Variant Caller Analysis Dashboard and Data Management System
Clone or download
ihodes Merge pull request #865 from hammerlab/nfs_bams
Don't look for BAMs on HDFS
Latest commit a436c4f Dec 18, 2015
Type Name Latest commit message Commit time
Failed to load latest commit information.

Build Status Coverage Status

Looking at a VCF on the "Examine Page"


Cycledash tracks runs of somatic variant callers on various (BAM) datasets and provides an interface with which to inspect, analyze, debug, and improve the resultant variant calls.

The primary feature of Cycledash is its "Examine Page" (screenshot above), which allows users to quickly filter, order, and examine variants. A user can use a SQL-like syntax to filter down to variants based on attributes of the genotype (e.g. DP or GQ), their position in the genome (e.g. X:1500000-3000000), or other annotations added by Cycledash workers (e.g. the gene a variant falls in).

We embed the pileup.js pileup viewer within this page, allowing users to explore the pileup at a variant's location.

How We Use Cycledash

At Hammerlab we're using Cycledash to help us improve our distributed somatic variant caller, Guacamole.

Our workflow is:

  1. Ketrew, our workflow engine, starts a Guacamole job.
  2. When the job is complete, the resulting VCF and metadata is posted via a JSON RESTful interface to Cycledash.
  3. Cycledash processes the VCFs and presents them in an easy-to-navigate interface (found in the screenshot, above).
  4. If a validation VCF is posted with the main VCF, Cycledash calculates statistics like precision and recall.

Cycledash can also be used by researchers interested in quickly browsing VCFs for variants of interest.

Developing Cycledash

Cycledash is a Python Flask app with a React.js frontend. We use PostgreSQL as our database, and use a worker queue to execute longer-running tasks such as importing VCFs into Postgres or annotating variants with gene names.

More information about developing Cycledash can be found in the file in this repository.

Deploying Cycledash

For a quick, barebones deployment, follow the develop instructions.

For a more robust deployment, we use:

  1. gunicorn (so that many server processes may run at once)
  2. unicornherder
  3. Upstart to keep things up and running.
  4. nginx acts as a reverse proxy and serves (and manages cache headers for) our static assets.


We welcome bug reports and feature requests, and handle them through GitHub's issue tracker.

Please search our GitHub issues before filing an issue.


You can find documentation at /about on a running Cycledash instance, or at