OUR‘EASY’
PROBLEMSCOULD
EASILYTAKE 100YEARSONATYPICAL
DESKTOPCOMPUTERthis string is ‘folded’ into a three-dimensional
shape. So far so good, but proteins can also
unfold and misfold.
As a basic analogy, when you crack open
an egg, the proteins will be in their natural
runny state, but if you apply heat, the proteins
in the egg white will start to unfold. The amino
acids that make up the different proteins
in the egg white will then mix, causing it to
change texture. That’s great for cooking an
egg, but proteins changing like this inside a
human body can have serious consequences,
including diseases such Alzheimer’s.
At the beginning, Folding@home was trying
to understand ‘how these little molecular
machines spontaneously self-assemble’,
says Bowman, but ‘now the focus has really
shifted to asking how they function and
malfunction, and how we can control that.
‘One of the things that my group is really
focusing on is hunting for what we call
cryptic pockets. The idea is that experimental
structures give you a snapshot of what a
protein usually looks like, but there are lots
of moving pieces that you don’t get a sense
of from these single snapshots. Watching
how the atoms in a protein move relative to
one another is important because it captures
valuable information that’s inaccessible by
any other means. Taking the experimental
structures as starting points, we can simulate
how all the atoms in the protein move.’
The problem is that protein science is
enormously complicated. ‘We’re simulating
these processes with atomic resolution,’
explains Bowman. ‘These things are sensitive
to changing just a few atoms out of many
thousands of atoms in a protein, so we really
need that level of detail to truly understand
them. Simulating how all these atoms in a
protein are moving as time progresses isextremely computationally expensive, and
besides our simulation there’s no real way to
observe it, say, experimentally.’Absolute units
Folding@home breaks up these enormously
complicated simulations into ‘work units’,
which are dished out to computers all over the
world running the Folding@home client. Each
work unit represents a fraction of an overall
protein simulation, enabling simulations
to be processed that would otherwise be
completely impractical.
‘Our “easy” problems could easily take 100
years on a typical desktop computer,’ says
Bowman, ‘and some of the harder problems
could take millions of years or longer. The
idea is to take these essentially impossible
calculations and break them up into lots of
small pieces that can be distributed to many
computers and run them in parallel on these
independent machines.’
Once your CPU or GPU has crunched
through the work unit, the result is sent back
to Folding@home, and you’re rewarded with
points. This points system has led to someThe Folding@home client Viewer lets you see a
visualisation of the molecule you’re processing