Skip to content

Welcome to our weekly research support coffee hour on Zoom! Click here for more information.

AUGUSTUS

AUGUSTUS is a program that predicts genes in eukaryotic genomic sequences. It can be used as an ab initio predictor and can also incorporate external evidence such as RNA-Seq alignments and protein homology.

License

Free to use and open source under the Artistic License 1.0.

Available

  • Roihu: 3.5.0, via the bio-apps module.

Usage

AUGUSTUS is part of the bio-apps collection on Roihu. Load the bio-apps module tree and then the AUGUSTUS module:

module load bio-apps/v202603
module load augustus/3.5.0

To predict genes using an existing species model:

augustus --species=human genome.fa > predictions.gff

Configuration directory

The module sets the AUGUSTUS_CONFIG_PATH environment variable, pointing to the configuration directory (species models, parameters) in the module installation. This installation is read-only, which is fine for running predictions with existing species models.

If you need to train a new species (or otherwise write to the configuration directory, for example with etraining or the autoAug scripts), copy the configuration to a writable location and point AUGUSTUS_CONFIG_PATH there:

cp -r $AUGUSTUS_CONFIG_PATH /scratch/<project>/augustus_config
export AUGUSTUS_CONFIG_PATH=/scratch/<project>/augustus_config

Example batch script

#!/bin/bash
#SBATCH --job-name=augustus
#SBATCH --account=<project>
#SBATCH --output=output_%j.txt
#SBATCH --error=errors_%j.txt
#SBATCH --partition=small
#SBATCH --time=08:00:00
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=1
#SBATCH --mem=8G

module load bio-apps/v202603
module load augustus/3.5.0

augustus --species=human genome.fa > predictions.gff

Replace <project> with your CSC project (for example project_2001234).

See creating a batch job script for Roihu for more information about running batch jobs.

Support

CSC Service Desk

More information

Do you find this page useful?
Thank you for your feedback!
Thank you! Your feedback is appreciated.

Please let us know how we could improve this page by contacting us or by creating an issue in GitHub. You can also edit this page yourself and contribute your improvements by creating a pull request.