1
2
3
4
5
6
7
8
9
10
|
On human chromosome 21 genes were predicted with augustus using sampling using
for T in 0 1 2 3 4 5 6 7; do
augustus --alternatives-from-sampling=true --species=human --sample=100 chr21.fa --minexonintronprob=0 --minmeanexonintronprob=0 --temperature=$T
done
The graph shows the posterior probability reported in the 6th column of the augustus output gff file versus the actual fraction of exons (CDS) that
match the reference annotation (RefSeq). The esimated curves for t=5,6,7 are not reliable for high posterior probabilities as they occurr too rarely.
For large temperatures exons with high posterior probability become increasingly rare. We recommend t=3 as it comes relatively close to the ideal curve,
the diagonal, for the most relavant range of posterior probabilities.
|
