Loom Workflow Engine: Collaboration through portable, shareable data analysis Nathan A. Hammond1, Isaac Liao1, Sowmi Utiramerur2, Somalee Datta1
1Stanford Center for Genomics and Personalized Medicine 2Stanford Heath Care
[email protected]–[email protected]@loomengine.org
Contact
• Aworkflowframeworkcanensurereproducibilityand
implementothercommonfuncAonsinaconsistentwayacrossallsteps.
• Thisiseasier,cleaner,andsaferthandependingonthe
pipelinedevelopertoimplementthesefuncAonsforeverystepinineveryworkflow.
Whyuseaworkflowframework?
Portabilityarchitecture
Reproducibility&Traceability
• LoomabstractsplaEorm-levelservicessuchasfilestorage,compute,and
databaseoperaAons.Withsimpleadaptors,itcanrunonmanydifferentplaEorms.
• Loom’sclient-serverarchitectureallowsittoscalefromasingleuser
runningitononedesktoptomanyuserssharingaremoteLoomserver.
DockermakesrunAmeenvironmentsbothportableandreproducible.
• PaloAltoVAHospital• PhilTsao• CuipingPan
• StanfordClinicalGenomicsService• StanfordHealthCare• LucilePackardChildren’sHospital
SpecialthankstoourpartnersResourcesStanfordCenterforGenomicsandPersonalizedMedicinehRp://scgpm.stanford.edu/hRps://github.com/StanfordBioinformaAcs/loomhRps://pypi.python.org/pypi/loomengine
Toreproduceananalysis,youneedtoreassembleandverify:• Inputdata• RunAmeenvironment• Commandsexecuted
LoomautomaAcallykeepstrackoftheseforyou.FilesarealwaysidenAfiedbyhash,andverifiedbydefault.RunAmeenvironmentissavedusingDockerandtrackedbyanimmutableimageID.
WithLoom,thesameworkflowyourunonyourlaptopcanberuninthecloudwithoutmodificaAon.
Ge@ngstartedLoomissAllinpre-releasebutweinviteyoutocheckitoutandletusknowwhatyouthink!Youcanfindusongithub:hRps://github.com/StanfordBioinformaAcs/loom
ThinkbacktoacomputaBonalresult
from3yearsago• Howdidyougeneratethisresult?
• Whatwastheinputdata?Canyouverifyit?
• WhatwerethesoFwareversions?Areyousure?
• Canyoureruntheanalysisandverifytheresult?
• Cansomeoneelseruntheanalysiswithoutyourhelp?
ACADEMICRESEARCHmustbe
REPRODUCIBLE
Stand-aloneserver
Cloudprovider