I´m using miRNA databases to predict if a particular sequence has the potential to encode for miRNAs. I wonder to know what are the range of e-value that I can consider significant.
As I know MFE used to predict miRNA's and it must be lower than -18 kcal/mol (-19, -20). In fact after passing several filters you should predict secondary structure and just miRNA's with minimum free enegry lower than -18 kca/mol could be candidate for new predicted miRNA.
It is not an easy question because e-values are dependent on database size. Maybe you should use the query cover and the identity. Unfortunately I do not have the expertise to indicate such values for miRNA.
As I know MFE used to predict miRNA's and it must be lower than -18 kcal/mol (-19, -20). In fact after passing several filters you should predict secondary structure and just miRNA's with minimum free enegry lower than -18 kca/mol could be candidate for new predicted miRNA.
As far as I know there is no way to calculate a p-value from an E-value.
But since a E-value is the "expected number of hits given a size of the database" it should be (much) smaller than 1/database.size (since this is the limit where you would expect to find 1 by chance).
On top of my head i think:
1 / (database.size * p)
where p is 95 or 100 sounds reasonable since you then your hit to not be expected to be found even if your database was 95 or 100 times larger, which are the thresholds as we usually use on p-values (0.05 and 0.01) - but just keep in mind they are not not the same and cannot be interpret the same way!