Performs variant calling on an input cram, using a re-write of (DeepVariant)[A universal SNP and small-indel variant caller using deep neural networks | Nature Biotechnology] which is adapted for Ultima Genomics data. There are three stages to the variant calling: (1) make_examples - Looks for “active regions” with potential candidates. Within these regions, it performs local assembly (haplotypes), re-aligns the reads, and defines candidate variant. Images of the reads in the vicinity of the candidates are saved as protos in a tfrecord format. (2) ca
This is a companion discussion topic for the original entry at github.com/rickymagner/healthomics-workflows/EfficientDV