laboratory studies. AI models that reason about proteins and molecular interactions could eventually accelerate many stages of discovery. However, laboratory assays, in vivo biology, manufacturing and controlled human studies remain necessary to establish what actually works.
What would count as progress
Watch for repeatable predictions validated in independent experiments, useful synthesized molecules, lower costs per successful discovery and improved clinical results. YottaBit intentionally separates a structure prediction from a validated treatment.