DGE - filter or not filter

[masterpiece]

Hi all,

Want to get your opinion on this matter.

On Digital Gene Expression (DGE), the low abundant tags (for example count = 1) consist of high percentage of total number of tags. This low abundant tags could due to sequencing error. So my question here,

1. Before running differential analysis (DE), is it important to filtered out any of these low abundant tags? How is this can effect the statistic analysis?

2. Different paper using different count to filter of low abundant tags. How we should set the cutoff value for low abundant tags?

Thanks in advance

-Kamal