How much reads really needs for de novo RNA-seq?

Question

[Whoknows]

Hi guys,

I challenged with an old question, how to deal with RNA-seq de novo sample size. I did a project with 90b length and 10 Gb data it gave good result. Also, another project on plant transcriptome with 90b length and 8Gb data works fine.

Here is my big challenge, we have number of options in read length 75, 100 and 150b It seems 150b should be better but its price goes out of our budget. My questions:

1. How much paired-end reads do we really needs for analyzing plant de novo RNA-seq?
2. Is it worth to reduce sample size instead of sample size? Is there a trade-off between length and sample size?
3. Does anybody know paper about this issue?

thanks