Hi all, I'm running a DNAscope to GVCFtyper joint calling pipeline with Ultima Genomics UG 100 WGS data (184 samples) and have a few questions.

Pipeline

Step 1. DNAscope

sentieon driver \
  -r Homo_sapiens_assembly38.fasta \
  -t 32 \
  -i sample-006.cram \
  --read_filter UltimaReadFilter \
  --algo DNAscope \
  --model dnascope.model \
  --emit_mode gvcf \
  sample-006.raw.gvcf.gz 

Step 2. DNAModelApply

sentieon driver \
  -r Homo_sapiens_assembly38.fasta \
  -t 32 \
  --algo DNAModelApply \
  --model dnascope.model \
  --vcf sample-006.raw.gvcf.gz \
  sample-006.gvcf.gz

Step 3. GVCFtyper (184 samples)

sentieon driver \
  -r Homo_sapiens_assembly38.fasta \
  -t 18 \
  --algo GVCFtyper \
  -v sample1.gvcf.gz \
  -v sample2.gvcf.gz \
  ... \
  joint.vcf.gz

Questions

Q1. In the GVCFtyper output, both ML and PL FORMAT fields are absent. Is this expected for Ultima Genomics DNAscope gVCFs?

Q2. The input gVCFs have no FILTER=MLrejected tags (FILTER column is .). Should ML filtering be applied before joint calling, or is this normal for gVCF input?

Q3. Does GVCFtyper use the ML field internally during joint genotyping, or does it rely only on PL?

Q4. Our gVCFs were generated without --pcr_indel_model none despite being PCR-free data. How much does this affect indel accuracy, and is regeneration necessary? Thanks in advance!