RNA-seq pre-processing about deepvariant HOT 4 CLOSED

Do you recommend read trimming before alignment using tools such as fastp?

No - we don't perform read trimming and in fact, this can hurt model performance (because we trained on untrimmed data). We observed this when running our models on processed with Opossum (which performs trimming):

I can see that this is the repository for the human reference genome, which DeepVariant recommends.

You can use a reference with or without alt contigs. With ALT contigs we have previously observed a slight increase in error rates with other models. I have not investigated whether they have an impact with our RNA-seq model.

Note: I have updated this after speaking with my teammates regarding this question.

from deepvariant.

Can you recommend a GTF annotation?

from deepvariant.

@malonzm1

I think that this should do fine. Can be found from the link I shared

GCA_000001405.15_GRCh38_full_analysis_set.refseq_annotation.gtf.gz

from deepvariant.

Hi @j0ntt , this has been inactive for a month, so I'll close it.
If there are remaining questions, please feel free to reopen with more details.

from deepvariant.