Hi,
I'm interested about your opinion on how often and when do you use, or don't use the visualization of individual reads from NGS. Or whether you most of the time just view the "piled-up" depictions (sometimes called "coverage") of the data.
Would it suffice, to you, to have only the "piled-up" representation for experiments such RNAseq, Chip-seq, xxx-Seq type of experiment?
Best regards
Edit 1
I'm not talking about the consensus sequence.
As in the following example, I'm interested, how often you find more helpful to look in the "hills" (row under the blue annotation), or the gray "lines" reads with mismatches highlighted in color.
Can you elaborate how you intend to generate that "piled-up" representation? Would that be a consensus sequence? At the end of the day no one practically looks at individual/single reads but people do look at individual reads when they are piling up in a location/gene etc when making decisions about SNP's etc.
Depiction of individual reads is very useful for structural variants (SVs) because some SV types do not cause a change in coverage (e.g., inversions, balanced translocations, ...).
Clinical DNA sequencing (whole exome) here. It really depends on the task.
If I want to check wether my gene got enough coverage at every relevant position the "pile-up" is sufficient.
If I want to have a look at a called SNP to judged if it might be an artifact (and if it is one, what went wrong?) I really need each individual read.
If I want to have a look at a called CNV I need both. The "pile-up" for first assesment. If I want to be shure to find the exact breaking points I have to look at indiviual reads.
...
I can't really quantify how much every task is done. But if any of this tasks could not be done with a genome browser I would not consider using it.
The core concept for DeepVariant comes from how human scientists look at a putative variant in a genome browser like IGV, evaluating the evidence: How many reads support the variant? Do the reads have good base and mapping quality scores? Are there any unexpected patterns in read mapping or other variants nearby?
Can you elaborate how you intend to generate that "piled-up" representation? Would that be a consensus sequence? At the end of the day no one practically looks at individual/single reads but people do look at individual reads when they are piling up in a location/gene etc when making decisions about SNP's etc.
98% of the time I just look at the coverage but once in a while, it is quite useful to have the individual reads as well.