dataflow-java

If you are ready to start coding, take a look at the information below. But if you are looking for a task-oriented list (e.g., How do I compute principal coordinate analysis with Google Genomics?), a better place to start is the Google Genomics Cookbook .

Getting started

• First, follow the Google Cloud Dataflow getting started instructions to set up your environment for Dataflow. You will need your Project ID and Google Cloud Storage bucket in the following steps.

• To use this code, build the client using Apache Maven:

  cd dataflow-java
  mvn compile
  mvn bundle:bundle
  

• Then, follow the Google Genomics sign up instructions to generate a valid client_secrets.json file.

• Move the client_secrets.json file into the dataflow-java directory. (Authentication will take place the first time you run a pipeline.)

• Then you can run a pipeline locally with the command line, passing in the Project ID and Google Cloud Storage bucket you made in the first step. This command runs the VariantSimilarity pipeline (which runs PCoA on a dataset):

  java -cp target/google-genomics-dataflow*.jar \
    com.google.cloud.genomics.dataflow.pipelines.VariantSimilarity \
    --project=your-project-id \
    --output=gs://your-bucket/output/localtest.txt \
    --secretsFile=client_secrets.json
  

  Note: when running locally, you may run into memory issues depending on the capacity of your local machine.

• To deploy your pipeline (which runs on Google Compute Engine), some additional command line arguments are required:

  java -cp target/google-genomics-dataflow*.jar \
    com.google.cloud.genomics.dataflow.pipelines.VariantSimilarity \
    --runner=BlockingDataflowPipelineRunner \
    --project=your-project-id \
    --stagingLocation=gs://your-bucket/staging \
    --output=gs://your-bucket/output/test.txt \
    --secretsFile=client_secrets.json \
    --numWorkers=4
  

  Note: By default, the max workers you can have without requesting more GCE quota is 16. (That's the default limit on VMs)

Command Line Options

Use --help to get more information about the command line options. Change the pipeline class name below to match the one you would like to run:

java -cp google-genomics-dataflow*.jar \
  com.google.cloud.genomics.dataflow.pipelines.VariantSimilarity --help

Code layout

The Main code directory contains several useful utilities:

coders:

includes Coder classes that are useful for Genomics pipelines. GenericJsonCoder can be used with any of the Java client library classes (like Read, Variant, etc)

functions:

contains common DoFns that can be reused as part of any pipeline. OutputPCoAFile is an example of a complex PTransform that provides a useful common analysis.

pipelines:

contains example pipelines which demonstrate how Google Cloud Dataflow can work with Google Genomics

• VariantSimilarity runs a principal coordinates analysis over a dataset containing variants, and writes a file of graph results that can be easily displayed by Google Sheets.
• IdentityByState runs IBS over a dataset containing variants. See the results/ibs directory for more information on how to use the pipeline's results.

readers:

contains functions that perform API calls to read data from the genomics API

utils:

contains utilities for running dataflow workflows against the genomics API

• DataflowWorkarounds contains workarounds needed to use the Google Cloud Dataflow APIs.
• GenomicsOptions.java and GenomicsDatasetOptions extend these classes for your command line options to take advantage of common command line functionality

Maven artifact

This code is also deployed as a Maven artifact through Sonatype. The utils-java readme has detailed instructions on how to deploy new versions.

To depend on this code, add the following to your pom.xml file:

<project>
  <dependencies>
    <dependency>
      <groupId>com.google.cloud.genomics</groupId>
      <artifactId>google-genomics-dataflow</artifactId>
      <version>LATEST</version>
    </dependency>
  </dependencies>
</project>

You can find the latest version in Maven's central repository

For an example pipeline that depends on this code in another GitHub repository, see https://github.com/googlegenomics/codelabs/tree/master/Java/PlatinumGenomes-variant-transformation.

Updating the prebuilt jar

The prebuilt jar is currently generated by hand, whenever the code seems stable and has enough new features to warrant the effort.

To update the jar, bundle and copy to a new location:

cd dataflow-java
mvn clean compile bundle:bundle
cp target/google-genomics-dataflow*.jar google-genomics-dataflow.jar

Project status

Goals

• Provide a Maven artifact which makes it easier to use Google Genomics within Google Cloud Dataflow.
• Provide some example pipelines which demonstrate how Dataflow can be used to analyze Genomics data.

Current status

This code is in active development:

• TODO: Refine the transmission probability pipeline
• TODO: Add more tests