PDCUB

A collection of tools for computing position-dependent codon usage bias (PDCUB) scores.

computeCodonStats.py

Performs a position-based computational survey for each non-stop codon given a FASTA file.

Required input:

• GENCODE gene FASTA
• Path to the result directory (must already exist)
• bin size in nucleotides, must be a multiple of 3
• CDS cutoff length in nucleotides (the maximum CDS position to read to)

Example: python computeCodonStats.py gene.fa result_dir 60 600

Script output:

• TSV files where the rows are codons, the columns are the bins, and the values are
  • codon raw counts per bin
  • codon z-scores per bin
  • codon p-values per bin
  • codon z-squared values per bin
  • codon log likelihood weights per bin

computeCodonStats_GC.py

Performs a similar computation as computeCodonStats.py, but computes for two categories:

• G/C nucleotides, or S nucleotides
• A/U nucleotides, or W nucleotides

Inputs are the same as computeCodonStats.py, with the omission of a bin size.

Example: python computeCodonStats_GC.py gene.fa result_dir 600

computeCodonStats_AUGC.py

Performs a similar computation as computeCodonStats.py, but computes for each nucleotide instead of each codon.

Inputs are the same as computeCodonStats.py, with the omission of a bin size.

Example: python computeCodonStats_AUGC.py gene.fa result_dir 600

findAUGScores.py

Finds the highest PDCUB scoring AUG and identifies whether it is the start codon for each transcript in an input FASTA file.

Required input:

• log-likelihood TSV that was computed from computeCodonStats.py, computeCodonStats_GC.py, or computeCodonStats_AUGC.py
• gene FASTA file
• option whether or not to choose a random AUG start codon to compare to the highest scoring AUG
  • enter "nullmodel" to generate a null model
  • enter "realmodel" to generate a PDCUB model
• bin size that the log-likelihood TSV was calculated
• the number of bins that the log-likelihood TSV was calculated with
• the type of model to generate, depending on what computing script was used

• enter "gc" if the log likelihood TSV came from computeCodonStats_GC.py
• enter "augc" if the log likelihood TSV came from computeCodonStats_AUGC.py
• enter "codon" if the log likelihood TSV came from computeCodonStats_GC.py

example: python findAUGScores.py log_gc.tsv gene.fa realmodel 3 100 gc

Script Output:

• TSV file with the following columns
• Transcript ID
• Gene ID
• Start Codon position
• 5' UTR position
• Whether or not the start codon was the highest scoring AUG

• 1 if the start codon was the highest scoring AUG
• 0 if the start codon was not the highest scoring AUG

• Transcript Length in nucleotides
• the score of the start codon

createHeatMap.py

Creates a PDF of a heatmap that is based on one of the computational TSV outputs from one of the computing scripts. In the paper, this script is used to visualize the z-scores of each non-stop codon across the CDS.

Required input:

• input TSV file, where the rows are codons, the columns are bin ranges, and the values are numeric in nature
• name of the output file
• name for the colormap used for the heatmap [colormap options] (https://matplotlib.org/stable/tutorials/colors/colormaps.html)
• The title of the color bar
• Organizational scheme for the heatmap

• 0: the codons are organized by amino acid and chemical group
• 1: the codons are organized by chemical group, amino acid, and the first column in increasing order
• 2: the codons are organized by the difference between the first and second column in increasing order

Example: python createHeatMap.py z-scores.tsv zScoreHeatMap.pdf plasma z-scores 0

createDendrogram.py

Creates a PDF of a heatmap where the codons are organized according to a dendrogram.

The input is the same as createHeatMap.py, but there is no option for the organization scheme of the heatmap.

Example: python createDendrogram.py z-scores.tsv zScoreDendrogram.py Reds z-scores

plotTai_quintiles.py

This python script reads in five evenly sized fasta files, each comprising one quintile of the GENCODE protein-coding human transcriptome sorted by PDCUB score. The output is a combined plot of the average local tAI scores for each quintile. These averages are calculated using a sliding window, the size of which is entered from the command line when executing the script.

plotSignalPeptideVenn.py

This script takes three inputs: a list of SignalP prediction results, a list of PrediSi prediction results, and the list of human protein-coding GENCODE transcripts with PDCUB scores considered significant. The output is a Venn diagram of all three inputs.

plotTaiVsPdcub_final.py

Python script that takes in a fasta file, a PDCUB score file, and a CDS window size W in nucleotides. The script computes local tAI for the first W nucleotides and the second W nucleotides. There are two outputs: A scatterplot of second-window local tAI against first-window local tAI for each transcript, and a scatterplot of first-plus-second-window local tAI against PDCUB score for each transcript.

plotCAI_localRA_final.py

Python script that takes in a fasta file, a PDCUB score file, and a CDS window size W in nucleotides. The script computes local CAI for the first W nucleotides and the second W nucleotides. The output is a scatterplot of second-window local CAI against first-window local CAI for each transcript.

plot_avg_ribosome_profile_quintiles_a_site.py

Python script that reads in five fasta files (quintiles) and a BAM file of ribo-seq reads both derived from the GENCODE protein-coding human transcriptome. This script plots the normalized ribosome profile of each quintile.