Bioinformatics Data Skills

Assignment 7

You will be creating some SLURM scripts that will run on the UofU CHPC Cluster.

1. Read through Chapter 4 at least twice

2. Watch the first 6 videos in the CHPC series in the course website resources

3. Write and submit the following 3 jobs via SLURM scripts:

Each of these three jobs should be self-contained SLURM scripts that you submit to the “kingspeak-shared” cluster

Job 1: Testing this out

  • Resources = 1 core, 1 minute of wall time, 10 Mb of memory

  • Task:

    cat <(ls -l1 ~)

Job 2: Downloading surface glycoprotein sequences

  • Resources = 1 core, 30 minutes of wall time, 6400 Mb of memory

  • Tasks:

    1. Download and setup the NCBI datasets program we used before
    2. Use datasets to download just the sars-cov-2 genomes deposited since 01/01/2023, extracting the CDS (CoDing Sequences)

    • That command should look like…

      PATH/TO/datasets download virus genome taxon sars-cov2 --include cds --filename sars-cov2.zip --released-after 01/01/2023 --annotated --complete-only
  1. Extract and format the nucleotide sequences for just the “surface glycoprotein” CDS
  2. Calculate the number of sequences you downloaded
  3. Calculate the number of unique nucleotide variants for that gene

Job 3: Clustering those sequences

  • Resources = 12 cores, 4 hours of wall time (to start with, at least…increase if necessary), 28000 Mb of memory

  • Modules = vsearch/040218

  • Tasks:

    1. Run the following program on the cds.fna file of surface glycoproteins (tweak file names/paths if necessary)

           vsearch --cluster_fast surface_glycoproteins.fasta -id 0.97 --centroids centroids.fasta --consout consensus.fasta --msaout msa.fasta --profile cluster_profiles.txt

     cat centroids.fasta | grep "^>" > centroid_ids.txt

     grep -Fwf centroid_ids.txt cds.fna

This vsearch program is clustering the nucleotide sequences based on similarity. In other words, it’s finding groups of nucleotide variants that are at least 97% similar to each other, and then profiling the makeup of these clusters at each nucleotide position. This is useful for several applications… Firstly, if you have 9999 identical sequences and 1 that has a single nucleotide change, that could be real, or it could be a sequencing error. Secondly, it’s performing an alignment step that can be useful for downstream processes. Thirdly, these clusters represent sequence similarity and we can get a quick feel for diversity in this locus, including where the clusters are geographically.

The second 2 lines of that script will return the full sequence names for the representative sequences of each identified cluster. Those full names have location information embedded in them!

Each of these 3 jobs will produce an outfile containing STDOUT.

What to turn in:

  • Your 3 SLURM scripts, concatenated into a single plantext file.
  • Your 3 outfiles, concatenated into a single plantext file.

Upload those plain-text files to Canvas. You’ll be graded on completeness, documentation, and readability.