Recommendations for variant call from long-read RNA-seq

[camelest]

Describe the issue: Thank you so much for the great tool.

I'm working on a heterozygous mouse long-read RNA-seq dataset from PacBio and would like to perform variant call + phasing at read-level. I'm wondering whether you have some recommendations regarding the points below:

• I'm currently using --model_type=PACBIO with the bam files processed with gatk SplitNCigarReads. Does this model consider RNA editing? Or should I use --model_type=WES? I saw some discussions mentioning WES model considers RNA-editing in https://github.com/google/deepvariant/issues/775
• Is there anyway that I could integrate the known variants from genomic data into the variant calling? Or should it be integrated after DeepVariant variant call at vcf-level?

Setup

• Operating system: Ubuntu 2.20
• DeepVariant version: v1.6.1
• Installation method (Docker, built from source, etc.): singularity
• Type of data: PacBio HiFi, mm10, long-read RNA-seq data

Steps to reproduce:

• Command:

  singularity run -B /usr/lib/locale/:/usr/lib/locale/ docker://google/deepvariant:1.6.1 \
  /opt/deepvariant/bin/run_deepvariant \
  --model_type=PACBIO \
  --ref=GRCm38.primary_assembly.genome.fa \
  --reads=SNCR.bam \
  --output_vcf=output.vcf.gz \
  --num_shards 16

Thank you so much for your kind help!

[AndrewCarroll]

Hi @camelest

Thanks for the issue. We have trained a DeepVariant isoseq model specifically for the type of data you have (I presume you have done: splitNCigarReads + FlagCorrection). We're planning to make that available to everyone soon. If you like, we can give you a custom model for this application in advance. If so, please email awcarroll@google.com and I can send you the files.

[camelest]

Hi, @AndrewCarroll,

Thank you so much for your kind offer. That would be really helpful. I will send an email at the address.

Best,