UCSC-LoweLab/tRNAviz-data

Data processing pipeline for tRNAviz

tRNAviz-data

Data processing pipeline for tRNAviz

Pipeline overview

0. Prerequisites

   • Python 3: pandas, Biopython
   • R: plyr, dplyr, tidyr, ggplot2, Biostrings, RColorBrewer
   • A list of taxids. If you need to, run parse-genomeinfodb.py to get this list.

1. Run tRNAscan-SE on all genomes.

   • We need .out, .iso, and .ss files, and to run with detailed output.

2. Find taxonomy information for list of species along with relevant filepaths

   • Prerequisites: List of NCBI Taxonomy IDs. Also helpful for this list to contain paths to tRNAscan-SE output. (taxids.tsv)
   • Scripts:
     • get-taxonomy.R (taxids.tsv -> genomes.tsv)
       • Output list of species, plus taxonomic info from NCBI.
     • generate_newick_tree.py (genomes.tsv -> newick-tree.txt)
       • Output a Newick tree that can be used for visualizing phylogenetic trees.

3. Parse tRNAscan-SE output

   • Prerequisites:
     • tRNAscan-SE output files (.out, .ss, .iso) separated into individual folders
   • A table containing paths for .out, .iso, and .ss files. Can be combined with genomes.tsv as additional columns. (genomes.tsv)
     • Covariance model specialized for alignment and numbering (specified by -n in parse-tRNAs.py)
     • Confidence set (if applicable) (confidence-set.txt)
   • Scripts:
     • parse-tRNAs.py: main driver script (genomes.tsv, .out, .iso, .ss, numbering.sto, confidence-set.txt -> trna-df.tsv)
     • tRNA_position.py: helper library that resolves Sprinzl tRNA positions
     • sstofa3 (.ss -> .fa)
       • Removes introns by parsing the .ss file. Note that the default output is the same name as tRNAscan-SE .fa output, which contains introns! Introns are not okay!

4. Resolve consensus features and count base frequencies

   • Prerequisites:
     • Parsed data frame with each position as a column (trnas.tsv)
   • Scripts:
     • resolve-consensus.py (taxonomy.tsv, trnas.tsv -> consensus.tsv)
       • consensus.tsv: For each taxonomic group, lists consensus features for each position and isotype.
     • count-freqs.py (taxonomy.tsv, trnas.tsv -> freqs.tsv)
       • freqs.tsv: For each taxonomic group, lists base and base pair counts by position and isotype.

Tuning the pipeline

Depending on your genome set, you will certainly need to make some changes to your workflow. Here's some examples:

• Adding a couple of lower scoring tRNAs is helpful for identifying minor conserved variants. For non-mammalian vertebrates though, I've locked in the tRNA set to the high confidence set, due to the sheer number of highly amplified tRNA genes.
• Fungi have more diverged tRNAs and score a bit lower using the eukaryotic covariance model, so a score threshold needs to be tuned to maximize the quality of your tRNA set.