QC of NGS analysis pipeline?

[mzpitman]

Sorry if this is a noob question, but how do you currently QC your NGS analysis workflow? - i.e. generate simulated data with known variants, then compare how your pipeline performed against known data... What are ways that you have set up a QC for your analysis workflow, pitfalls to avoid, etc.

[adaptivegenome]

Your best bet is to start with GATK Best Practices...

[mzpitman]

Your best bet is to start with GATK Best Practices...

Thanks for this - it looks like a good start, but I was wondering about specifically simulating data from different sequencers that mimics real data - like a region of interest - so that you can QC how your pipeline would react to different data types, maybe decide which sequencer would be best for a particular experiment...

[adaptivegenome]

Thanks for this - it looks like a good start, but I was wondering about specifically simulating data from different sequencers that mimics real data - like a region of interest - so that you can QC how your pipeline would react to different data types, maybe decide which sequencer would be best for a particular experiment...

This is of broad interest to many on this board. I don't have simple answer for you but I agree this should be an area of focus for all of us.

[turnersd]

This is of broad interest to many on this board. I don't have simple answer for you but I agree this should be an area of focus for all of us.

I agree. I hope this discussion continues.

What sort of NGS application are you considering? DNA? RNA-seq? ChIP-seq? Methylation?

Perhaps this would make for a good how to wiki page - techniques and software for pre- and post- alignment and assembly quality assessment / quality control.

[mzpitman]

This is of broad interest to many on this board. I don't have simple answer for you but I agree this should be an area of focus for all of us.

Thanks, that validates that at least I'm thinking in the right direction... I'm working on some tools for this type of analysis QC/data simulation. So, I'll let everybody know when I have something to try.

Still open to pointers from anyone who has tried something similar...

[mzpitman]

I agree. I hope this discussion continues.

What sort of NGS application are you considering? DNA? RNA-seq? ChIP-seq? Methylation?

Perhaps this would make for a good how to wiki page - techniques and software for pre- and post- alignment and assembly quality assessment / quality control.

So, at the moment I have mostly DNA data, but I'm also working on 1 RNA-seq project and 1 ChIP-Seq project as well.

I think a wiki page would be very cool for this, but the discussion needs to progress further. At this point this is mostly just an idea I have in my head a few hacked py scripts that aren't doing what they're told to, and some interest on this forum.

I hope this discussion does keep going because I think analysis QC is easily overlooked, but necessary - even critical in some settings.

[Alex Renwick]

...
Still open to pointers from anyone who has tried something similar...

You might start with wgsim for read simulation. It's purpose is more for assessing mapper accuracy than evaluating the whole analysis pipeline. Sequencing errors are limited to substitutions (no indels), and insertion variants are of limited length. But it's small and fast and reliable for what it does.

[mzpitman]

You might start with wgsim for read simulation. It's purpose is more for assessing mapper accuracy than evaluating the whole analysis pipeline. Sequencing errors are limited to substitutions (no indels), and insertion variants are of limited length. But it's small and fast and reliable for what it does.

Thanks. I'd seen this before, but it was a good reminder because I'd forgotten about it.