This gene is in forward strand and corresponds to the first gene predicted both by genscan and geneid. Each program predicted 3 exons (2 internals and a terminal) wich were nearly completely equal, except for the first internal one which begins slightly (31 nucleotides) before in the geneid prediction. This high coincidence makes the predictions very strong, though they aren't reinforced by no human EST.
Then, we run blastp to find human proteins homologous to the predicted ones. The results both for the geneid and the genscan predicted proteins have a similarity of 100% with the same protein (UDP-N-acetyl-alpha-D-galactosamine: polypeptide N-acetylgalactosaminyltransferase 7 (GALNT7)). This result let us accept as good the begin of the first internal predicted by geneid.
From the match we obtain, we can deduce that the first part of this gene is not in our contig, because we have a 100% of homology between all of the aminoacids of our predicted protein and the final part of the homologous one, that is to say, the last 132 aminoacids of the 639 the protein is made of.
We can conclude that it is a real gene which codes for a known human protein but which is not complete in our contig.
