BioCompute Objects: Tools for Communicating Next Generation Sequencing Data and Analysis

May 14, 2019, FDA, Maryland, USA

Jonas S Almeida, NCI/DCEG Chief Data Scientist

 

BCO as a manifest for orchestrating Web API ecosystems.

 

Modern cloud execution models such as “serverless computing” (FaaS) create stateless API ecosystems that allow the code to move to the data, rather than the more conventional have the data travel to where the application is deployed. The latter is much harder to scale and is nearly impossible to govern transparently. This presentation will illustrate this mechanism with applications to population studies, from genomics to pathology. It will also explore how a computational ecosystems approach using popular APIs for computational genomics, such as GA4GH and GDC, depends on the ability to operate without loss of governance. In that regard BCOs represent a particularly interesting opportunity to define computational manifests that orchestrate the authenticated operation reproducibly.

mathbiol.github.io/bco

 

github.com/mathbiol/bco

 

Web execution

try it yourself

---

 

Go to a domain you don't control, say Washington Post.

 

Inject bco library: $.getScript('https://mathbiol.github.io/bco/bco.js')

 

Execute a BCO:
egfr = await BCO.webExe('https://mathbiol.github.io/bco/BCOexamples/demo.json')

---

Discussion

~3/4 of workflows not reproducible

APIs can be tested is environments we don't control

I was defeated by FDA this morning

Consumer-facing NGS is on FHIR ^*

__________________________|