Hi,

Could you give some insight into the mass accuracy of the experiment you performed, the tolerance set for your database search, the size of your dataset (# queries), the size (# sequences) of your forward database and that of your reversed database? Are your reversed sequences contained within your normal database (so is it one and the same database)?

These parameters can probably shed some light on the issue you're encountering.

I make tolerance narrow, i used homo sapiens uniprot database and reversed this. 

Hello,

Regarding false discovery rates, what you need to control for is the proportion of decoy hits and hits from the normal (forward) database. For this you can combine the results of your two searches (normal and decoy database), order the peptide / protein hits by decreasing score, or by increasing p-values and while going down through the list, check at which point the number of decoy hits is more than the desired proportion / FDR. At this place you can stop accepting additional protein / peptide identifications.

Figure 1 from this paper illustrates this approach:

http://www.biomedcentral.com/content/pdf/1471-2105-10-179.pdf