VIGOR, an annotation program for small viral genomes

Wang, Shiliang; Sundaram, Jaideep P; Spiro, David

Two ab initio gene prediction programs, ZCURVE_-CoV [28] and GeneDecipher [29] , were specifically trained and the program parameters were adjusted for SARS coronavirus genomes. Both programs can correctly predict the major large protein coding genes and structural protein coding genes like polyprotein orf1a, orf1ab, spike gene, nucleocapsid gene, envelope gene and membrane protein gene. However, short peptide genes and embedded genes were often missing on the predicted gene list (false negative) [29] , although the exact function of these small peptide genes is unknown. Mature peptide prediction is not a designed function for these two programs. 17 of the 18 SARS-CoV genomes tested by GeneDecipher were used to evaluate VIGOR predictions. Since there were no annotations for most of these genomes in GenBank, SARS-CoV genome TOR2 genome was annotated in GenBank and predictions were listed [29] , a detail comparison was done for this genome (data not shown). The predictions of VIGOR for SARS-CoV genome TOR2 were exactly same as the GenBank annotations. GeneDeciper didn't pick 6 small non-structural protein genes and predicted one gene incorrectly. The genome structure and genes of the other 16 tested SARS-CoV genomes are same as SARS-CoV genome TOR 2, 14 genes and one non-functional non-structural protein gene were detected in these genomes by VIGOR.