Project PCA (with a reference)

You can leverage reference panel information to see how samples in your data cluster on a “global” scale. PCs are computed using 1KG+HGDP dataset as a reference panel, and then samples in the input dataset are projected onto the 1KG+HGDP PC space. A random forest classifier model, adopted from gnomAD, is then used to assign population ancestries in the input dataset

Below is a code on how you can run projection PCA via the command-line or inside a Python script Use

1. Python (inside a Python script)

   import gwaspy.pca as pca
   pca.pca.pca(data_dirname="gs://my-gcs/bucket/test_data/", data_basename="my_data_basename",
               out_dir="gs://my-gcs/bucket/test_data/", input_type="my_input_type", reference="GRCh37",
               pca_type="project")
   

2. Command line

   pca --data-dirname gs://my-gcs/bucket/test_data/ --data-basename my_data_basename --out-dir gs://my-gcs/bucket/test_data/--input-type my_input_type --reference grch37 --pca-type project