Demo code showing how one could possibly do genotype-phenotype analysis using ADAM. This is work-in-progress. Currently, we implement a simple case/control analysis using a Chi squared test.
To build, install Maven. Then run:
mvn package
Maven will automatically pull down and install all of the necessary dependencies.
Occasionally, building in Maven will fail due to memory issues. You can work around this
by setting the MAVEN_OPTS environment variable to -Xmx2g -XX:MaxPermSize=1g.
To run, you'll need to install Spark. If you are just evaluating locally, you can use a prebuilt Spark distribution. If you'd like to use a cluster, refer to Spark's cluster overview.
Once Spark is installed, set the environment variable SPARK_HOME to point to the Spark
installation root directory. Then, you can run gnocchi via ./bin/gnocchi-submit.
We include test data. You can run with the test data by running:
./bin/gnocchi-submit regressPhenotypes testData/sample.vcf testData/samplePhenotypes.csv testData/associations -saveAsText
We accept phenotype inputs in a CSV format:
Sample,Phenotype,Has Phenotype
mySample,a phenotype,true
The has phenotype column is binary true/false. See the test data for more descriptions.
This project is released under an Apache 2.0 license.