AnnotationRNAseqAssembly

This repository consists of a Snakemake workflow to produce a CDS-annotated genome annotation using Stringtie and TransDecoder. It consists of the following steps:

• Build a STAR aligner index for the genome if it doesn't already exist
• Perform 2-pass STAR alignment of provided paired-end RNA-seq reads to the genome
• Assembles transcripts from those alignments with Stringtie
• Predicts CDS and UTR features for the annotation with TransDecoder
• Adds back into the TransDecoder annotation those features for which open reading framers(ORFs) could not be found
  • TransDecoder normally discards things w/o detectable ORFs which we view as being less than ideal, so we fix that

To run the workflow, you need to create a directory called data directory within the workflow base directory that contains three subdirectories:

• fastq: contains the paired-end RNA-seq fastq files
• genome: contains the genome fasta, and a STAR index if already generated
• blast: this should be a protein blast database that predicted ORFs can be searched against. This is necessary to invoke a TransDecoder feature that, when selecting from a set of candidate ORFs, has TransDecoder prefer an ORF with a blastp hit over one that doesn't.

You then need to:

• edit sampletable.tsv so it specifies your sample ids, and the full path names of the R1 and R2 fastq files for each sample. Note, if you wish to point to RNA-seq fastq files to a different location other than data/fastq, you need to do it in the sample sheet, in which case creating the data/fastq directory will not be necessary.
• edit config/config.yaml, especially if you are already supplying a STAR index--therefore need to indicate its location, and set StarIndexExists to TRUE, but you also need to supply the genome fasta file name (assuming it resides in data/genome), and the prefix of the files in the protein blast database.
• if you are running the workflow on an HPC cluster, you will need to specify slurm partition names in profiles/slurm/config.yaml.

Note that the yaml file in profiles/slurm is a workflow-specific profile that sets resources related to rules that are used to execute particular workflow steps. This is distinct from a global profile, which sets options for the cluster environment in which the workflow is being executed. On a linux system, one's global profile will typically located in your home directory, e.g. $HOME/.config/snakemake. An example global profile for a cluster running SLURM might look like this, and would be called, somewhat confusingly, config.yaml:

executor: slurm
use-conda: True
jobs: 100
latency-wait: 100
retries: 0


default-resources:
    slurm_account: "" # add your account here
    slurm_partition: "" # add your partition list here
    runtime: 360 
    mem_mb: 6000
    threads: 1

If you do not need to modify a global profile from the defaults, then all that is needed is to specify the workflow profile.In this case, you can then execute the Snakemake worflow as follows:

snakemake --snakefile workflow/Snakefile --workflow-profile ./profiles/slurm

If you also needed to specify a particular global profile, if you put the global profile yaml file in a directory called my_global and placed that directory in $HOME/.config/snakemake/, you could then specify both the local and the qworkflow-specific profile like this:

snakemake --snakefile workflow/Snakefile --workflow-profile ./profiles/slurm --profile my_global

For more information on global and workflow profiles consult the Snakemake profiles documentation.

With either of these options, the command can easily be wrapped as a cluster submission job. As it simply manages conda package installs, submission of cluster data analysis jobs, and runs some low-memory serial jobs for file conversions, it can be submitted with one core, and with a modest amount of memory, e.g. a few Gb.