25 July 2020 | New Scientist | 33innate metabolism, which means different
people get different doses of metabolites
from their food.”
All of which suggests there is a food
mountain to climb. But there is a way
forward. There are many parallels, says
Barabási, between our understanding of
nutrition and health and our knowledge
of genetic epidemiology before the human
genome was sequenced.
In those pre-genome days, only around
1.4 per cent of human DNA – the chunk
that codes for proteins – was considered
important, and the other 98.6 per cent
was considered junk. The idea of sequencing
the whole genome was often dismissed as a
waste of time and resources, says Barabási.
And yet we now know that about two-thirds
of sequences linked to disease are in those
“junk” regions, many of which are stretches
of DNA that control gene expression.
That doesn’t imply that the majority of
associations between diet and health will
be found in nutritional dark matter. But
it would be foolish to dismiss 99.5 per cent
of the compounds we eat as unimportant,
says Barabási.
There is only one answer, he says: to fully
map the dark matter. His team and others are
already working on it. That inevitably means
having to analyse food the old fashioned way,
with slow and laborious lab techniques suchas mass spectrometry and nuclear magnetic
resonance. To make this a tractable problem,
these methods may have to be re-engineered
to make them faster, he says.Diet and disease
But there are other avenues. Barabási and
his team, who are physical scientists rather
than nutritionists, specialise in using big
data and the science of complex networks
to probe, among other things, the genetic
origin of human disease. This is what led
them to nutrition. “Genetic data can only
explain about 5 to 20 per cent of disease
causation,” he says. “We are interested in
finding the rest, and food is the next big
opportunity to fill the gap. Also, big data
has not yet touched nutrition.”
To test the potential of such approaches,
Barabási and his colleagues created a
text-mining tool to dredge the scientific
literature for scraps of information about
the constituents of food. This unearthed
some of the missing concentration figures
for garlic – for example, it found one for
diallyl disulphide, another pungent flavour
molecule with possible health benefits. It also
discovered a further 96 components of garlic
that aren’t even listed in FooDB. Doing
the same for cocoa, the researchers found
238 novel compounds. This suggests there is
a good deal of information out there already
which has yet to be pulled into the databases.
But garlic and cocoa are just two of about
2000 natural ingredients people consume.
“We are currently expanding to multiple
foods including milk, apple and basil,” says
Giulia Menichetti of the Network Science
Institute at Northeastern University in
Massachusetts, who co-developed the
text-analysis tool, called FoodMine. “We
aim to cover the most common staple
foods, to have a good grasp of the chemical
complexity of our diet.”
As if all that wasn’t hard enough, there
is the added complication that cooking
transforms chemical components into
others, sometimes with health implications.
When sugars and amino acids react at high
temperatures, the result is new molecules
that make roasted and grilled food >