theSIGNAL SCIENCE
6 September 2026
"Life is a relationship between molecules."
NEWSBUSINESSTECHNOLOGYWORLDSCIENCECULTUREENVIRONMENTSPORT
Science

The AI that learned proteins can change shape

Japanese researchers have tricked AlphaFold3 into seeing proteins as the restless, shape-shifting machines they really are.

Imagine a lock that can reshape itself into a key. That is, roughly, what proteins do — and it has been driving biologists quietly mad for decades. Proteins are chains of amino acids that fold into precise three-dimensional shapes. But they do not stay still.

They flex, twist, and snap between different "conformational states" in response to signals from other molecules, and those shape changes are often exactly what makes them useful — or dangerous — inside a living cell. In 2020, Google DeepMind's AlphaFold rewrote what was thought possible. Its AI could predict a protein's folded structure from its amino acid sequence alone, a problem that had stumped scientists for fifty years. John Jumper and Demis Hassabis shared the 2024 Nobel Prize in Chemistry for the achievement.

But AlphaFold had a peculiar blind spot: ask it about a protein that naturally exists in several shapes, and it would hand you just one. For drug designers, who need to understand every shape a target protein can take, that was a serious limitation. Now, researchers at Japan's Institute for Molecular Science — Jun Ohnuki and Kei-ichi Okazaki — have found a quietly clever fix. They introduced a repulsive force between the structures that AlphaFold3 generates.

Photo: Terry Vlisidis / Unsplash
A protein's shifting shapes, once invisible to AI, can now be coaxed into view by a repulsive-force trick.
Discuss
  • Why did AlphaFold's inability to show multiple protein shapes matter for medicine?
  • How does the 'repulsive force' trick force AlphaFold3 to explore different structural possibilities?
  • In what ways is a protein switching shapes similar to, or different from, a machine changing gears?
Science

Tiny Arrows Rewrite Human Migration Story

Inside the Obi-Rakhmat rock shelter in Uzbekistan, researchers have found something startling: triangular stone arrowheads, 80,000 years old, bearing the impact scars of actual use. That is 25,000 years before identical points appeared in France's Rhône Valley — carried, it seems, by Homo sapiens pushing into Neanderthal territory. For over a century, prehistoric timelines were built almost entirely from French caves.…
  • Should prehistoric timelines be rebuilt from non-European evidence first?
∞
AlphaFold is known to predict only a single conformation for many proteins, limiting its applicability to the life sciences, including drug design.
Jun Ohnuki, Institute for Molecular Science, Japan
Science

A Black Hole the Size of a Truck

In 1974, Stephen Hawking shocked physics: black holes actually glow, leaking energy until they vanish. Tiny ones should disappear fast. So how could one survive inside a star today? Here is the strange answer. Neutron stars may capture ultraheavy dark-matter particles, which pile up at the core and collapse into a minuscule "endoparasitic" black hole — roughly 40 metric tons, about a loaded semi-truck. Then a race begins. Hawking evaporation tries to shrink it; accreted stellar matter and fresh dark matter try to grow it.…
  • If dark matter feeds black holes inside stars, what does that mean for stellar lifetimes?
AMERICAS · Science
Genetically modified pig kidney keeps man alive for record nine months
OCEANIA · Science
Pacific islands to face food shortages due to global warming: Scientists
WORLD · Science
Ancient “living fossils” may reveal how complex life began
EUROPE · Science
Deadly floods of 2021—Researchers identify gaps in hazard mapping
theSIGNAL IN THE LAB
1VOCABULARY
conformational statesrepulsive force
structural landscapeimpact scarsendoparasitic
Hawking evaporationantibiotic resistance
2GRAMMAR FOCUS
Nominalisation — converting verbs to nouns for formal register
Nominalisation means turning a verb or adjective into a noun form, often using suffixes like -tion, -ment, -ance, or -al. This creates a more formal, academic register and allows complex ideas to function as the subject or object of a sentence.
prediction · discovery · migration · limitation · exploration · survival · collapse
  1. AlphaFold's of protein structures from amino acid sequences alone was considered impossible for fifty years.
  2. The of tiny stone arrowheads in Uzbekistan has forced researchers to rethink prehistoric timelines entirely.
  3. A key of early AlphaFold models was their inability to show the multiple shapes a single protein can take.
  4. The of the structural landscape, rather than settling on one answer, is exactly what the Japanese researchers' repulsive-force method encourages.
  5. Human into Europe may have been rehearsed first across Central Asian territories that archaeologists had long overlooked.
  6. The of a minuscule black hole at the core of a neutron star depends on whether accreted matter outpaces Hawking evaporation.
3COLLOCATIONS
Match the verb on the left with the noun phrase it best collocates with on the right.
  1. map
  2. bear
  3. share
  4. collapse into
  5. park in
4CRITICAL THINKING
The Obi-Rakhmat discovery suggests that a century of prehistoric science was skewed because researchers concentrated on one region. In what other scientific fields might geographic bias in data collection be quietly distorting our understanding? Who has the responsibility to correct that bias, and how?
5CREATIVE · HEADLINES
Write a headline for the top story in each of the following styles. One line each, no explanation:
  • TABLOID NEWSPAPER
  • LUXURY MAGAZINE
  • ACTIVIST BLOG
6WRITING
A parasitic black hole inside a dying star, an AI nudged into new territory, and tiny arrowheads that rewrite a continent's history — choose the discovery you find most unsettling and explain in one paragraph why uncertainty can be more valuable to science than a clean answer.
7DEGREES OF EXTREMITY
Complete each ladder from mild to strong.
  • curious→→
  • shift→→
  • useful→→
  • glance→→
  • hint→→
  • shrink→→
8SPEAKING
  1. Which article's finding would most change everyday human life?
  2. Could AI tools ever fully replace traditional laboratory experimentation?
  3. What surprises you most about how stars might actually die?
  4. Does finding older arrowheads in Uzbekistan change your idea of civilisation?