The result is in gff 2 format. There are additional comment lines with predicted protein sequence. The gff file describes the predicted gene models. The amino acids sequence is the translation of the predicted coding sequence of the first predicted gene on contig 1 (between start codon (included as 'M') and stop codon). You can use BlastP vs NR to for a quick search.
However, your output is most likely bogus and cannot be used, because you used human training data (at least that is what -species=human switch indicates), but your genome is fungal. This doesn't work. For predicting eukaryote genes, you need appropriate training data from your organism or closely related organisms, e.g. RNA-seq data, full length cDNA, related organism's protein sequences, etc.
If you blast your predicted protein sequence from the example, one gets only very weak hits, none significant, of course this might be an exception. You could check all predictions like that if you don't believe me about the importance of training data; a very large proportion of predicted AA might not have significant hits, indicating that the prediction is not good.
If you want a state-of-the-art gene prediction, you should look at pipelines like MAKER, which include several tools, like Augustus, Snap, integrate evidence, proper repeat masking, and re-training.
•
written
by
bioinformaticssrm2011
