This repository contains a Nextflow pipeline that allows eukaryotic contigs to be isolated from assemblies.
The current prerequisites are Singularity and the tool images of Tiara, DeepMicroClass and CAT.
To easily test this pipeline, you can use the predefinied test use case :
nextflow run main.nf -profile test,ebi_slurm
For practical cases, you can run its current version like this :
nextflow run main.nf -profile ebi_slurm --contigsPath {pathToFasta}
For non-EBI users, note that you'll need to create your own profile and associated config file in order to precise the container paths and versions.
Note that you can specify the DeepMicroClass parameters by adding these when running the nextflow script :
nextflow run main.nf -profile ebi_slurm --contigsPath {pathToFasta} --model {path} --encoding {onehot,embedding} --mode {hybrid,single} --singleLen {n} --device {cpu,cuda}
By default, the model contained in the Singularity image deepmicroclass.sif is used.
In addition, although it is possible to specify the use of CUDA, please first check that the used image corresponds to the GPU version of DeepMicroClass.
This pipeline use the following decision rules to classify the contigs :
The final fasta other_kingdoms.fa includes the contigs not classified as eukaryotes by CAT and the contigs not classified as eukaryotes during the first classification stage.
