GATK Best Practices Workshop: Data Pre-Processing

GATK Best Practices Workshop
Data Pre-Processing

This past Monday and Tuesday I was able to attend the GATK Best Practices Seminars being held at the Broad Institute in Cambridge, MA.

Here's the download of the Monday morning session:

And by way of note...all slides from the workshop can be found on a link via the GATK forum.

By way of introduction...

• GATK doesn't really do 'mapping' though they have suggestions for tools of preferred use for mapping.
• GATK is a post-processing tool.
• The integrated genome viewer (IGV) is a user friendly tool for visualizing whole genome data.
• Be sure to pay attention to your study design...are you 'deep' sequencing of 'shallow' sequencing? This will affect data processing and interpretation.

With deep sequencing designs:

• You will have increased sensitivity for variant detection
•  More accurate genotyping
• Caveat: No information about multiple samples as a deep sequencing design often times means you can only do 1 or very few samples.

With shallow sequencing designs:

• Your sensitivity will depend on the frequency of the variation of interest.
• Not as accurate genotyping potentially.
• You may discover more 'total' variants across more samples, however your confidence in real variants versus 'error' due to not deep enough sequencing may be reduced.

 Best Practices:

• Based on human whole genome or whole exome analysis
• Definitely hit up the documentation, it is quite extensive
• Not necessarily applicable to all datasets, can be used as a general guideline but if you are working with bacteria or viruses (like me) then some parts may not be applicable or calibrated the way you need them to be so you'll have to 'play around'
• Depends on design (see above)
• Use the forum, a great place to lob questions at the developers regarding the tools.

So here's the layout: Also available on their website

credit:  http://www.broadinstitute.org/gatk/guide/best-practices

Other suggestions before we jumped in per the Exome sequencing they have been doing:

• Add 50 bp of 'padding' on either side of your intervals (ie. exome, genes, loci, regions, etc...)
• Run at least 50 samples/run
• If you don't have 50 samples you can pull from the 1000 genomes project and do 'joint-calling', if you don't work on humans...pull samples from your preferred database make sure formatting, meta-information, is the same though so you'll have to do some more manipulation potentially.
• You can also use 'hard filters' per the best practices recommendations.

BTW....Tangent if you ever get the opportunity to attend seminar/talk with Eric Banks, do it. He's a fabulous speaker who answers any question great or small.

Let's get started...

Blog Series: WoG, Cesky Krumlov: Day 8: Transcriptomics

Where did days 6 and 7 go??? 

Lost inevitably in the chaos of RStudio Lab tutorial, Python Lab Tutorial, Cesky Krumlov Castle touring and wine tasting followed by copious amounts of wandering, eating, more wandering, snow ball fights, freezing feet, beer drinking and sleeping.

The tutorials are well written, given there was no lecture for these, I've simply linked the tutorials...go nutz.

Remember I've discussed python in previous blogs in the programming prep blog for instance and linked Tyghe's website for further tutorials using python--the resources are at your fingertips, have at it! Feel free to direct python related questions to Tyghe's blog or if you happen to know Daniel McDonald from the Univ. of Colorado feel free to harass him as well...I will not link his email in case he hunts me down for giving you all his information and challenges my husband to a python duel--coding at dawn!

Now...on to week 2!!!