18 | New Scientist | 29 February 2020
Tiniest tadpoles suck
on bubbles to get air
MOST tadpoles have to breathe air
to live but hatchlings are too feeble
to break the “skin” on a pond’s
surface caused by water tension –
so they suck air bubbles instead.
While tadpoles have gills, most
also develop lungs and frequently
surface to breathe air, which is
essential for survival in water
containing low levels of oxygen.
Kurt Schwenk at the University
of Connecticut saw the bubble
AI finds best way
to charge batteries
ARTIFICIAL intelligence could help
to create longer-lasting electric
car batteries that charge faster.
William Chueh at Stanford
University in California and his
colleagues have developed an AI
that optimises recharging while
also maximising battery lifespan.
There are many ways to
recharge, says Chueh. Standard
electric vehicle (EV) batteries tend
to be recharged fast at first, and
then more slowly, for example.
How a battery is recharged is
crucial because a sub-optimal
charging pattern has the potential
to significantly reduce battery
lifetime, but reducing charging
time is also important.
While a petrol car takes only a
few minutes to refuel, electric cars
can take between 45 minutes and
2 hours to fully charge. Reducing
this time would help improve the
driving experience for people who
Electric cars^ Animal behaviour
YOU might think a person’s risk of
catching HIV is solely governed by
exposure to the virus. But a new
study shows genetics may account
for up to 42 per cent of the variation
in the likelihood of getting infected.
These findings could help us
to develop new treatments or
vaccines against HIV, says Timothy
Powell at King’s College London.
He was interested in the fact that
millions of people are exposed to
HIV every year but don’t catch it.
For instance, only about a third of
babies get infected after being born
to HIV-positive mothers who aren’t
taking drug treatment for the virus.
In 2013, another team looked
at about 6000 people with HIV
and 7000 similar people without it
to see if there were relevant genetic
differences between the groups.
That team identified one gene,
which encodes a protein called
CCR5. This is on the surface of
immune cells and HIV attaches to it
in order to enter such cells (pictured,
in blue). A small percentage of
people lack this protein and seem to
be virtually immune to catching HIV.
Powell and his team reanalysed
the data from this 2013 study and
found many other genetic variants
that play a role. They also looked at
a separate large study of health and
genes and found that high genetic
risk of HIV infection was linked to
lower blood levels of a molecule
called CCL17, which is involved in
signalling between immune cells
(Scientific Reports, doi.org/dnch).
If confirmed, this suggests
vaccine developers could focus
on people with low CCL17 when
developing their products, says
Powell. Clare Wilson
Genetic influence on risk
of contracting HIV revealed
sucking by chance while studying
salamanders feeding on tadpoles.
“I assumed this had been described
before but it hadn’t,” he says.
Schwenk and Jackson Phillips,
a PhD student at the university,
filmed tadpole hatchlings from
five frog species swimming up
to take a breath. Due to surface
tension they couldn’t break
through to gulp air. Instead, they
stuck their open mouths to the
underside of the water’s surface.
By dropping the floor of the
mouth, tadpoles suck at the water
surface and create an air pocket
that they can pinch off by quickly
closing their jaws. This forms a
bubble in the mouth that contains
fresh air and a bit of exhaled air.
Raising the floor of the mouth
squeezes the bubble, forcing air
into the lungs (Proceedings of the
Royal Society B, doi.org/dm6f).
A second study showed that
even when tadpoles were older
and big enough to breach, those
from two species of frog preferred
this method. James Urquhart
use electric vehicles, says Chueh.
The researchers set out to
explore how best to recharge an
EV battery to 80 per cent within
10 minutes. They trained an AI on
existing information from 41 EV
batteries that had been used to
the point of failure, which took
about 1000 charge cycles.
The AI was given only data from
the first 100 cycles, and needed to
find early indicators to predict
battery life. It achieved a close
correlation between the predicted
and actual battery performance.
The team then used the AI to find
which of 224 potential charging
patterns was most efficient.
It took the AI 16 days to
determine which charge patterns
were best (Nature, doi.org/dm6c).
Given that an EV battery is meant
to last up to 1000 charge cycles, or
around 10 years of regular driving,
evaluating how charging patterns
affect lifetime was previously a
lengthy process: fully exhausting
a battery in testing took more
than 500 days. Donna Lu
STEVE GSCHMEISSNER/SCIENCE PHOTO LIBRARY/ALAMY
News In brief