Release Pipeline

From GO Wiki
Revision as of 10:34, 5 June 2018 by Vanaukenk (talk | contribs)
Jump to navigation Jump to search

This documentation is currently a work in progress.

Overview

The GO Consortium (GOC) is now publically releasing data on a monthly basis. Data includes annotation files, ontology files, GO-CAM models, and... Official monthly releases are versioned and archived so that analyses performed with these data can be reproduced at any point in the future. Additionally, daily snapshot releases of GO data are available for internal use by GOC members. This allows annotators, for example, to have access to the most up-to-date version of the ontology for their curation. However, data generated using snapshot releases will not be officially released until the monthly public release.

Release Cycle

For both the daily and monthly releases, the pipeline starts at midnight and finishes within 24 hours; starting nightly for the daily snapshot release and the first of the month (or as close as can be obtained if there are failures) for the monthly public release.

Annotations

The public release cycle is monthly; these are intended for end users. Snapshot releases are created daily; these are intended only for internal consumption by GOC members. In both cases, the pipeline starts at midnight and finishes within 24hrs

The pipeline is driven by YAML metadata files in the metadata/datasets folder on the go-site repo. Please see the README.md in this folder for a description of the structure

For example: https://github.com/geneontology/go-site/blob/master/metadata/datasets/tair.yaml

The most important field here is the source tag. This dictates where GO Central pulls each contributing group's GAF from. This can be an FTP URL on a FTP site managed by that group, an S3 bucket, an HTTP server, etc.--anything as long as it resolves to the latest submitted GAF.

The pipeline will then run checks on this (see below) and repair any auto-repairable issues (for example, migrating annotations to merged terms). It will then public the processes GAFs, GPADs, GPIs, etc. to a public site, where it is available for the public to download.

The "publish" sites that are currently part of the pipeline are:

All pipeline runs start at midnight (12am) PDT, and currently take about 14hrs (this will be decreased in the future). The `release`/`current` pipeline runs are attempted on the first of every month. As a note to that, the `snapshot` run does not currently run on the day of the `release`.

Per-group curator QC reports

We run the full pipeline every day, sans the SVN writeback and software deployments; we call this snapshot. As part of this run, just as in the release, we generate products and reports that are of use to db admins and curators of the various contributing groups (MODs, UniProt, etc) to the GO Consortium.

Summary

This is a basic summary of the parsing of your GAF file. It functionally replaces the old "Mike's script"

These are found in reports

Example: http://snapshot.geneontology.org/reports/dictybase.report.md

These report basic syntax errors and implement a subset of checks in the GO QC Rules

Prediction Report and OWLTools Checks

Predictions

Technical Details

See the README.md in the pipeline GitHub repo: https://github.com/geneontology/pipeline

GO Consortium Dataflow

https://github.com/geneontology/go-site/blob/master/docs/go-consortium-dataflow.png