cgeb2001's DokuWiki!

Augustus is an ab initio gene predictor that employs Hidden Markov Models (HMMs) pre-trained on existing datasets. Training a custom HMM model with your own data is possible, but in this basic tutorial, we are only going to consider pre-existing models.

source activate augustus-3.5.0

# only required for species=generic
export AUGUSTUS_CONFIG_PATH="/misc/scratch3/jasons/protist_gene_prediction/software/custom_augustus_config"

augustus \
    --species=generic \
    <your genome> \
    --gff3=on \
    --outfile=<outfile name>.gff3

conda deactivate

Note that the species here is set to generic to minimize biases for a divergent organism. If your organism is closely related to one of the pre-trained species below, you can specify that instead to yield a better prediction.

If you decide to use a pre-trained species, then you don't have to include the export line. The reason for including the export line is because augustus had stopped shipping probability files with generic species around 3.3.3. Why have they done that, you might ask? Well, unfortunately it's another great mystery for which science cannot explain.