This gist is a follow up on https://gist.github.com/pichuan/7ad09bf1fa8f519facf6806eca835ea6 (which was trained with v1.4.0 codebase and training data)
I've trained a WGS and WES model on v1.5.0 codebase and training data, and added the additional allele frequency channel.
For more details about this work, see
This is an example from pichuan@ to run vg giraffe, so we can go from FASTQs --> BAM.
To test this, I got a machine using the command listed here: https://github.com/google/deepvariant/blob/r1.5/docs/deepvariant-details.md#command-for-a-cpu-only-machine-on-google-cloud-platform. I added more disks because 300G is not enough for the example below. I used --boot-disk-size "1000" for now.
(Note: the disk issue might be because I didn't run with --rm in docker.)
This gist is a follow up on https://gist.github.com/pichuan/64d73bc965300645470eb29a66116593 (which was trained with v1.1.0 codebase and training data)
In v1.4.0, our WGS
and WES models have a new default insert_size channel. Based on the requests
from our users, I've trained a WGS and WES model on v1.4.0 codebase and training
data, and added the additional allele frequency channel.
https://gist.github.com/pichuan/64d73bc965300645470eb29a66116593
Here is an example of how to apply the model described in Improving variant calling using population data and deep learning . The example is similar to DeepVariant WGS case study.
I got a machine using a command like this: https://github.com/google/deepvariant/blob/r1.1/docs/deepvariant-details.md#command-for-a-cpu-only-machine-on-google-cloud-platform
Using --variant_caller=vcf_candidate_importer in make_examples, we can use DeepVariant to act like a genotyper to force call candidates in the VCF file specified in --proposed_variants.
Before reading this guide, read the WES Case Study and metrics: