germline at least 30x, somatic depends on your question.

have a look at this paper: http://jmd.amjpathol.org/article/S1525-1578%2817%2930025-9/fulltext

I suggest a couple of readings for theory and some figures.

https://rtsf.natsci.msu.edu/_rtsf/assets/File/depth%20and%20coverage.pdf

https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-14-195

We are successful with 100X paired end sequencing in whole exome analysis (Illumina 2500 plateform, sure selectec libraries). 

Depends on the pipeline and analysis tools that are being used.

We have used 20X and got success, some people have no success with 70X.

The more coverage and depth you have, greater is the chance of success.

Best Wishes.

below 20x it should not be considered, especially cause now it´s quite easy to pull the average coverage around 100x.

we do 100X for control sample(blood for germline variation) and 300X for tumor sample.

200 x for somatic mutations in cancer samples and 50 x for germline mutations