Bowmanalsogivestheexampleofsome
recentresearchintoproteinsinvolvedwith
antibioticresistance,anotherpublichealth
threatatthemoment.‘Wefoundoneofthese
crypticbindingsitesandexperimentally
confirmedthatit’sthere,’hesays.‘Thenwe
subsequentlycomputationallydesigned,
andexperimentallyverified,a coupleofsmall
moleculesthatbindtothispocketandshutthe
proteindown.Sothosearenowinteresting
startingpointsforfurtherdevelopment.’
Anotherexampleis a proteinfromEbola
virus,whichBowmansays‘hastypically
beenassumedtobeundruggable,because
if youlookatthestructuralsnapshotsfrom
experiments,therearenopocketswherea
smallmoleculedrugis likelytobeabletobind
tightly,andshutofftheprotein’sfunction.But
again,wefoundoneofthesecrypticpockets,
andwe’vedonetheexperimentstoprovethat
it exists.Wewerejuststartingtoscreenfor
smallmoleculeswhenwehadtoshutdown
forthestay-at-homeworkers.’
WeaskBowmanif Folding@home’s
workcouldpotentiallybeusedina COVID-
19 vaccinedesign.‘Yes,I thinkthereare
opportunitiesthere,’hereplies.‘We’realso
startingtogetintoa littlebitofantibody
design,whichI thinkcouldbeusefulonthe
vaccineside.’
GPUfolding
If youwantyourPCtodoasmuchworkfor
Folding@homeaspossible,you’llwantto
enablefoldingonyourGPU.Withloadsof
memorybandwidthandhugebanksoftiny
streamprocessorsavailable,graphicscards
areidealforthehighlyparallelworkinvolved
inFolding@home’s simulations.
‘It’sreallythisideathatyoucanperformthis
onecommonoperationonmanydifferent
inputsinparallel,’saysBowman.‘It matches
reallywellwithoursimulations,where
weessentiallywanttoperformthesame
calculationoneachatom– howit’sbeing
pushedonandpulledonbyitsneighbours–
inparallelasquicklyaspossible.
TheGPUfoldingclienthaschangedover
theyears.Fora longwhile,youcouldonlyfold
onATI(nowAMD)GPUs,andthenanNvidia
foldingclientcamealongthatoutperformed
ATIGPUs.
GPUfoldinginvolvesusinga GPUcompute
applicationprogrammersinterface(API),in
muchthesamewayasgaminggraphicsuses
VulkanorDirectX.ThemainAPIsforGPU
computeareOpenCLandDirectCompute,
whichrunonbothAMDandNvidiaGPUs,and
CUDA,whichonlyrunsonNvidiaGPUs.The
currentGPUfoldingclientusesOpenCLon
bothAMDandNvidiaGPUs,butBowmansays
therewillsoonbea versionthatmakesuseof
Nvidia’sproprietaryCUDAAPIaswell.
TheFolding@homeclientusesOpenMM
(openmm.org), whichis a toolkitformolecular
simulation.PeterKennethEastman,developer
oftheFolding@homeGPU-basedsimulation
codeexplainsthat‘OpenMMcanuseeither
CUDAorOpenCL.It’ssupportedbothformany
years.It’s just a matter of creating a Folding@
home core that uses CUDA, which is mostly
just a packaging/licensing issue’.
At the moment, CUDA gives Nvidia a
performance advantage in OpenMM. ‘It’s
around 30 per cent faster when using CUDA
instead of OpenCL,’ says Eastman. Also, while
both AMD and Nvidia GPUs can use OpenCL,
the Nvidia ones are usually quicker. ‘In terms ofoverall speed, Nvidia processors are around
twotimesfasterthansimilarlypricedAMD
processors,’saysEastman,adding,’I suspect
this is more due to deficiencies in AMD’s
OpenCL compiler and runtime, rather than
the hardware itself.’
That doesn’t mean you need to spend a
lot of money on a top-end Nvidia GeForce
RTX GPU to get the best folding performance
though. AMD GPUs are still much faster than
CPUs at folding, and you can also use older
Pascal Nvidia GPUs to great effect for folding,
as the client doesn’t make use of the RT or
Tensor cores found in Nvidia’s RTX Turing
chipsanyway.‘It doesn’tusethem,’confirms
Eastman.’We haven’t figured out anything
they’re useful for in molecular dynamics.’Keep folding
The move to COVID-19 research during
the current pandemic has brought a lot of
attention to Folding@home, as well as loads
more computing power. PC Master Race
(pcmasterrace.org) has been promoting
it hard on Reddit and Twitter, and it’s
subsequently been pushed by Nvidia and
various component makers.
The Custom PC team has been folding
since 2004, and we used to regularly feature
monthly stats on how the team was doing in
the magazine, as well as reporting on various
users’ milestones, which we’ve brought back
on p107. However, after several years of solid
folding work, general interest in our Folding@
home team started to wane a few years ago,
particularly during the GPU cryptocurrency
mining boom. We’re hoping that we can
now get it all going again and help to make
a decent contribution towards the project’s
COVID-19 research.
‘It’s really exciting to see people coming
together over this – joining forces to tackle
this shared enemy,’ says Bowman. But he
also adds that ‘we’re only in a position to start
doing this because we already had this base
of 30,000 users who were excited to help
with our work on cancer and Alzheimer’s
and other diseases. I hope when we finally
get this COVID-19 situation under control
that people will continue their appreciation
for this, and help us to attack these other
maladies as well.’
In the same way, we’d also like to give a
big ‘thank you’ to everyone who has folded
for the Custom PC & bit-tech team (35947)
throughout the years.Folding@home’s research on cryptic pockets, screening small molecules
and antibody design could be useful for creating a COVID-19 vaccineFEATURE/ ANALYSIS