A Snakemake workflow for running TALON, FLAIR, and pipeline-nanopore-ref-isoforms. Performs a comparative analyses of results using tools such as GFFcompare.

1. Snakemake 7.3.1
   A full snakemake instalation is recommended
2. Singularity 3.7.0

Clone the repository to desired location.

1. Set parameters in config.yaml
2. run: snakemake -p --use-singularity --singularity-prefix "resources" --singularity-args "--bind *" --use-conda -j ** all --configfile "config/config.yaml"

Note * : You should provide your own directory for the --bind command so that the data is accesible from the singularity containers.
Note ** : Specify number of available threads here.

You can run snakemake --report report.html AFTER the workflow finished to create a report containing results.

The GTF files located in the 03_combined and 05_matched_transcripts have a column called TPM. This is actuallu the raw number of counts. The attribute is hijacked to pass counts to GFFCompare.

When testing the workflow it took about 18 hours on 10 threads with 100g memory to process 6 Human samples. Running with a much smaller RNA-virus dataset it took about 8 hours for 6 samples.

The main bottleneck is TranscriptClean which requires many hours and high memory to correct all samples.

Transcriptclean requires the reference genome Fasta file to only have one string per header. In order to run TranscriptClean you must edit the headers.

There seems to be an issue with Snakemake 7.3.1 when building conda environments. If a time out error occurs you can try running the workflow with an older version of snakemake such as version 5.3.2.

MIT, see LICENSE