This entirely depends on the application of the sequencing information you are interested in. e.g. if you are only interested in consensus level sequence information, 100x is allready very good data. If you want to study the presence of low frequency mutations (quasispecies), higher coverage is needed (towards 1000x). What is also important is the distribution of coverage depth over the genome, which is influenced by the protocol to generate the data (influenza targeted or random), the viral titer of your sample, etc., and may influence how easy your sequence assembly will be.

I agree with Steven. Just to add a bit about NGS for intrahost analysis: with random amplification - which can be influenced by GC content and RNA structure - I typically shoot for an average of 5000x coverage per site with no less than 1000x. That is not always achievable but is my gold standard.

Thanks Steven and Nathan very much for your reply. Yes, it is about quasispecies of an influenza virus from allantoic fluid. I got coverage from 60 to 198 with frequency 16 to 47%.