Figure 1.

Annotation of the protein coding regions of a part of the human ENCODE region ENm007. The line labeled 'VEGA_Known' shows one known gene on the forward strand. The ab initio program AUGUSTUS (labeled 'AUGUSTUS') predicts this gene almost correctly but completely misses the 9th exon annotated around position 318,600. Furthermore, as an ab initio program, AUGUSTUS predicts a false positive gene on the reverse strand around position 310,000. The lines labeled 'hints' show the hints derived from a comparison to the mouse genome. The height of the rectangles depends on their estimated reliability. The hints indicate the presence of an exon where AUGUSTUS missed the annotated exon. Also, there are no hints about coding regions where AUGUSTUS predicted a gene on the reverse strand. When the given hints are used by AUGUSTUS (labeled 'AUGUSTUS+mouse'), the missed exon is correctly predicted and the false positive gene is not predicted anymore. The former is a consequence of the bonus effect and the latter a consequence of the malus effect. Note that the hint about the exon around position 318,600 was helpful, although that exon is more likely to be on the reverse strand according to the hints alone. This plot has been obtained using gff2ps [28].