Does VQSR help or obscure new variation?

[Chfranck]

I have been reading about VQRS and I have a question. It seems that VQSL fits a Gaussian mixture model model which describes the behavior of a few statistics for known variants, then alters genotype calls to favor variants that are known to exist in the population.

By favoring known variants when making your calls, is it true that you punish your ability to detect something new? *If your known set is missing key variants that exist in the population, then wouldn't you worsen your chance of discovering the unknown variants by favoring what you already "know"? Or am I completely missing something?

What are the key advantages of using VQSR? Thanks!

[nilshomer]

Everything has a prior. If you know the sun has come up the past 50 days, then it's prior is quite high. Know variants should be accepted with less evidence than new mutations for this reason.

[adaptivegenome]

So how is it useful? Does VQSR make any adjustments to the confidence of mutations not previously known? And it seems that you need a lot of data for this to be useful, which we have for SNPs. How useful is this for indels?