Question on re-aligned bam in deeptrio

[sounkou-bioinfo]

Hi all,

My question is around the re-aligned bam emitted for the trio case
I run deeptrio to check some result (variant is not called in the proband, both parents are heterozyguous on the site). No variant is called in deepvariant nor deeptrio for the target sample in this site. I thought to check the bam but i am having a hard time interpreting the trio bam. Below the code and the screen shots (deepvariant then deeptrio re-aligned bams, the side of interest is the snp)

time docker run \
  -v "${inputBamsDir}":"/input" \
  -v "${outputDir}":"/output" \
  -v "${fastqfiledir}":"/ref" \
  google/deepvariant:deeptrio-"${BIN_VERSION}" \
  /opt/deepvariant/bin/deeptrio/run_deeptrio \
  --model_type=${model} \
  --ref=/ref/${fastfilename} \
  --reads_child=/input/${names_child}.bam \
  --reads_parent1=/input/${names_parent1}.bam \
  --reads_parent2=/input/${names_parent2}.bam \
  --output_vcf_child /output/${names_child}_${regname}.output.vcf.gz \
  --output_vcf_parent1 /output/${names_parent1}_${regname}.output.vcf.gz \
  --output_vcf_parent2 /output/${names_parent2}_${regname}.output.vcf.gz \
  --sample_name_child ${names_child} \
  --sample_name_parent1 ${names_parent1} \
  --sample_name_parent2 ${names_parent2} \
  --num_shards ${nproc}  \
  --regions ${regions}   \
  --intermediate_results_dir /output/intermediate_results_dir \
  --output_gvcf_child /output/${names_child}_${regname}.g.vcf.gz \
  --output_gvcf_parent1 /output/${names_parent1}_${regname}.g.vcf.gz \
  --output_gvcf_parent2 /output/${names_parent2}_${regname}.g.vcf.gz \
  --make_examples_extra_args="ws_use_window_selector_model=false,emit_realigned_reads=true,realigner_diagnostics=/output/realigned_reads" \
  --disable_small_model=true


  time docker run \
  -v "${inputBamsDir}":"/input" \
  -v "${outputDir}":"/output" \
  -v "${fastqfiledir}":"/ref" \
  google/deepvariant:"${BIN_VERSION}" \
  /opt/deepvariant/bin/run_deepvariant \
  --model_type=${model} \
  --ref=/ref/${fastfilename} \
  --reads=/input/${names_child}.bam \
  --output_vcf=/output/${names_child}_${regname}_single.output.vcf.gz \
  --num_shards ${nproc}  \
  --regions ${regions}   \
  --intermediate_results_dir /output/intermediate_results_dir \
  --make_examples_extra_args="ws_use_window_selector_model=false,emit_realigned_reads=true,realigner_diagnostics=/output/realigned_reads_single" \
  --disable_small_model=true

Thank you

[lucasbrambrink]

Hi,

I'm happy to help! Unfortunately I need a little more information about what you are looking to understand specifically. Are you saying that the top row of the IGV screenshot is the emitted realigned BAM file from DeepVariant, and the bottom row is from DeepTrio? And if so, what is the aspect that is unclear to you?

If you would like, you can provide the 2 bam files and the specific variant site in question and I can take a look in IGV myself.

Thanks!

[sounkou-bioinfo]

Hi @lucasbrambrink,

Thank you for your reply !

The top (re-aligned ) reads are from DeepVariant and the bottom ones are from DeepTrio.

My question is about which reads are used for the DeepTrio re-aligned pipelup, is it only the proband or the three samples ?

Note that the DeepVariant re-aligned pileup confirms the no call (looks about the same without re-alignment)

I don t have access to the bams now for sharing with you

I don t know if this is relevant, by i used the dragmap-os aligner

[lucasbrambrink]

DeepTrio uses all 3 samples in the pileup image, the order of [parent1, child, parent2]. That means that reads from all 3 samples are used. I should add that a pileup image is generated for all candidates, regardless of which sample it came from.

DeepVariant assumes a single sample, so only reads from the one sample are used.

Let me know if that answers your question!

[sounkou-bioinfo]

DeepTrio uses all 3 samples in the pileup image, the order of [parent1, child, parent2]. That means that reads from all 3 samples are used. I should add that a pileup image is generated for all candidates, regardless of which sample it came from.

DeepVariant assumes a single sample, so only reads from the one sample are used.

Let me know if that answers your question!

Thank you @lucasbrambrink
This answere the question