Correct me if I am wrong, but I believe that SNPEff doesn't check the inserted sequence when computing the effect of an insertion?
For instance an insertion of TTTA will be reported as a frameshift mutation. In the majority of cases this is fine, however if the current reading frame from the ORF annotation is the negative strand and the correct frame is selected the new codon will be:
The TAA is actually a stop codon, so the real effect of this insert should be insertion of a STOP CODON, not a frame shift in this instance. This is minor I know, but I though to report and make people aware.
I am using version 2.02 of SNPEff, but I don't see this fix reported up to 2.04.